Machine learning methods for cytometry

Abstract

A technique for determining a respective cell type for each of at least a portion of a plurality of cells, the technique including: obtaining cytometry data for a biological sample from a subject, the biological sample including a plurality of cells including a first cell, the cytometry data including first cytometry data for the first cell; and determining a respective type for each of at least a portion of the plurality of cells using a machine learning model hierarchy corresponding to the cell type hierarchy, the determining including determining the first type for the first cell by processing the first cytometry data with a first subset of the machine learning model hierarchy.

Description

[Technical field]

[0001] Related Applications This application claims the benefit under 35 U.S.C. Section 119(e) of the filing date of U.S. Provisional Patent Application No. 63 / 304,990, entitled "MACHINE LEARNING TECHNIQUES FOR CYTOMETRY," filed on January 31, 2022 (Attorney Docket No. B1462.70030US00), the entire contents of which are incorporated herein by reference. [Background technology]

[0002] Cytometry is a laboratory technique used to analyze single cells or particles in biological samples. Cytometry is used in a variety of applications, including immunology and molecular biology. Cytometry can be used to measure the characteristics of individual cells or particles. There are two types of cytometry: flow cytometry and mass cytometry.

[0003] Flow cytometry measures the intensity produced by fluorescently labeled markers used to label cells in a biological sample. For example, cells labeled with a marker, or a particular combination of markers, can be processed by a flow cytometry platform that measures the fluorescence intensity of the markers on a cell-by-cell basis. The measured intensity, or "marker value," of such markers can then be used to determine the type of each cell.

[0004] Mass cytometry measures the intensity of heavy metal ion tags used to label cells in biological samples. For example, cells labeled with a marker, or a particular combination of markers, can be processed by a mass cytometry platform that measures the relative intensity or abundance of the markers on a cell-by-cell basis. The intensity of such markers, or "marker value," can then be used to determine the type of each cell.

[0005] When obtaining cytometry data for a biological sample, the biological sample may be divided into multiple sub-samples. Each sub-sample may be processed using a different "panel" of markers. A panel of markers is a set of markers used to label cells in the biological sample or sub-sample. Obtaining cytometry data using different panels of markers allows for the identification of different cell types in the biological sample, since different markers bind to different cell types or subtypes. Summary of the Invention [Means for solving the problem]

[0006] Some embodiments provide a method that includes performing, using at least one computer hardware processor, obtaining flow cytometry data for a biological sample from a subject, the biological sample comprising a plurality of cells including a first cell, and the flow cytometry data including first flow cytometry data for the first cell; and determining a respective type for each of at least a portion of the plurality of cells using a machine learning model hierarchy corresponding to the cell type hierarchy, where determining includes determining the first type for the first cell by processing the first flow cytometry data using a first subset of the machine learning model hierarchy.

[0007] Some embodiments provide a system including at least one computer hardware processor; and at least one non-transitory computer readable storage medium storing processor-executable instructions that, when executed by the at least one computer hardware processor, cause the at least one computer hardware processor to: obtain flow cytometry data for a biological sample from a subject, the biological sample including a plurality of cells including a first cell, and the flow cytometry data including first flow cytometry data for the first cell; and determine a respective type for each of at least a portion of the plurality of cells using a machine learning model hierarchy corresponding to the cell type hierarchy, wherein determining includes determining the first type for the first cell by processing the first flow cytometry data using a first subset of the machine learning model hierarchy.

[0008] Some embodiments provide at least one non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by at least one computer hardware processor, cause at least one computer hardware processor to perform the following: obtaining flow cytometry data for a biological sample from a subject, the biological sample comprising a plurality of cells including a first cell, and the flow cytometry data including first flow cytometry data for the first cell; and determining a respective type for each of at least a portion of the plurality of cells using a machine learning model hierarchy corresponding to the cell type hierarchy, wherein determining includes determining the first type for the first cell by processing the first flow cytometry data using a first subset of the machine learning model hierarchy.

[0009] In some embodiments, the plurality of cells includes a second cell of a different type than the first cell. In some embodiments, the flow cytometry data includes second flow cytometry data for the second cell. In some embodiments, the determining further includes processing the second flow cytometry data using a second subset of the machine learning model hierarchy, the second subset of the machine learning model hierarchy being different from the first subset of the machine learning model hierarchy.

[0010] In some embodiments, the determining is performed on each of at least 10,000 cells from among the plurality of cells.

[0011] In some embodiments, processing the first flow cytometry data using the first subset of the machine learning model hierarchy includes processing the first flow cytometry data using a first machine learning model in the first subset of the machine learning model hierarchy; identifying a second machine learning model in the first subset of the machine learning model hierarchy based on a first output of the first machine learning model; and obtaining a second output by processing the first flow cytometry data using the second machine learning model.

[0012] In some embodiments, the first output is indicative of a first type for the first cell. In some embodiments, the second output is indicative of a subtype of the first type for the first cell.

[0013] In some embodiments, the first type comprises a white blood cell and a subtype of the first type comprises a granulocyte.

[0014] In some embodiments, the first type comprises lymphocytes and a subtype of the first type comprises B cells.

[0015] In some embodiments, the first type comprises T helper cells and a subtype of the first type comprises memory T helper cells.

[0016] In some embodiments, processing the first flow cytometry data using the first subset of the machine learning model hierarchy further includes identifying a third machine learning model in the first subset of the machine learning model hierarchy based on the first output of the first machine learning model; and processing the first flow cytometry data using the third machine learning model.

[0017] In some embodiments, processing the first flow cytometry data using the first subset of the machine learning model hierarchy further includes identifying a third machine learning model in the first subset of the machine learning model hierarchy based on a second output of the second machine learning model; and processing the first flow cytometry data using the third machine learning model.

[0018] In some embodiments, the machine learning models in the machine learning model hierarchy include a decision tree classifier, a gradient boosting decision tree classifier, or a neural network.

[0019] In some embodiments, the machine learning models in the machine learning model hierarchy include an ensemble of machine learning models.

[0020] Some embodiments further include determining a respective cellular composition percentage for each of at least some of the determined cell types, where determining includes determining a first cellular composition percentage for a first cell type.

[0021] In some embodiments, determining the first cellular composition percentage comprises determining a ratio between a number of cells of a first type and a total cell number of at least a portion of the plurality of cells.

[0022] Some embodiments further include comparing the first cellular composition percentage to a cellular composition percentage range associated with a patient cohort; and identifying the subject as a member of the patient cohort based on a result of the comparison.

[0023] In some embodiments, the patient cohorts include a healthy cohort, a cohort of patients with a disease, and / or a cohort of patients who have received treatment.

[0024] Some embodiments further include generating a visualization of the determined cellular composition percentage, the visualization indicating a comparison of the first cellular composition percentage to a cellular composition percentage range associated with the patient cohort.

[0025] Some embodiments further include comparing the first cellular composition percentage to a cellular composition percentage range associated with a study, the study evaluating the effectiveness of one or more treatments in treating the disease; and identifying a treatment for the subject based on the results of the comparison.

[0026] Some embodiments further include using the cell type hierarchy to generate a visualization of the determined cellular composition percentages.

[0027] In some embodiments, the visualization includes an indication of a cell type subset of a cell type hierarchy, hi some embodiments, the determined cell type subset percentages include aberrant cell composition percentages compared to a reference cell composition percentage.

[0028] In some embodiments, the visualization includes a plurality of nodes organized in a hierarchy with at least some pairs of nodes connected by respective edges, in some embodiments, the plurality of nodes includes a first node representing a first cell type and a second node representing a subtype of the first cell type, and the edges include a first edge connecting the first node and the second node.

[0029] In some embodiments, the visualization further includes, for the first cell type, a first number indicating a percentage of cells in the biological sample having the first cell type, the first number shown proximate to the first node in the visualization.

[0030] In some embodiments, determining each cellular composition percentage further comprises determining one or more second cellular composition percentages for one or more second cell types, in some embodiments, the one or more second cell types are subtypes of the first cell type.

[0031] Some embodiments further include comparing a sum of the one or more second cellular composition percentages to the first cellular composition percentage; determining a normalization factor based on a result of the comparison; and applying the normalization factor to the one or more second cellular composition percentages.

[0032] Some embodiments further include determining whether the one or more second cell types include all subtypes of the first cell type; and when the one or more second cell types include all subtypes of the first cell type, applying a normalization factor to the one or more second cellular composition percentages.

[0033] In some embodiments, the normalization factor comprises a ratio between the first cellular composition percentage and the sum of one or more second cellular composition percentages.

[0034] Some embodiments further include comparing a sum of the one or more second cellular composition percentages to the first cellular composition percentage; determining a type-specific normalization factor for each of the one or more second cell types based on a result of the comparison; and applying the type-specific normalization factor to each of the one or more cellular composition percentages.

[0035] In some embodiments, the flow cytometry data obtained includes noise transformed flow cytometry data.

[0036] In some embodiments, the flow cytometry data includes a fluorescence intensity value for each of at least a portion of a plurality of markers for each of at least a portion of the plurality of cells.

[0037] In some embodiments, the plurality of cells includes a first plurality of cells and a second plurality of cells. In some embodiments, the flow cytometry data includes a first flow cytometry data subset for the first plurality of cells and a second flow cytometry data subset for the second plurality of cells, the first flow cytometry data subset includes fluorescence intensity values ​​for a first subset of the plurality of markers for each of at least a portion of the first plurality of cells, and the second flow cytometry data subset includes fluorescence intensity values ​​for a second subset of the plurality of markers for each of at least a portion of the second plurality of cells. In some embodiments, the first subset of the plurality of markers and the second subset of the plurality of markers are different.

[0038] In some embodiments, the first flow cytometry data subset comprises data from a first panel. In some embodiments, the second flow cytometry data subset comprises data from a second panel.

[0039] In some embodiments, determining a respective type for each of at least a portion of the plurality of cells comprises determining a respective first plurality of types for a first plurality of cells, and determining a respective second plurality of types for a second plurality of cells. Some embodiments further include determining a first plurality of cellular composition percentages using the first plurality of cell types, where determining comprises determining a respective cellular composition percentage for each of at least a portion of the first plurality of cell types; and determining a second plurality of cellular composition percentages using the second plurality of cell types, where determining comprises determining a respective cellular composition percentage for each of at least a portion of the second plurality of cell types.

[0040] Some embodiments further include determining a respective cellular composition percentage for each of at least a portion of the determined cell types, where determining includes combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages.

[0041] Some embodiments further include combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages based on estimated composition percentages of cell types included in the first plurality of cell types and the second plurality of cell types.

[0042] Some embodiments further include combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages based on data obtained using beads contained in the biological sample.

[0043] In some embodiments, the biological sample includes a plurality of particles. In some embodiments, the flow cytometry data includes flow cytometry data for a first particle of the plurality of particles, where the first particle is a debris, a cell doublet, or a bead.

[0044] Some embodiments further include determining a respective particle type for each of at least a portion of the plurality of particles using the machine learning model hierarchy, where determining includes determining a first particle type for a first particle using the machine learning model hierarchy.

[0045] Some embodiments further include determining whether the first particle comprises a bead, debris, a cell doublet, or the first particle is unidentifiable.

[0046] In some embodiments, obtaining the flow cytometry data comprises processing the biological sample using a flow cytometry platform.

[0047] In some embodiments, the machine learning model hierarchy includes at least 250 machine learning models.

[0048] In some embodiments, the machine learning model hierarchy includes at least 50 machine learning models.

[0049] In some embodiments, the first machine learning model includes 20 hyperparameters.

[0050] Some embodiments provide a method that includes performing, using at least one computer hardware processor, obtaining flow cytometry data for a biological sample from a subject, the biological sample including a plurality of cells including a first cell and a second cell of a different type than the first cell, the flow cytometry data including first flow cytometry data for the first cell and second flow cytometry data for the second cell; determining a respective type for each of at least a portion of the plurality of cells using a plurality of machine learning models, the determining including determining a first type for the first cell by processing the first flow cytometry data using a first subset of the plurality of machine learning models, the first subset including the first machine learning model; and determining a second type for the second cell by processing the second flow cytometry data using a second subset of the plurality of machine learning models, the second subset including a second machine learning model that is not in the first subset of the plurality of machine learning models.

[0051] In some embodiments, determining the respective type for each of at least a portion of the plurality of cells is performed for each of at least 10,000 cells from among the plurality of cells.

[0052] In some embodiments, the first machine learning model is the first machine learning model in the first subset.In some embodiments, determining the first type of the first cell comprises: using the first machine learning model to process the first flow cytometry data; identifying a second machine learning model in the first subset of the multiple machine learning models based on the output of the first machine learning model; and using the second machine learning model in the first subset to process the first flow cytometry data to obtain a second output.

[0053] In some embodiments, the first output is indicative of a type for the first cell. In some embodiments, the second output is indicative of a subtype of the type for the first cell.

[0054] In some embodiments, the type for the first cell comprises a leukocyte and a subtype of the type comprises a granulocyte.

[0055] In some embodiments, the type for the first cell comprises a lymphocyte and a subtype of the type comprises a B cell.

[0056] In some embodiments, the first type comprises T helper cells and a subtype of the type comprises memory T helper cells.

[0057] In some embodiments, the machine learning models in the machine learning model hierarchy include a decision tree classifier, a gradient boosting decision tree classifier, or a neural network.

[0058] In some embodiments, the machine learning models in the machine learning model hierarchy include an ensemble of machine learning models.

[0059] Some embodiments further include determining a respective cellular composition percentage for each of at least some of the determined cell types, where determining includes determining a first cellular composition percentage for a first cell type.

[0060] In some embodiments, determining the first cellular composition percentage comprises determining a ratio between a number of cells of the first type and a total cell number of at least a portion of the plurality of cells.

[0061] Some embodiments further include comparing the first cellular composition percentage to a cellular composition percentage range associated with a patient cohort; and identifying the subject as a member of the patient cohort based on a result of the comparison.

[0062] In some embodiments, the patient cohorts include a healthy cohort, a cohort of patients with a disease, and / or a cohort of patients who have received treatment.

[0063] Some embodiments further include generating a visualization of the determined cellular composition percentage, the visualization indicating a comparison of the first cellular composition percentage to a cellular composition percentage range associated with the patient cohort.

[0064] Some embodiments further include comparing the first cellular composition percentage to a cellular composition percentage range associated with a study, the study evaluating the effectiveness of one or more treatments in treating the disease; and identifying a treatment for the subject based on the results of the comparison.

[0065] Some embodiments further include using the cell type hierarchy to generate a visualization of the determined cellular composition percentages.

[0066] In some embodiments, the visualization includes an indication of a cell type subset of a cell type hierarchy, hi some embodiments, the determined cell type subset percentages include aberrant cell composition percentages compared to a reference cell composition percentage.

[0067] In some embodiments, the visualization includes a plurality of nodes organized in a hierarchy with at least some pairs of nodes connected by respective edges, in some embodiments, the plurality of nodes includes a first node representing a first cell type and a second node representing a subtype of the first cell type, and the edges include a first edge connecting the first node and the second node.

[0068] In some embodiments, the visualization further includes, for the first cell type, a first number indicating a percentage of cells in the biological sample having the first cell type, the first number shown proximate to the first node in the visualization.

[0069] In some embodiments, the determining further comprises determining one or more cell subtype composition percentages for one or more cell subtypes of the first cell type.

[0070] Some embodiments further include comparing a sum of the one or more cell subtype composition percentages to the first cell composition percentage; determining a normalization factor based on a result of the comparison; and applying the normalization factor to the one or more cell subtype composition percentages.

[0071] Some embodiments further include determining whether the one or more cell subtypes include all subtypes of the first cell type; and applying a normalization factor to the one or more cell subtype composition percentages when the one or more cell subtypes include all subtypes of the first cell type.

[0072] In some embodiments, the normalization factor comprises a ratio between the first cell composition percentage and the sum of the one or more cell subtype composition percentages.

[0073] Some embodiments further include comparing a sum of the one or more cell subtype composition percentages to the first cell subtype composition percentage; determining a type-specific normalization factor for each of the one or more cell subtypes based on a result of the comparison; and applying the type-specific normalization factor to each of the one or more cell subtype composition percentages.

[0074] In some embodiments, the flow cytometry data obtained includes noise transformed flow cytometry data.

[0075] In some embodiments, the flow cytometry data includes a fluorescence intensity value for each of at least a portion of a plurality of markers for each of at least a portion of the plurality of cells.

[0076] In some embodiments, the plurality of cells includes a first plurality of cells and a second plurality of cells. In some embodiments, the flow cytometry data includes a first flow cytometry data subset for the first plurality of cells and a second flow cytometry data subset for the second plurality of cells, the first flow cytometry data subset includes fluorescence intensity values ​​for a first subset of the plurality of markers for each of at least a portion of the first plurality of cells, and the second flow cytometry data subset includes fluorescence intensity values ​​for a second subset of the plurality of markers for each of at least a portion of the second plurality of cells. In some embodiments, the first subset of the plurality of markers and the second subset of the plurality of markers are different.

[0077] In some embodiments, the first flow cytometry data subset comprises data from a first panel. In some embodiments, the second flow cytometry data subset comprises data from a second panel.

[0078] In some embodiments, the determining comprises determining a first plurality of respective types for the first plurality of cells and determining a second plurality of respective types for the second plurality of cells. Some embodiments further comprise: determining a first plurality of cellular composition percentages using the first plurality of cell types, where determining comprises determining a respective cellular composition percentage for each of at least a portion of the first plurality of cell types; and determining a second plurality of cellular composition percentages using the second plurality of cell types, where determining comprises determining a respective cellular composition percentage for each of at least a portion of the second plurality of cell types.

[0079] Some embodiments further include determining a respective cellular composition percentage for each of at least a portion of the determined cell types, where determining includes combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages.

[0080] Some embodiments further include combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages based on estimated composition percentages of cell types included in the first plurality of cell types and the second plurality of cell types.

[0081] Some embodiments further include combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages based on data obtained using beads contained in the biological sample.

[0082] In some embodiments, the biological sample includes a plurality of particles. In some embodiments, the flow cytometry data includes flow cytometry data for a first particle of the plurality of particles, where the first particle is a debris, a cell doublet, or a bead.

[0083] Some embodiments further include determining a respective particle type for at least a portion of the plurality of particles using a plurality of machine learning models, where determining includes determining a first particle type for a first particle using the plurality of machine learning models.

[0084] Some embodiments further include determining whether the first particle comprises a bead, debris, a cell doublet, or the first particle is unidentifiable.

[0085] In some embodiments, obtaining the flow cytometry data comprises processing the biological sample using a flow cytometry platform.

[0086] In some embodiments, the machine learning model hierarchy includes at least 250 machine learning models.

[0087] In some embodiments, the machine learning model hierarchy includes at least 50 machine learning models.

[0088] In some embodiments, the first machine learning model includes 20 hyperparameters.

[0089] Some embodiments include a method for obtaining, using at least one computer hardware processor, cytometry data for a biological sample previously obtained from a subject, the biological sample comprising a plurality of cells, the cytometry data including cytometry measurements obtained during respective cytometry events, the cytometry events corresponding to particular objects in the biological sample under measurement by the cytometry platform, the cytometry events including an event subset corresponding to cells in the biological sample under measurement by the cytometry platform; obtaining a respective plurality of cell types by identifying types of cells in the plurality of cells using a plurality of machine learning models, the plurality of machine learning models including a first machine learning model and a second machine learning model different from the first machine learning model, the identifying including identifying a type of each particular event in the event subset. obtaining, for each event, from the cytometry data, a cytometry measurement value corresponding to the particular event; determining an event type for the particular event by processing the cytometry measurement value corresponding to the particular event with a first machine learning model, where the event type indicates whether the particular event corresponds to a cell under measurement by the cytometry platform, debris under measurement by the cytometry platform, or beads under measurement by the cytometry platform; and when the determined event type indicates that the particular event corresponds to a cell under measurement by the cytometry platform, processing the cytometry measurement value corresponding to the particular event with a second machine learning model to determine a type of cell.

[0090] Some embodiments, when executed by at least one computer hardware processor, include obtaining cytometry data for a biological sample previously obtained from a subject, the biological sample comprising a plurality of cells, the cytometry data including cytometry measurements obtained during respective cytometry events, the cytometry events corresponding to particular objects in the biological sample under measurement by the cytometry platform, the cytometry events including an event subset corresponding to cells in the biological sample under measurement by the cytometry platform; obtaining a respective plurality of cell types by identifying types of cells in the plurality of cells using a plurality of machine learning models, the plurality of machine learning models including a first machine learning model and a second machine learning model different from the first machine learning model, wherein the identifying includes, for each particular event in the event subset, obtaining from the cytometry data a cytometry measurement corresponding to the particular event. obtaining cytometry measurements; determining an event type for the particular event by processing the cytometry measurements corresponding to the particular event using a first machine learning model, where the event type indicates whether the particular event corresponds to a cell under measurement by the cytometry platform, debris under measurement by the cytometry platform, or beads under measurement by the cytometry platform; and when the determined event type indicates that the particular event corresponds to a cell under measurement by the cytometry platform, determining a type of cell by processing the cytometry measurements corresponding to the particular event using a second machine learning model.

[0091] Some embodiments include a method for identifying a cell type in the plurality of cells by using a plurality of machine learning models to obtain a respective plurality of cell types, the method comprising: at least one computer hardware processor; and, upon execution by the at least one computer hardware processor, obtaining cytometry data for a biological sample previously obtained from a subject, the biological sample comprising a plurality of cells, the cytometry data including cytometry measurements obtained during respective cytometry events, the cytometry events corresponding to particular objects in the biological sample under measurement by the cytometry platform, the cytometry events including an event subset corresponding to cells in the biological sample under measurement by the cytometry platform; obtaining a respective plurality of cell types by identifying a type of cell in the plurality of cells using a plurality of machine learning models, the plurality of machine learning models including a first machine learning model and a second machine learning model different from the first machine learning model, the identifying including, for each particular event in the event subset, obtaining a particular cell type from the cytometry data. and at least one non-transitory computer-readable storage medium having processor-executable instructions stored thereon to cause at least one computer hardware processor to perform a method comprising: obtaining cytometry measurements corresponding to an event; determining an event type for the particular event by processing the cytometry measurements corresponding to the particular event using a first machine learning model, where the event type indicates whether the particular event corresponds to a cell under measurement by the cytometry platform, debris under measurement by the cytometry platform, or beads under measurement by the cytometry platform; and when the determined event type indicates that the particular event corresponds to a cell under measurement by the cytometry platform, determining a type of cell by processing the cytometry measurements corresponding to the particular event using a second machine learning model.

[0092] In some embodiments, the event subset includes at least 10,000 events.

[0093] In some embodiments, the event subset includes at least 100,000 events.

[0094] In some embodiments, the first machine learning model includes a first multi-class classifier and the second machine learning model includes a second multi-class classifier.

[0095] In some embodiments, the first machine learning model includes a first decision tree classifier, a first gradient boosting decision tree classifier, or a first neural network, and the second machine learning model includes a second decision tree classifier, a second gradient boosting decision tree classifier, or a second neural network.

[0096] Some embodiments further include determining a percentage cellular composition of different cell types in the biological sample based on the identified plurality of cell types.

[0097] In some embodiments, determining the cellular composition percentage comprises determining a first cellular composition percentage for the first cell type by determining a ratio between a number of cells identified as being of the first type in the plurality of cells and a total number of cells in the plurality of cells.

[0098] In some embodiments, the subject has, is suspected of having, or is at risk of having cancer, and the method further comprises identifying a treatment for the subject based on the determined cellular composition percentage.

[0099] Some embodiments further include administering the identified treatment to the subject.

[0100] In some embodiments, identifying a treatment for the subject based on the determined cellular composition percentage includes identifying ipilimumab for the subject when the cellular composition percentage of peripheral blood mononuclear cells (PBMCs) is below a threshold value.

[0101] In some embodiments, identifying a treatment for the subject based on the determined cellular percentage comprises determining a ratio between a cellular percentage of CD8+PD-1+ cells and a cellular percentage of CD4+PD-1; and identifying an immune checkpoint blockade therapy for the subject when the determined ratio is above a threshold.

[0102] Some embodiments further include comparing a cellular composition percentage among the determined cellular composition percentages to a cellular composition percentage range associated with a patient cohort; and identifying the subject as a member of the patient cohort based on a result of the comparison.

[0103] In some embodiments, the patient cohort comprises a healthy cohort, a cohort of patients with a disease, or a cohort of patients who have undergone treatment.

[0104] Some embodiments further include comparing a cellular composition percentage among the determined cellular composition percentages to a cellular composition percentage range associated with a study, the study evaluating the effectiveness of one or more treatments in treating the disease; and identifying a treatment for the subject based on the results of the comparison.

[0105] In some embodiments, the event subset corresponding to cells in the biological sample under measurement by the cytometry platform comprises a first event subset, and the cytometry events further comprise a second event subset corresponding to beads in the biological sample under measurement by the cytometry platform, and a third event subset corresponding to debris in the biological sample under measurement by the cytometry platform.

[0106] In some embodiments, the cytometric measurements corresponding to a particular event include fluorescence intensity values ​​for at least a portion of the plurality of markers.

[0107] In some embodiments, the plurality of events includes a first plurality of events and a second plurality of events, and the cytometry data includes first cytometry data for the first plurality of events and second cytometry data for the second plurality of events, the first cytometry data includes measurements obtained for a first marker of the plurality of markers during each of at least a portion of the first plurality of events, and the second cytometry data includes measurements obtained for a second marker of the plurality of markers during each of at least a portion of the second plurality of events, where the first marker of the plurality of markers and the second marker of the plurality of markers are different.

[0108] In some embodiments, the first cytometry data includes data from a first panel, and the second cytometry data includes data from a second panel that is different from the first panel.

[0109] In some embodiments, obtaining cytometry data for the biological sample includes obtaining flow cytometry data for the biological sample, and the cytometry measurements obtained during each cytometry event include flow cytometry measurements obtained during each flow cytometry event. [Brief description of the drawings]

[0110] [Figure 1A] FIG. 1 depicts an exemplary technique for determining respective types for one or more events based on cytometry data according to some embodiments of the technology described herein. [Figure 1B] 1 is a table showing example cytometry data for multiple events corresponding to cells and / or particles in a biological sample according to some embodiments of the technology described herein. [Figure 1C] 1 is a table showing example cytometry data including example marker values ​​for multiple events according to some embodiments of the technology described herein. [Figure 1D] 1 is a block diagram of a system 150 including an example of a computing device 108 and software 112 according to some embodiments of the technology described herein. [Figure 1E] FIG. 1 is a block diagram of a system 180 for determining one or more types for a cell by processing cytometry data 132-1 for the cell, according to some embodiments of the technology described herein. [Figure 2A] 2 is a flowchart of an exemplary process 200 for determining the respective type of one or more cells using cytometry data and a machine learning model hierarchy according to some embodiments of the technology described herein. [Figure 2B] 1 is a flowchart illustrating an example of a processing implementation 206a of a process 200 for determining a first type for a first cell, according to some embodiments of the technology described herein. [Figure 2C] 2 is a flow chart of an exemplary process 250 for determining a type for an event corresponding to an entity in a biological sample under measurement by a cytometry platform according to some embodiments of the technology described herein. [Figure 3A] 1 is an example diagram of determining a type of event based on the output of a binary class classifier, according to some embodiments of the technology described herein. [Figure 3B] 1 is an example diagram of determining a type of event based on the output of a multi-class classifier, according to some embodiments of the technology described herein; [Figure 4] 4 depicts an example hierarchy 400 of cell / particle types and shows an illustrative example of determining one or more types for cells / particles based on the hierarchy 400, according to some embodiments of the technology described herein. [Figure 5A] 5 is a flowchart of an exemplary process 500 for identifying a subject as a member of a patient cohort according to some embodiments of the technology described herein. [Figure 5B]1 is a flowchart illustrating an example implementation of operation 506 of process 500 for determining a cellular composition percentage based on an estimated cellular composition percentage of common cell types, according to some embodiments of the technology described herein. [Figure 5C] 1 is a flowchart illustrating an example implementation of operation 506 of process 500 for determining cellular composition percentage based on bead percentage according to some embodiments of the technology described herein. [Figure 5D] 1 is a flowchart illustrating a processing implementation 508 of the process 500 for normalizing cellular composition percentages with respect to hierarchical relationships between cell types according to some embodiments of the technology described herein. [Figure 6A] 1 depicts an illustrative example of determining a cellular composition percentage based on cell type determined using cytometry from a single panel according to some embodiments of the technology described herein. [Figure 6B] 1 depicts an illustrative example of determining a cellular composition percentage based on cell type determined using cytometry data from different panels according to some embodiments of the technology described herein. [Figure 6C] FIG. 1 depicts an illustrative example of determining a cellular composition percentage based on cell types determined using cytometry data from different panels according to some embodiments of the technology described herein. [Figure 7A] 1 is an example of a hierarchical visualization of cellular composition percentages according to some embodiments of the technology described herein. [Figure 7B] 1 is an example of a hierarchical visualization of cellular composition percentages according to some embodiments of the technology described herein. [Figure 7C] 1 is an example of a hierarchical visualization of cellular composition percentages according to some embodiments of the technology described herein. [Figure 7D] 13 is a screenshot of an example report showing biomarker assessment based on determined cellular composition percentages, according to some embodiments. [Figure 7E]13 is a screenshot of an example report indicating the cellular composition percentage in a biological sample from a patient according to some embodiments of the technology described herein. [Figure 7F] 13 is a screenshot of an example report showing deviations of cellular composition percentage values ​​compared to normal ranges of cellular composition percentage values ​​associated with a reference cohort according to some embodiments of the technology described herein. [Figure 7G] 13 is a screenshot of an example report showing deviations of cellular composition percentage values ​​compared to normal ranges of cellular composition percentage values ​​associated with a reference cohort according to some embodiments of the technology described herein. [Figure 8] 8 is a flowchart illustrating an example process 800 for training multiple machine learning models to determine whether a cell is of a particular type according to some embodiments of the technology described herein. [Figure 9A] 13 shows the results of clustering cytometry data that has not been subjected to noise transformation according to some embodiments of the technology described herein. [Figure 9B] 13 shows the results of clustering noise-transformed cytometry data according to some embodiments of the technology described herein. [Figure 9C] 1 shows a distribution of marker intensities obtained from cytometry data that has not been subjected to noise transformation, according to some embodiments of the technology described herein. [Figure 9D] 1 shows a distribution of marker intensities obtained from noise transformed cytometry data according to some embodiments of the technology described herein. [Figure 9E] 1 is a plot used by conventional techniques to manually identify different types of events (e.g., cells or particles) based on marker values, according to some embodiments of the technology described herein. [Figure 9F] 1 is a plot showing event clusters that have been manually given display names indicating event type, according to some embodiments of the technology described herein. [Figure 10]1 illustrates an exemplary implementation of a computer system that may be used in conjunction with some embodiments of the technology described herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0111] There are many different cell populations in the human immune system, many of which play important roles in fighting disease and protecting the body from foreign substances.In some cases, the immune system can also include diseased cell populations (e.g., diseased B cells).For example, when cells undergo mutation, it can cause uncontrolled growth of one or more cell populations (e.g., forming cancerous cell populations).

[0112] Understanding the composition of cells of the immune system is useful in determining diagnosis, developing treatment strategies, and conducting research.For example, if the ratio of certain cell populations is known, it can be used to diagnose subjects with disease or predict whether a subject will respond to a certain treatment.For example, both CD4+ and CD8+ cells are favored in patients treated with rituximab.

[0113] Cytometry is a tool that can be used to identify the type of individual cells in a biological sample. Traditional techniques involve manual, time-consuming analysis of data generated during cytometry. In particular, such techniques involve plotting the cytometry data into a series of two-dimensional plots and manually identifying regions of interest in each plot, commonly referred to as "gating." To identify such regions of interest, an operator defines a boundary around a group of plotted points. Cells are identified as being of a particular cell type when they have marker values ​​that fall within the range of the region of interest and / or within a combination of regions of interest.

[0114] The inventors have recognized that conventional techniques for identifying cell types using cytometry have several problems. One such problem is that identifying regions of interest is a subjective procedure, which creates challenges in reproducibility. In particular, different operators make different decisions regarding where to place boundaries. For example, some operators may place more extensive boundaries, resulting in a greater number of points, while others may place more restrictive boundaries, resulting in a smaller number of points. As a result, there will be variability in the data used to determine the type of individual cells, leading to variability in the results of such analyses. This challenge is even more pronounced when analyzing large amounts of cytometry data, which involves generating and identifying regions of interest on a larger number of plots, leading to greater overall variability in the data used to determine cell types. As a result, different operators may classify the same cells differently (e.g., one operator may classify a cell as one type, while another operator may classify the cell as a different type). As a result, different operators will produce different estimates of cell population percentages (eg, estimates of the proportion of each of one or more cell types in the entire cell population).

[0115] Another problem with the conventional technique of identifying cell types using cytometry is that such data is extremely inefficient to analyze manually. This poses a challenge to large-scale studies, such as patient cohort studies and drug screening studies, where complex multi-dimensional data sets are generated. Manual processing of such data sets is extremely time-consuming, leading to high costs, which in turn affects the quality of data processing results or the entire study. For example, analysis of two-dimensional cytometry data plots can produce low-quality data, such as when the boundaries around plotted marker values ​​are not carefully defined. Determining cell types based on such data can be inaccurate, especially since closely related cell types share highly similar marker values, thus increasing the risk of being grouped together by boundaries that do not carefully distinguish them from each other.

[0116] Inaccurate and unreliable determinations of the cell type of a sample using conventional cytometry techniques also affect the accuracy and reliability of estimating the cell composition of a sample. The cell composition of a sample can be estimated based on the relative number of cells of each cell type in the sample. When based on an inaccurate cell type, the estimated cell composition will not accurately reflect the relative number of cells in each cell population. As explained above, the cell composition can be used to diagnose a subject with a disease or predict whether a subject will respond to a particular treatment. When the estimated cell composition is inaccurate, conventional techniques can diagnose a subject with the wrong disease or erroneously predict how a subject may respond to a treatment. For example, a relatively high cell composition percentage of CD4+ cells can indicate that the patient will respond well to rituximab. If many cells in a patient's sample are incorrectly determined to be CD4+ cells, then conventional techniques can erroneously predict that the patient will respond well to rituximab, when in fact the patient may respond negatively or not at all.

[0117] Moreover, variability in the data used to determine cell types will result in variability in the estimates of cell composition. For example, different operators may determine different types for the same cells, resulting in differences in the estimates of cell numbers for each cell population. Such variability makes it difficult to compare data collected from different studies, since the same sample may be estimated to have different cell compositions by different operators. As a result, it is difficult to derive meaningful insights from the data, such as correlations between cell composition and treatment response, diagnosis, and other clinically relevant outcomes. For example, clinicians may receive conflicting cell composition estimates from different cytometry operators, making it difficult to diagnose or select treatments for patients based on such estimates.

[0118] The present inventors have developed a technique that addresses the problems of the above-mentioned conventional techniques for determining the type of cells contained in a biological sample more accurately, more reliably, and more efficiently based on cytometry data. The technique includes processing cytometry data using multiple machine learning models to identify the types of cells present in the biological sample. In some embodiments, the cytometry data includes cytometry measurements (e.g., marker values) obtained during each cytometry event ("event").

[0119] An "event" corresponds to an object (e.g., a cell, debris, bead, doublet, or unknown object) in a biological sample under measurement by a cytometry platform (e.g., a flow cytometry platform or a mass cytometry platform). For example, an event may correspond to a cell in a biological sample under measurement by a cytometry platform, and measurements obtained during an event may be included in the cytometry data. In some embodiments, the multiple machine learning models used to process the cytometry data include a first machine learning model and a second machine learning model different from the first machine learning model. In some embodiments, the first machine learning model is used to determine an event type for a particular event by processing a cytometry measurement corresponding to the particular event. The event type may indicate whether the event corresponds to a cell under measurement by the cytometry platform, a debris under measurement by the cytometry platform, or a bead under measurement by the cytometry platform. For example, the first machine learning model may include a multi-class classifier trained to distinguish between at least some event types. In some embodiments, when the determined event type indicates that the cell being measured by the cytometry platform for the particular event corresponds to the cell type for the particular event, a second machine learning model is used to determine the cell type for the particular event by processing the cytometry measurement value corresponding to the particular event.For example, the second machine learning model may include a multi-class classifier trained to distinguish between at least some event cell types.

[0120] In some embodiments, the techniques include processing cytometry data for cells (e.g., types of events) in a biological sample using a machine learning model hierarchy corresponding to a cell type hierarchy. A machine learning model in the machine learning model hierarchy may be trained to predict a specific type for a cell using cytometry data corresponding to the cell. Additionally or alternatively, a machine learning model in the machine learning model hierarchy may include a multi-class classifier trained to distinguish between at least some cell types at a particular level in the hierarchy. Different levels of the machine learning model hierarchy may be used to predict types (e.g., general cell types or specific subtypes) for cells with different levels of specificity.

[0121] Such techniques improve cytometry by improving the accuracy and reproducibility of cell type determination results. In particular, at least one machine learning model is specifically trained to predict the type of cells in a biological sample based on cytometry data. In some embodiments, the machine learning model determines a confidence that the cell is of that particular type, and will not identify a type for the cell when the confidence does not exceed a threshold. This prevents inaccurate cell type identification of cells that may falsely appear to be of that particular type (e.g., have similar marker values ​​as other cells of that type). This use of one or more machine learning models removes the subjective process of conventional techniques, including relying on human operators to identify points to be included or excluded from a region of interest. As described above, such conventional techniques can produce over- or under-inclusive results, leading to inaccurate and inconsistent identification of cell types. In contrast, the systems and methods described herein using machine learning models produce more accurate and reproducible results. Accurate and reproducible identification of cell types is important for applications of cytometry where cell counts and / or differential cellular composition are used to provide diagnostic information and / or provide therapeutic implications.

[0122] Furthermore, the systems and methods described herein include techniques for identifying particles in biological samples, which contribute to the accuracy of cell type determination results and improved cytometry. In particular, at least one machine learning model (e.g., in a hierarchy of machine learning models) is trained to determine the type (e.g., beads, doublets, or debris) of particles in the biological sample. Rather than discarding all particles from the analysis or treating them as a whole, particles can be used to distinguish between each other and produce more accurate results. For example, particles identified as doublets can be analyzed separately as single cells for further analysis, which can improve the accuracy of total cell counts and specific types of cell counts. Additionally or alternatively, particles identified as beads can be used to determine the percentage of beads contained in the biological sample, which can be used to improve the analysis of cells contained in the biological sample (e.g., by providing information on the estimate of the percentage of cell composition).

[0123] Furthermore, the use of machine learning model hierarchy improves the specificity of cell type determination results. In particular, different levels of the hierarchy are used to determine cell types with different levels of specificity. The machine learning models at each level are trained to distinguish between cell types at the same level, allowing for the identification of subtle differences in cytometry data that are not readily apparent using conventional techniques. Cell type specificity is important for cytometry applications where the presence of a particular cell type and the degree to which a cell type is present relative to other cell types are predictive and / or prognostic biomarkers.

[0124] Thus, some embodiments provide computer-implemented techniques for determining the types of cells present in a biological sample using cytometry and multiple machine learning models. In some embodiments, the techniques include obtaining cytometry data (e.g., flow cytometry data or mass cytometry data) for a biological sample (blood, saliva, biopsy, etc.) previously obtained from a subject, e.g., a subject having, suspected of having, or at risk of having cancer (e.g., lymphoma) or an immune-related disease (e.g., rheumatoid arthritis). The biological sample may include a plurality of cells. The cytometry data may include cytometry measurements (e.g., marker values, fluorescence intensity values, intensity of heavy metal ion tags, etc.) obtained during each cytometry event ("event"). In some embodiments, an event corresponds to a particular object (e.g., a cell, debris, or bead) in the biological sample under measurement by the cytometry platform. The cytometry events may include a subset of events corresponding to cells in the biological sample under measurement by the cytometry platform. For example, the event subset may include one, some, or all of the cytometry events.

[0125] In some embodiments, the techniques include identifying a type of cell in the plurality of cells using a plurality of machine learning models, including a first machine learning model and a second machine learning model different from the first machine learning model. In some embodiments, the identifying is performed for each particular event in the event subset. In some embodiments, the identifying includes, for each particular event, obtaining a cytometry measurement corresponding to the particular event. For example, the cytometry measurement may be obtained from cytometry data obtained for the biological sample. In some embodiments, the identifying further includes determining an event type for the particular event by processing the cytometry measurement corresponding to the particular event using the first machine learning model. For example, the event type may indicate whether the particular event corresponds to a particular object (e.g., a cell, debris, or bead) under measurement by the cytometry platform. In some embodiments, when the determined event type indicates that the particular event corresponds to a cell under measurement by the cytometry platform, the identifying further includes determining a type of cell (e.g., one or more of the cell types listed in Table 1), which in some embodiments includes processing the cytometry measurement corresponding to the particular event using the second machine learning model.

[0126] In some embodiments, the event subset includes at least 5,000 events, at least 10,000 events, at least 20,000 events, at least 50,000 events, at least 100,000 events, at least 500,000 events, at least 600,000 events, at least 900,000 events, between 500 events and 1 million events, between 5,000 events and 900,000 events, or between 20,000 events and 700,000 events.

[0127] In some embodiments, the first machine learning model includes a first multi-class classifier and the second machine learning model includes a second multi-class classifier. Additionally or alternatively, the first machine learning model and / or the second machine learning model may include one or more binary class classifiers. In some embodiments, the first machine learning model includes a first decision tree classifier, a first gradient boosting decision tree classifier, or a first neural network, and the second machine learning model includes a second decision tree classifier, a second gradient boosting decision tree classifier, or a second neural network.

[0128] Some embodiments further include determining a cellular composition percentage of different cell types in the biological sample based on the identified plurality of cell types. For example, this may include determining a cellular composition percentage for one or more cell types (or subtypes) listed in Table 1. In some embodiments, determining the cellular composition percentage includes determining a first cellular composition percentage for the first cell type by determining a ratio between the number of cells identified as being of the first type in the plurality of cells and the total number of cells in the plurality of cells.

[0129] In some embodiments, the determined cell composition percentage is used to identify a treatment for the subject when the subject has, is suspected of, or is at risk of having cancer. For example, in some embodiments, identifying a treatment for the subject based on the determined cell composition percentage includes determining a ratio between a cell composition percentage of CD8+PD-1+ cells and a cell composition percentage of CD4+PD-1, and identifying an immune checkpoint blockade therapy for the subject when the determined ratio is above a threshold. As another example, in some embodiments, identifying a treatment for the subject based on the determined cell composition percentage includes identifying ipilimumab for the subject when the cell composition percentage of peripheral blood mononuclear cells (PBMCs) is below a threshold. In some embodiments, the identified treatment is administered to the subject (e.g., used to treat the subject).

[0130] In some embodiments, the determined cellular composition percentages are used to identify the subject as a member of a cohort. For example, a patient cohort may include a healthy cohort, a cohort of patients with a disease (e.g., cancer or immune-related disease), or a cohort of patients who have undergone treatment. Identifying the subject as a member of a cohort, in some embodiments, includes comparing a cellular composition percentage of the determined cellular composition percentages to a cellular composition percentage range associated with the patient cohort; and identifying the subject as a member of the patient cohort based on the results of the comparison. For example, the cellular composition percentage ranges may be obtained from a data store and / or may be obtained from one or more other subjects whose cellular composition percentages have been determined.

[0131] In some embodiments, the cytometry events additionally include a second subset of events corresponding to beads in the biological sample under measurement by the cytometry platform, and a third subset of events corresponding to debris in the biological sample under measurement by the cytometry platform. In some embodiments, the techniques include determining an event type for one or more events by processing cytometry measurements corresponding to one or more events in the second and / or third subset of events. For example, this may include determining an event type for one or more events by processing the cytometry measurements for the one or more events with a first machine learning model.

[0132] In some embodiments, the plurality of events includes a first plurality of events and a second plurality of events. The cytometry data may include first cytometry data for the first plurality of events (e.g., corresponding to objects in a first subsample of the biological sample under measurement on the cytometry platform) and second cytometry data for the second plurality of events (e.g., corresponding to objects in a first subsample of the biological sample under measurement on the cytometry platform). The first cytometry data may include values ​​(e.g., fluorescence intensity values, relative intensities of heavy metal ion tags) for a first subset (e.g., a first panel) of the plurality of markers for each of at least a portion of the first plurality of events, and the second cytometry data may include values ​​(e.g., fluorescence intensity values, relative intensities of heavy metal ion tags) for a second subset (e.g., a second panel) of the plurality of markers for each of at least a portion of the second plurality of events. In some embodiments, the first and second subsets of the plurality of markers are different. For example, the first subset of the plurality of markers may include none, some, half, or most of the markers included in the second subset of the plurality of markers. In some embodiments, the first cytometric data includes data from a first panel, and the second cytometric data includes data from a second panel that is different from the first panel.

[0133] Some embodiments provide computer-implemented techniques for determining a type for cells in a biological sample using cytometry data and a machine learning model hierarchy. In some embodiments, the techniques include obtaining cytometry data for a biological sample (e.g., blood, saliva, biopsy, etc.) from a subject (e.g., a subject having, suspected of having, or at risk of having cancer (e.g., lymphoma) or an immune-related disease (e.g., rheumatoid arthritis). The biological sample may include a plurality of cells including a first cell, and the cytometry data may include first cytometry data (e.g., one or more marker values) for the first cell. In some embodiments, the techniques include determining a respective type for each of at least a portion (e.g., at least 10%, at least 30%, at least 50%, at least 70%, at least 90%, etc.) of the plurality of cells using a machine learning model hierarchy (e.g., a decision tree classifier, a gradient boosting decision tree classifier, etc.) corresponding to a cell type hierarchy (e.g., cell types and cell subtypes). For example, a machine learning model in the machine learning model hierarchy may be trained to determine whether a cell is a particular type corresponding to the cell hierarchy. In some embodiments, the determining includes determining a first type for a first cell by processing the first cytometry data using a first subset (e.g., a first plurality) of the machine learning model hierarchy. For example, the first machine learning model subset may correspond to a particular path through the hierarchy.

[0134] In some embodiments, the plurality of cells includes a second cell of a different type than the first cell, and the cytometry data includes second cytometry data (e.g., one or more marker values) for the second cell. In some embodiments, determining the respective types of the cells further includes processing the second cytometry data using a second subset (e.g., a second plurality) of the machine learning model hierarchy, the second subset of the machine learning model hierarchy being different from the first subset of the machine learning model hierarchy. For example, the second subset of the machine learning model hierarchy may represent a second path through the hierarchy. In some embodiments, the first and second subsets of the machine learning model hierarchy may include one or more of the same machine learning models. Additionally or alternatively, the first and second subsets may not include any of the same machine learning models.

[0135] In some embodiments, the techniques include determining the respective type for each of at least 5,000 cells, at least 10,000 cells, at least 20,000 cells, at least 50,000 cells, at least 100,000 cells, at least 500,000 cells, at least 600,000 cells, or at least 900,000 cells.

[0136] In some embodiments, processing the first cytometry data using a first subset of the machine learning model hierarchy includes processing the first cytometry data using a first machine learning model (e.g., a machine learning model trained to determine whether the first cell is a first type). For example, the first machine learning model may be trained to determine whether the first cell is a lymphocyte. The processing may further include identifying a second machine learning model in the first subset of the machine learning model hierarchy (e.g., a machine learning model trained to determine whether the first cell is a second type) based on the output of the first machine learning model (e.g., an indication of whether the first cell is a first type). For example, if the output of the first machine learning model indicates that the first cell is a first type (e.g., a lymphocyte), then the technique may include a second machine learning model trained to determine whether the first cell is a subtype of the first cell type (e.g., a T cell). In some embodiments, the techniques include processing the first cytometry data with a second machine learning model to obtain a second output (e.g., an indication of whether the first cell is of a second type).

[0137] In some embodiments, the first output indicates a first type for the first cell and the second output indicates a subtype for the first cell. For example, the first type may include white blood cells while a subtype of the first type includes granulocytes. As another example, the first type may include lymphocytes while a subtype of the first type includes B cells. As yet another example, the first cell type may include T helper cells while a subtype of the first type may include memory T helper cells. Examples of cell types and relationships between cell types are provided herein, including at least those in Table 1.

[0138] In some embodiments, processing the first cytometry data using the first subset of machine learning models further includes identifying a third machine learning model in the first subset of the machine learning model hierarchy (e.g., one trained to determine whether the cell is of a third type) based on the first output of the first machine learning model (e.g., a first type for the first cell). For example, identifying the third machine learning model may include identifying a machine learning model trained to determine whether the cell is a different subtype (e.g., a B cell) of the first cell type (e.g., a lymphocyte). The techniques may further include processing the first cytometry data using the third machine learning model.

[0139] In some embodiments, processing the first cytometry data using the first subset of the machine learning model hierarchy further includes identifying a third machine learning model (e.g., trained to determine whether the cell is a third type) in the first subset of the machine learning model hierarchy based on the second output of the second machine learning model (e.g., a second type for the second cell). The third machine learning model may be trained to determine whether the cell is a subtype of the second cell type determined by the second machine learning model. For example, if the second output of the second machine learning model indicates that the cell is a B cell, the third machine learning model may be trained to determine whether the cell is a memory B cell. In some embodiments, the technique then includes processing the first cytometry data using the third machine learning model.

[0140] In some embodiments, the machine learning models in the machine learning model hierarchy may include a decision tree classifier, a gradient boosting decision tree classifier, a neural network, or any other suitable type of machine learning model, as aspects of the technology described herein are not limited in this respect. In some embodiments, the machine learning models in the machine learning model hierarchy may include an ensemble of any suitable type of machine learning models.

[0141] In some embodiments, the techniques further include determining a respective cellular composition percentage for at least some of the determined cell types. This may include determining a first cellular composition percentage for a first cell type. For example, this may include estimating the percentage of B cells in the biological sample. In some embodiments, determining the first cellular composition percentage includes determining a ratio between the number of cells estimated to be of a first type (e.g., based on cytometry data) and the total cell number of at least some of the plurality of cells in the biological sample.

[0142] In some embodiments, the techniques include comparing the determined cell composition percentage (e.g., a first cell composition percentage) to a cell composition percentage range associated with a patient cohort. For example, the patient cohort may include patients who have been diagnosed with a disease, patients receiving a particular treatment, healthy patients, and / or patients with one or more other conditions or characteristics. The techniques may include identifying the subject as a member of the patient cohort based on the results of the comparison. For example, this may include identifying the subject as healthy, as having a particular disease, as likely to respond to a particular therapy, and / or as having another condition.

[0143] In some embodiments, the techniques include generating a visualization of the determined cellular composition percentage, the visualization indicating a comparison of the first cellular composition percentage to a cellular composition percentage range associated with the patient cohort. For example, the visualization may include a graphic indicating a range of reference cellular composition percentages and an indication of where the first cellular composition percentage falls within the range.

[0144] In some embodiments, the techniques include comparing the first cell composition percentage to a cell composition percentage range associated with a study. In some embodiments, the study evaluates the effectiveness of one or more treatments in treating a disease, and identifying a treatment for the subject based on the results of the comparison. For example, such a treatment may be identified by comparing the determined cell composition percentage to the cell composition percentages of patients with good or bad survival rates after receiving different treatments.

[0145] In some embodiments, the techniques include generating a visualization of the determined cellular composition percentages using a cell type hierarchy. The visualization may include a plurality of nodes organized in a hierarchy, at least some of the plurality of nodes being connected by respective edges. The plurality of nodes may include a first node representing a first cell type and a second node representing a subtype of the first cell type, while the edges may include a first edge connecting the first node and the second node.

[0146] In some embodiments, the visualization includes an indication of a cell type subset of the cell type hierarchy. In some embodiments, the determined cell type subset cell composition percentages include aberrant cell composition percentages compared to a reference cell composition percentage. For example, aberrant cell composition percentages may include those that fall outside of a cell composition percentage range associated with a patient cohort (e.g., of healthy patients, of patients diagnosed with a particular disease, and / or of patients undergoing a particular treatment).

[0147] In some embodiments, the visualization further includes a first number for the first cell type, indicating a percentage of cells having the first cell type in the biological sample. For example, the first number may be shown adjacent to the first node. In some embodiments, the size of the node compared to the size of other nodes in the visualization may represent the relative proportion of cells having a particular cell type in the biological sample. For example, a larger node may represent a higher percentage of cells having a particular cell type.

[0148] In some embodiments, determining each cellular composition percentage further includes determining one or more second cellular composition percentages for one or more second cell types that are subtypes of the first cell type. By way of non-limiting example, this may include determining a first cellular composition percentage for lymphocytes and a second cellular composition percentage for B cells and T cells.

[0149] In some embodiments, the techniques include comparing a sum of one or more second cellular composition percentages (e.g., a sum of the percentages of B cells and T cells) to a first cellular composition percentage (e.g., a percentage of lymphocytes). The techniques may include determining a normalization factor (e.g., a ratio between the first cellular composition percentage and the sum of the one or more second cellular composition percentages) based on a result of the comparison, and applying the normalization factor to the one or more second cellular composition percentages. For example, the normalization factor may take into account a difference between the first cellular composition percentage and the sum of the one or more second cellular composition percentages.

[0150] In some embodiments, the techniques further include determining whether the one or more second cell types include all subtypes of the first cell type, and applying a normalization factor to the one or more second cellular composition percentages when the one or more second cell types include all subtypes of the first cell type. As a non-limiting example, the normalization factor may not be applied to the cellular composition percentages of B cells and T cells because B cells and T cells do not represent all subtypes of lymphocytes.

[0151] In some embodiments, the techniques include comparing a sum of one or more second cellular composition percentages (e.g., a sum of the percentages of B cells and T cells) to a first cellular composition percentage (e.g., a percentage of lymphocytes). The techniques may include determining a type-specific normalization factor for each of the one or more second cell types based on a result of the comparison. For example, this may include determining a normalization factor for B cells and a normalization factor for T cells. In some embodiments, the techniques include applying the type-specific normalization factor to each of the one or more cellular composition percentages.

[0152] In some embodiments, the flow cytometry data obtained includes noise-transformed flow cytometry data. For example, noise may be reduced using any suitable noise transformation technique.

[0153] In some embodiments, the cytometry data includes values ​​(e.g., fluorescence intensity measurements) for each of at least a portion of a plurality of markers (e.g., fluorescently labeled antibodies or fluorescent dyes or stains, heavy metal ion tags) for each of at least a portion of the plurality of cells.

[0154] In some embodiments, the plurality of cells includes a first plurality of cells and a second plurality of cells. The cytometry data may include a first subset of cytometry data for the first plurality of cells (e.g., in a first subsample of the biological sample) and a second subset of cytometry data for the second plurality of cells (e.g., in a second subsample of the biological sample). The first subset of cytometry data may include values ​​(e.g., fluorescence intensity values, relative intensities of heavy metal ion tags) for a first subset (e.g., a first panel) of a plurality of markers for each of at least a portion of the first plurality of cells, and the second subset of cytometry data may include values ​​(e.g., fluorescence intensity values, relative intensities of heavy metal ion tags) for a second subset (e.g., a second panel) of a plurality of markers for each of at least a portion of the second plurality of markers. In some embodiments, the first and second subsets of a plurality of markers are different. For example, a first subset of the plurality of markers may include none, some, half, or most of the markers included in a second subset of the plurality of markers.

[0155] In some embodiments, determining the respective types for at least a portion of each of the plurality of cells includes determining a first plurality of respective types for the first plurality of cells and determining a second plurality of respective types for the second plurality of cells. For example, this may include determining a type for cells in a first subsample of the biological sample and determining a type for cells in a second subsample of the biological sample. The technique may include using the first plurality of cell types to determine a first plurality of cellular composition percentages. This may include determining a respective cellular composition percentage for each of at least a portion of the first plurality of cell types. The technique may further include using the second plurality of cell types to determine a second plurality of cellular composition percentages for each of at least a portion of the second plurality of cell types.

[0156] In some embodiments, the techniques include determining a respective cellular composition percentage for each of at least a portion of the determined cell types. For example, this may include determining a cellular composition percentage for each of at least a portion of the plurality of cell types based on the total number of cells in both the first and second plurality of cells. In some embodiments, the determining may include combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages.

[0157] In some embodiments, combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages may be based on estimated composition percentages of cell types contained in the first plurality of cells and the second plurality of cells. For example, each of at least a portion of the first plurality of cellular composition percentages may be normalized to an estimated composition percentage of a common cell type in the first plurality of cells. Similarly, each of at least a portion of the first plurality of cellular composition percentages may be normalized to an estimated composition percentage of a common cell type in the second plurality of cells.

[0158] In some embodiments, the techniques include combining at least a portion of the first plurality of cellular composition percentages with at least a portion of the second plurality of cellular composition percentages based on data obtained using beads contained in the biological sample. For example, this may include normalizing a portion of the first plurality of cellular composition percentages to the concentration of beads contained in the first plurality of cells, and normalizing a portion of the second plurality of cellular composition percentages to the concentration of beads contained in the second plurality of cells.

[0159] In some embodiments, the biological sample includes a plurality of particles (e.g., debris, doublets, beads, etc.), and the cytometry data includes cytometry data (e.g., a marker value) for a first particle of the plurality of particles.

[0160] In some embodiments, the techniques include using a machine learning model hierarchy to determine a respective particle type for each of at least a portion of the plurality of particles. This may include using the machine learning model hierarchy to determine a first particle type for a first particle. A machine learning model in the machine learning model hierarchy may be trained to determine whether a particle or cell is a particular particle type. For example, a machine learning model may be trained to determine whether a particle or cell is a bead (or debris or doublet). In some embodiments, this may also include determining whether the first particle is unidentifiable.

[0161] In some embodiments, obtaining the cytometry data includes processing the biological sample using a cytometry platform.

[0162] In some embodiments, the machine learning model hierarchy includes at least 50 machine learning models, at least 100 machine learning models, at least 200 machine learning models, at least 250 machine learning models, at least 300 machine learning models, at least 350 machine learning models, or at least 400 machine learning models.

[0163] The following is a more detailed description of various concepts related to the cell type determination system and method developed by the inventors, and embodiments thereof. It should be understood that the various aspects described herein may be implemented in any of numerous ways. Specific examples of implementations are provided herein, but are for illustrative purposes only. In addition, the various aspects represented in the following embodiments may be used alone or in any combination, and are not limited to the combinations expressly described herein.

[0164] FIG. 1A depicts an exemplary technique 100 for determining a respective type 110 for each of one or more events. As described herein, in some embodiments, an event corresponds to obtaining a measurement for an object in a biological sample 102. Obtaining a measurement for an object includes, in some embodiments, obtaining cytometry data, such as cytometry data 106. In some embodiments, the cytometry data 106 is obtained by processing the object using a cytometry platform 104. Additionally or alternatively, the cytometry data 106 may have been previously obtained using the cytometry platform 104. The respective type 110 for each of the one or more events is determined by processing the cytometry data 106 using a computing device 108. In some embodiments, the computing device 108 may be part of the cytometry platform 104. In other embodiments, the computing device 108 may be separate from the cytometry platform 104 and may receive the cytometry data 106 directly or indirectly from the cytometry platform 104.

[0165] In some embodiments, the illustrated technique 100 may be implemented in a clinical or laboratory setting. For example, the illustrated technique 100 may be implemented in a computing device 108 located within a clinical or laboratory setting. In some embodiments, the computing device 108 may obtain the cytometry data 106 directly from a cytometry platform 104 located within a clinical or laboratory setting. For example, a computing device 108 included within a cytometry platform 104 may obtain the cytometry data 106 directly from the cytometry platform 104. In some embodiments, the computing device 108 may obtain the cytometry data 106 indirectly from a cytometry platform 104 located within or outside of a clinical or laboratory setting. For example, a computing device 108 located within a clinical or laboratory setting may obtain the cytometry data 106 via a communications network, such as the Internet, or via any other suitable network, as aspects of the technology described herein are not limited to any particular communications network.

[0166] Additionally or alternatively, the illustrated technique 100 may be implemented in a setting that is remote from a clinical or laboratory setting. For example, the illustrated technique 100 may be implemented on a computing device 108 located outside of a clinical or laboratory setting. In this case, the computing device 108 may indirectly obtain cytometry data 106 generated using a cytometry platform 104 located within or outside of the clinical or laboratory setting. For example, the cytometry data 106 may be provided to the computing device 108 via a communications network, such as the Internet, or via any other suitable network, as aspects of the technology described herein are not limited to any particular communications network.

[0167] As shown in FIG. 1A, the technique 100 involves processing a biological sample 102 using a cytometry platform 104, which generates cytometry data 106. The biological sample 102 may be obtained from a subject having, suspected of having, or at risk of having cancer or any immune-related disease. The biological sample 102 may be obtained by performing a biopsy or by obtaining a blood sample, saliva sample, or any other suitable biological sample from the subject. The biological sample 102 may include diseased tissue (e.g., cancerous) and / or healthy tissue. In some embodiments, the method of deriving or preparing the biological sample may include any of the embodiments described herein, including with respect to the "Biological Sample" section.

[0168] In some embodiments, the cytometry platform 104 includes any suitable equipment and / or system configured to perform cytometry, such that aspects of the technology described herein are not limited to any particular type of cytometry system. For example, the cytometry platform may include any suitable flow cytometry platform. Additionally or alternatively, the cytometry platform may include any suitable mass cytometry platform. In some embodiments, the biological sample 102 may be prepared according to a manufacturer's protocol associated with the cytometry platform 104. In some embodiments, the biological sample may be prepared according to any suitable protocol, such that embodiments of the technology described herein are not limited to any particular preparation protocol. In some embodiments, the flow cytometry technique may include any of the embodiments described herein, including with respect to the "Flow Cytometry" section. In some embodiments, the mass cytometry technique may include any of the embodiments described herein, including with respect to the "Mass Cytometry" section.

[0169] In some embodiments, the cytometry data 106 includes cytometry data for each of one or more events. Each event may correspond to obtaining a cytometry measurement for an object in a biological sample using a cytometry platform. In some embodiments, the object includes a cell, a particle, and / or an unknown object. In some embodiments, the particle includes a bead, debris, and / or a doublet. A "bead," or calibration bead, is a particle of known concentration that can be mixed with a known volume of a biological sample prior to processing by a flow cytometer or mass cytometer. The ratio of beads detected and identified in the cytometry data for a sub-sample can be used to determine the number of cells in the sub-sample and / or the number of cells of a particular type in the sub-sample. A "doublet" is a pair of two independent particles or cells that are processed and classified as a single particle by the cytometry platform. This occurs when two cells or particles pass close to each other through the cytometry platform. The cytometry data 106 is further described herein, including with respect to at least Figures 1B-1C.

[0170] In some embodiments, the cytometry data 106 is processed using a computing device 108. In some embodiments, the computing device 108 may be one or more computing devices of any suitable type. For example, the computing device 108 may be a portable computing device (e.g., laptop, smartphone) or a stationary computing device (e.g., desktop computer, server). When the computing device 108 includes multiple computing devices, the one or more devices may be physically co-located (e.g., in a single room) or distributed across multiple physical locations. In some embodiments, the computing device 108 may be part of a cloud computing infrastructure. In some embodiments, one or more computers 108 may be co-located at a facility operated by an entity (e.g., hospital, research institute). In some embodiments, one or more computing devices 108 may be physically co-located with a medical instrument, such as a cytometry platform 104. For example, the cytometry platform 104 may include a computing device 108.

[0171] In some embodiments, the computing device 108 may be operated by a user, such as a physician, clinician, researcher, patient, or other individual. For example, a user may provide the cytometry data 106 as input to the computing device 108 (e.g., by uploading a file) and / or may provide user input that specifies a process or other method to be performed using the cytometry data 106.

[0172] In some embodiments, the computing device 108 includes software configured to perform various functions on the cytometry data 106. Examples of computing devices 108 including such software are described herein, including with respect to at least FIG. 1D. In some embodiments, the software on the computing device 108 is configured to process the cytometry data to identify a respective cell or particle type 110 for each of the one or more events. Example techniques for processing the cytometry data 106 are described herein, including with respect to at least FIG. 1E.

[0173] In some embodiments, technique 100 additionally includes determining one or more cellular composition percentages for cell types in the biological sample by processing the cytometry data and / or the identified cell or particle types using computing device 108. The cellular composition percentages indicate the proportion of a particular cell type present in the biological sample 102.

[0174] In some embodiments, the cell composition percentage for the biological sample 102 is compared to the cell composition percentage associated with the cohort to predict a diagnosis for the subject, predict how the subject is likely to respond to a particular treatment, select a treatment for the subject, or perform any other suitable application, as aspects of the technology described herein are not limited in this respect. For example, if the cell composition percentage determined for the biological sample 102 for the subject is comparable to the cell composition percentage associated with a cohort of patients who responded well to a particular treatment, it may indicate that the subject is likely to respond well to the treatment. Additionally or alternatively, if the cell composition percentage determined for the biological sample 102 for the subject is comparable to the cell composition percentage associated with a cohort of patients diagnosed with a particular disease, it may indicate that the subject is likely to have the disease.

[0175] In some embodiments, technique 100 may include generating a report indicating the determined cell and / or particle types 110, the cell composition percentages, the predicted treatments, the predicted diagnoses, and / or any other suitable data obtained from technique 100. In some embodiments, the report may include graphics and / or text. In some embodiments, the report may be stored in a memory or displayed on a user interface (e.g., a graphical user interface (GUI)) of a computing device (e.g., computing device 108). Example report generation techniques are described herein, including with respect to at least FIG. 1D and FIG. 7A-G.

[0176] As a non-limiting example, technique 100 may be performed to determine the cellular composition percentages of various different cell types in a subject suspected of having leukemia. A blood sample may be obtained by a physician and processed using a cytometry platform to obtain cytometry data. The cytometry data may be processed using a computing device to determine the respective cell or particle type for each event, and the cellular composition percentages may be determined for the determined cell and particle types. The cellular composition percentages indicate the proportions of different cell populations (e.g., populations of different cell types) present in the blood sample. For example, this may include determining the percentage of T cells in the blood sample. In some embodiments, this cellular composition percentage is compared to a cellular composition percentage associated with a different patient cohort. For example, the subject's estimated percentage of T cells may be compared to the percentage of T cells associated with a cohort of patients diagnosed with a particular disease. If the subject has a comparable T cell composition percentage, it may be an indication that the subject is a member of the cohort (e.g., has the disease). A report may be generated directing visualization of the cohorts identified for the subject and the cell populations in the blood sample.

[0177] 1B, the cytometry data 106 includes cytometry data for each of a plurality of events, in this example, Event 1 through Event N. The cytometry data 106 includes cytometry data 132-1 for a first event, cytometry data 132-2 for a second event, cytometry data 132-3 for a third event, etc. In some embodiments, all of Events 1-N correspond to cells. In some embodiments, some of Events 1-N correspond to cells, while some of Events 1-N correspond to particles.

[0178] In some embodiments, the cytometry data 106 indicates one or more values ​​for one or more markers used to obtain the cytometry data. A "marker" may include a protein found in a particular cell type or types. A "marker value" may be indicative of the expression of such a protein.

[0179] In some embodiments, the markers may be fluorescently labeled and the flow cytometry platform may measure the intensity of the fluorescence that a particular cell emits when it is processed: cells that express a higher expression level of the marker will have a higher marker value (e.g., a higher intensity measurement).

[0180] In some embodiments, different markers are labeled with fluorescent proteins of different colors. This helps distinguish between the expression of different markers. In some embodiments, the fluorescence intensity is measured for each color of fluorescence emitted by a cell, each of which can be associated with a particular respective marker. For example, if a cell fluoresces green, red, and blue, it can indicate that the cell expresses three different markers.

[0181] Additionally or alternatively, in some embodiments, the markers may be labeled using heavy metal ion tags, and the mass cytometry platform may measure the relative intensity (or abundance) of the heavy metal ion tags. The relative intensity of the tags quantifies the amount of ions produced relative to the amount of the most abundant ion. In some embodiments, a relative intensity is measured for each heavy metal ion tag detected from a cell, each of which may be associated with a particular respective marker.

[0182] FIG. 1C shows example marker values ​​contained in the cytometry data 106 for example markers (eg, CD3, CD62L, CD27, CD45+, and IgA+).

[0183] According to some embodiments, each of the example markers is labeled with a fluorescent protein of a particular color, and the corresponding marker value indicates the intensity of the fluorescence of that color emitted by the cells. For example, consider the marker CD3, which is labeled with green fluorescent protein. The marker value of CD3 for a number of cells represents the intensity of the green fluorescence emitted by those cells.

[0184] According to some embodiments, each of the exemplary markers is labeled with a particular heavy metal ion tag, and the corresponding marker value indicates the intensity of the tag relative to the most abundant ion.

[0185] It should be understood that the markers shown in Figure 1C are non-limiting examples, and any suitable marker or combination of markers may be used in conjunction with some embodiments of the technology described herein. Exemplary markers are described herein, including with respect to at least Table 2.

[0186] In some embodiments, the computing device 108 includes software 112 configured to perform various functions with respect to the cytometry data 106. In some embodiments, the software 112 includes a number of modules. The modules may include processor-executable instructions that, when executed by at least one computer hardware processor, cause the at least one computer hardware processor to perform one or more functions of the module. Such modules, sometimes referred to herein as "software modules," each include processor-executable instructions configured to perform one or more processes, such as those described herein, including with respect to at least Figures 2A-2C, 5A-5D, and 8.

[0187] 1D is a block diagram of a system 150 including an example of a computing device 108 and software 112 according to some embodiments of the technology described herein. The software 112 includes one or more software modules for processing cytometry data, such as an event type determination module 172, a cell composition percentage module 174, a cohort identification module 176, and a report generation module 178. In some embodiments, the software 112 additionally includes a user interface module 170, a cytometry platform interface module 162, and / or a data store interface module 160 for obtaining data (e.g., user input, cytometry data, one or more machine learning models). In some embodiments, data is obtained from a cytometry platform 152, a cytometry data store 154, and / or a machine learning model data store 166. In some embodiments, the software 112 further includes a machine learning model training module 164 for training one or more machine learning models (e.g., stored in the machine learning model data store 166).

[0188] In some embodiments, data obtained from the cytometry data store 154 and / or the cytometry platform 152 and machine learning models obtained from the machine learning model data store 166 are used by the event type determination module 172 to determine the type of one or more events. In some embodiments, the obtained data includes cytometry data for a biological sample from a subject. For example, the cytometry data may include cytometry data for each of a plurality of events in the biological sample, such as first cytometry data for a first event. In some embodiments, the machine learning models obtained from the machine learning model data store 166 include machine learning models organized in a hierarchy of machine learning models corresponding to a cell type hierarchy. In some embodiments, the machine learning models obtained may include one, some, most, or all of the machine learning models included in the machine learning model hierarchy.

[0189] In some embodiments, the event type determination module 172 uses at least one machine learning model to determine a respective type for at least a portion of the events in the biological sample. For example, in some embodiments, this includes determining an event type for a particular event by processing cytometry measurements corresponding to the event using a first machine learning model. For example, the first machine learning model may include a multi-class classifier trained to predict the event type among multiple event types. Additionally or alternatively, the first machine learning model may include one or more binary class classifiers each trained to predict whether an event is a particular event type. When the event type indicates that a particular event corresponds to a cell under processing in the cytometry platform, then the event type determination module 172, in some embodiments, determines a cell type for the event by processing cytometry measurements corresponding to the particular event using a second machine learning model. For example, the second machine learning model may include one or more multi-class classifiers trained to predict a cell type for the event among multiple cell types. Additionally or alternatively, the second machine learning model may include one or more binary class classifiers, each trained to predict whether the event corresponds to obtaining a measurement for a particular type of cell. In some embodiments, the first machine learning model is different from the second machine learning model.

[0190] In some embodiments, the event type determination module 172 determines a respective type for each of at least a portion of the events in the biological sample using a machine learning model hierarchy corresponding to the event type hierarchy. In some embodiments, this includes determining a first type for a first event in the biological sample by processing the first cytometry data with a subset of the machine learning model hierarchy. For example, the subset of the machine learning model hierarchy includes one or more of the machine learning models obtained from the machine learning model data store 166. In some embodiments, the event type determination module 172 determines a type for an event contained in the biological sample by processing the cytometry data according to techniques described herein, including with respect to at least FIGS. 2A-2C.

[0191] In some embodiments, the event type determination module 172 obtains the cytometry data and / or the machine learning models from one or more interface modules. In some embodiments, the interface modules include a cytometry platform interface module 162 and a data store interface module 160. The cytometry platform interface module 162 may be configured to obtain (either pull or be provided) the cytometry data from the cytometry platform 152. The data store interface module 160 may be configured to obtain (either pull or be provided) the cytometry data and / or the machine learning models from the cytometry data store 154 and / or the machine learning model data store 166, respectively. The data and / or the machine learning models may be provided via a communications network (not shown), such as the Internet, or via any other suitable network, as aspects of the technology described herein are not limited to any particular communications network.

[0192] In some embodiments, the cytometry data store 154 includes any suitable data store, such as a flat file, a data store, a multifile, or any suitable type of data storage, as aspects of the technology described herein are not limited to any particular type of data store. The cytometry data store 154 may be part of the software 108 (not shown) or may be separate from the software 108, as shown in FIG. 1D.

[0193] In some embodiments, the cytometry data store 154 stores cytometry data obtained from one or more biological samples of one or more subjects. In some embodiments, the cytometry data may be cytometry data from the cytometry platform 152 and / or cytometry data obtained from one or more public data stores and / or studies. In some embodiments, a portion of the cytometry data may be processed by the event type determination module 172 to determine a type for an event associated with the cytometry data. In some embodiments, one or more machine learning models may be trained (e.g., by the machine learning model training module 164) using the portion of the cytometry data. In some embodiments, the portion of the cytometry data may include additional data (e.g., event type and / or cellular composition percentage) and may be associated with one or more patients in a cohort. This portion of the cytometry may be used, for example, by the cohort identification module 176 to identify the subject as a member of a cohort.

[0194] In some embodiments, the machine learning model datastore 166 includes any suitable data store, such as a flat file, a data store, a multifile, or any suitable type of data storage, as aspects of the technology described herein are not limited to any particular type of data store. The machine learning model datastore 166 may be part of the software 108 (not shown) or may be separate from the software 108, as shown in FIG. 1D.

[0195] In some embodiments, the machine learning model datastore 166 stores one or more machine learning models. For example, the machine learning model datastore 166 may store a machine learning model trained to predict an event type for an event. Additionally or alternatively, the machine learning model datastore 166 may store one or more machine learning models trained to predict a cell type for an event. In some embodiments, the machine learning model datastore 166 stores a machine learning model hierarchy used to determine the type of event based on the cytometry data. In some embodiments, the machine learning model hierarchy corresponds to an event type hierarchy. Relationships between the machine learning models in the hierarchy may be stored in the machine learning model datastore 166. For example, the machine learning model datastore 166 may store a relationship between a machine learning model trained to predict a particular event type and a machine learning model trained to predict a subtype of the event type.

[0196] In some embodiments, the cellular composition percentage module 174 estimates the cellular composition percentage for different types of cell populations in the biological sample. In doing so, the cellular composition percentage module 174 may use labeled cytometry data that indicates an event type for the event for which the cytometry data was obtained. For example, the labeled cytometry data may include cytometry data for a first event that is labeled with a type for the first event. In some embodiments, the label may be determined by the event type determination module 172 and / or through alternative means. In some embodiments, the cellular composition percentage module 174 estimates the cellular composition percentage for a cell type contained in the biological sample. Additionally or alternatively, the cellular composition percentage module 174 may estimate the cellular composition percentage for a cell type contained in a sub-sample of the biological sample. Examples of techniques for estimating the cellular composition percentage are described herein, including with respect to at least Figures 5B-5C and 6A-6C.

[0197] In some embodiments, the cohort identification module 176 identifies a patient cohort to which a subject (e.g., the subject from whom a biological sample was obtained) belongs. This may include comparing the subject's cellular composition percentage to cellular composition percentages associated with one or more patient cohorts. In some embodiments, the cohort identification module 176 may obtain data associated with one or more patient cohorts from the data store interface module 160. Additionally or alternatively, the cohort identification module 176 may obtain input from a user 168 via the user interface module 170 indicating one or more cohorts (and their associated cellular composition percentages) to which the subject's cellular composition percentage should be compared. Example techniques for identifying a subject as a member of a cohort are described herein, including with respect to at least FIG. 5A.

[0198] In some embodiments, the report generation module 178 generates one or more reports by processing results obtained from the event type determination module 172, the cellular composition percentage module 174, and / or the cohort identification module 176. For example, the one or more reports may indicate the event types contained in the biological sample, the proportions of cell populations (e.g., cellular composition percentages) in the biological sample, and / or one or more cohorts to which the subject belongs. Additionally or alternatively, the one or more reports may indicate any other suitable information, such as, for example, a diagnosis for the subject, a proposed treatment, and / or relationships between cell populations. In some embodiments, the reports may include visualizations, such as charts, graphs, tables, and / or any other suitable visualizations displaying data. Example reports are described herein, including with respect to at least FIGS. 7A-7G.

[0199] User interface 170 may be a graphical user interface (GUI), a text-based user interface, and / or any other suitable type of interface through which a user may provide input. For example, in some embodiments, the user interface may be a web page or web application accessible from an internet browser. In some embodiments, the user interface may be a graphical user interface (GUI) of an app running on a user's mobile device. In some embodiments, the user interface may include a number of selectable elements through which a user may interact. For example, the user interface may include drop-down lists, check boxes, text areas, or any other suitable elements.

[0200] In some embodiments, the machine learning model training module 164, referred to herein as training module 164, is configured to train one or more machine learning models used to determine a type for an event. This may include training a machine learning model to determine whether an event is of a particular type. In some embodiments, the training module 164 trains the machine learning model using a training set of cytometry data. For example, the training module 164 may obtain the training data from the data store interface module 160. In some embodiments, the training module 164 may provide the trained machine learning model to the machine learning model data store 166 via the data store interface module 160. Techniques for training machine learning models are described herein, including with respect to at least FIG. 8.

[0201] 1E depicts an example technique 180 for determining a respective type for one or more events (e.g., cells or particles) by processing the cytometry data 106 using a computing device 108. In some embodiments, the example technique 190 includes providing the first cytometry data 132-1 for the first event as an input to a machine learning model hierarchy that is used to determine one or more event types 184, 190 for the first event.

[0202] As described with respect to FIG. 1B, the cytometry data 106 may include cytometry data for each of a plurality of cells and particles processed using the cytometry platform 104. For example, the cytometry data 106 includes first cytometry data 132-1 for a first event (e.g., “Event 1” in FIGS. 1B-1C). FIG. 1E illustrates processing the first cytometry data 132-1 to determine one or more types for the first event. However, it should be understood that the example technique 180 may be used to process cytometry data for any suitable number of events, such as second cytometry data for a second event (e.g., “Event 2” in FIGS. 1B-1C), as aspects of the technology described herein are not limited to processing cytometry data for any particular number of events.

[0203] In some embodiments, technique 180 includes processing the first cytometry data 132-1 with a machine learning model hierarchy (e.g., in event type determination module 172 of FIG. 1D). FIG. 1E shows an example of a machine learning model hierarchy that includes machine learning models 182a-c, 186a-b, 188.

[0204] In some embodiments, the machine learning model may be trained to determine whether a first event is of a particular type based on the first cytometry data 132-1. In some embodiments, this may include determining a probability that the first event is of a particular type. As an example, the machine learning model 182a may be trained to determine whether a first event is of type A. As another example, the machine learning model 186b may be trained to determine whether a first event is of type E.

[0205] Additionally or alternatively, the machine learning model may include a multi-class classifier trained to determine whether the first event is one of a plurality of different event types based on the first cytometry data 132-1. For example, the machine learning model A 182a may be trained to determine whether the first event is type A1, type A2, or type A3. For example, the machine learning model may output the most probable type (e.g., type A1, type A2, or type A3) for the first event. Such a machine learning model may output the type and / or the probability that the event is of the identified type. For example, the machine learning model A 182a may identify that the event is more likely to be type A2 than type A1 or type A3, along with the probability that the event is type A2.

[0206] In some embodiments, different levels of the machine learning model hierarchy may be used to determine the event type with different levels of specificity. For example, machine learning models 182a-c may be used to determine that a first event is of type B 184, while machine learning models 186a-b may be used to determine that the first event is of type E 190, which is a subtype 184 of type B.

[0207] In some embodiments, the output of the machine learning models 182a-c is used to provide information about which one or more machine learning models of the hierarchy will be subsequently used to process the first cytometry data 132-1. For example, the output of the machine learning models 182a-c may indicate which event type associated with each of these models has the highest probability of being the event type for the first event. As shown in this example, the output of the machine learning models 182a-c indicates that the first event is type B 184. The technique 180 may continue to determine whether the first event is a subtype of type B 184 based on this output. Thus, in some embodiments, the machine learning models 186a-b trained to determine whether an event is a subtype of type B 184 may be used to process the first cytometry data 132-1.

[0208] In some embodiments, a level of the machine learning model hierarchy may not indicate any type for the first event. For example, the level of the hierarchy that includes the machine learning model 188 does not indicate a type for the first event. In some embodiments, this may indicate that none of the machine learning models at that level of the hierarchy predicted the first event to be the specific event type associated with the machine learning model (e.g., what the machine learning model was trained to determine). For example, the machine learning model 188 predicted that the first event was not type F. In some embodiments, if a level of the hierarchy does not indicate an event type, then the event type indicated at the previous hierarchical level may be determined to be the type for the first event. For example, type E 190 may be determined as the type for the first event. In this case, type E 190 corresponds to the most specific type for the first event because type E 190 is a subtype of type B 184.

[0209] As a non-limiting example of technique 180, first cytometry data 132-1 may be provided to three machine learning models (e.g., machine learning models 182a-c) at a first level of a hierarchy. The three machine learning models 182a-c may be trained to determine the probability that the first event is a monocyte (e.g., type A), a lymphocyte (e.g., type B), or a granulocyte (e.g., type C), respectively. Based on the output of the machine learning models 182a-c, it is determined that the first event is most likely a lymphocyte. Technique 180 may then proceed to determine whether the first event is a subtype of lymphocyte. For example, this may include processing the first cytometry data 132-1 using machine learning models 186a-b, which may be trained to determine whether the first event is a T cell (e.g., type D) or a B cell (e.g., type E), respectively. As shown, based on the output of the machine learning models 186a-b, the first event is determined to be a T cell. The machine learning model 188 is then used to process the first cytometry data 132-1 to determine whether the first event is a CD4 T cell (e.g., type F), which is a subtype of T cell. Because the output of the machine learning model 188 indicates that the first event is unlikely to be a CD4 T cell, the technique 180 ends. As a result, the first event is determined to be a T cell.

[0210] 2A is a flow chart depicting an example process 200 for determining a respective type for one or more cells using cytometry data and a machine learning model hierarchy in some embodiments of the technology described herein. Process 200 may be performed on a laptop computer, a desktop computer, one or more servers, a cloud computing environment, by a computing device 108 as described herein with respect to FIG. 1D, a computing device 1000 as described herein with respect to FIG. 10, or in any other suitable manner.

[0211] Process 200 begins at operation 202, where cytometry data is obtained for a biological sample from a subject. In some embodiments, obtaining the cytometry data includes obtaining the cytometry data from a subject in a clinical or research setting and / or from a data store storing such information. In some embodiments, the biological sample includes a plurality of cells, such as a first cell and a second cell. In some embodiments, the biological sample may additionally include particles, such as debris or beads. For example, beads may be added to the biological sample to determine the cellular composition percentage of different cell types in the sample, as described herein, including with respect to at least FIG. 5C and FIG. 6C. In some embodiments, the cytometry data includes first cytometry data for the first cell. For example, the first cytometry data may include data indicating a value of a marker for the first cell. In some embodiments, the cytometry data includes cytometry data for particles and / or doublets (e.g., when two or more cells are processed as a single cell) in the biological sample. Examples of cytometry data are described herein, including at least the "Flow Cytometry" and "Mass Cytometry" sections.

[0212] In process 204, the cytometry data is processed using one or more data processing techniques. In some embodiments, this may include applying a non-linear transformation to the cytometry data. For example, this may include applying a hyperbolic sine function to transform the cytometry data, such as the function shown in Equation 1.

number

[0213] Additionally or alternatively, in some embodiments, processing the data may include normalizing the data to account for asymmetric data distribution. For example, this may include applying a skewness-adjusted normalization. In some embodiments, this may further include scaling the data. For example, M. Hubert and E. Vandervieren ("An adjusted boxplot for skewed distributions," in Computational Statistics&Data Analysis, vol.52, no.12, pp.5186-5201, 2008), which is incorporated herein by reference in its entirety, describe examples of scaling techniques. However, it should be understood that any suitable data processing technique may be used to process the cytometry data, such that embodiments of the technology described herein are not limited to any particular data processing technique.

[0214] Process 200 then proceeds to operation 206, where a respective type is determined for each of at least a portion of the plurality of cells using a machine learning model hierarchy corresponding to the cell type hierarchy. In some embodiments, the machine learning model hierarchy may include multiple machine learning machines, each of which is trained to determine whether a cell is of a particular type or subtype. For example, each machine learning model may be trained to determine the probability that a cell is of a particular type or subtype. Table 1 shows a non-limiting example of a cell type hierarchy. The machine learning models in the machine learning model hierarchy may include one or more of the machine learning models described in the "Machine Learning" section.

[0215] In some embodiments, a hierarchy of machine learning models may include different levels of machine learning models. For example, a machine learning model at one level of the hierarchy may be trained to determine a relatively general cell type for a cell, while a machine learning model at a subsequent level may be trained to determine a more specific cell type (e.g., subtype) for a cell. As a non-limiting example, one level of the hierarchy may be trained to determine whether a cell is a lymphocyte or monocyte, while a subsequent level of the hierarchy may determine whether a cell is a subtype of lymphocyte or monocyte.

[0216] In some embodiments, determining a type for a particular cell may include processing the cytometry data for the cell using a subset of the machine learning models included in the hierarchy. Process 206a includes determining a first type for a first cell by processing the first cytometry data using a first subset of the machine learning model hierarchy. In some embodiments, the machine learning models included in the first subset are identified during the performance of process 200. After determining a first type for the first cell, process 200 proceeds to process 206b, which determines whether the cytometry data includes cytometry data for another cell of the plurality of cells. For example, this includes determining whether the cytometry data includes second cytometry data for a second cell. If process 206b determines that the cytometry data includes cytometry data for another cell, then process 200 returns to process 206a, where a type is determined for the cell by processing the cytometry data (e.g., the second cytometry data) using a subset (e.g., the second subset) of the machine learning model hierarchy. If, in operation 206b, it is determined that the cytometry data does not include cytometry data for another cell of the plurality of cells, then the process 200 ends.

[0217] 2B illustrates an implementation 206a of a process for determining a first type for a first cell, beginning with process 222, which includes processing first cytometry data with one or more machine learning models at a first level of a hierarchy of machine learning models.

[0218] In some embodiments, the output of the one or more machine learning models indicates the probability that a cell is one or more cell types. For example, a machine learning model at a first level of the hierarchy may be trained to determine the probability that a cell is type A. In some embodiments, process 224 includes determining whether any of the outputs (e.g., probabilities) exceed a specified threshold. If at least one of the outputs does not exceed the threshold, then the implementation may end. For example, if none of the outputs exceed the threshold, then the implementation may end.

[0219] If at least one of the outputs exceeds a threshold in operation 224, the implementation proceeds to operation 226, where a first cell type is determined for the first cell based on the output of one or more machine learning models at a first level of the machine learning model hierarchy. In some embodiments, this may include identifying an output that indicates the highest probability that the cell is of a particular type. For example, the output of a first machine learning model at the first level of the hierarchy may indicate that there is a 70% probability that the cell is of type A, while a second machine learning model at the first level of the hierarchy may indicate that there is a 30% probability that the cell is of type B. In this example, type A may be identified as the first cell type for the first cell because type A corresponds to the highest relative probability.

[0220] In operation 228, process implementation 206a includes determining whether the first cytometry data for the first cell should be processed using another machine learning model in the machine learning model hierarchy. For example, this may include determining whether the first type determined for the first cell has any subtypes. For example, if the first cell type is a lymphocyte, then operation 228 may include determining that T cells and B cells are subtypes of lymphocytes.

[0221] Additionally or alternatively, in process 228, determining whether to process the first cytometry data for the first cell using another machine learning model may be based on an input (e.g., user input) that indicates to what degree of specificity the cell type should be determined for the first cell. For example, the input may indicate that the cell type should be determined broadly or finely for the first cell. In some embodiments, the fewer levels of the machine learning model hierarchy that processes the cytometry data, the broader the classification that is obtained, while the more levels of the machine learning model hierarchy that processes the cytometry data, the more fine the classification that is obtained.

[0222] If in operation 228 it is determined that another machine learning model should not be used, then the implementation 206a ends. If in operation 228 it is determined that another machine learning model should be applied, then the implementation 206a proceeds to operation 230.

[0223] In process 230, one or more machine learning models in a second level of the machine learning model hierarchy are identified based on the first cell type determined for the first cell. In some embodiments, this may include identifying a machine learning model trained to determine whether the cell is a subtype of the first type. For example, if in process 226 the first cell type determined for the cell is a memory B cell, then in process 230 a machine learning model trained to determine whether the cell is a class switched memory B cell and a machine learning model trained to determine whether the cell is an unswitched memory B cell may be identified.

[0224] Implementation 206a then returns to operation 222, where the first cytometry is processed using one or more machine learning models identified in operation 230.

[0225] Although only the first and second levels of the machine learning model hierarchy are described with respect to implementation 206a, it should be understood that the machine learning model hierarchy may include any suitable number of machine learning models and any suitable number of levels, such that aspects of the technology described herein are not limited to any particular number of machine learning models or to any particular number of machine learning model levels. For example, the hierarchy may include a third level of machine learning models, each of which is trained to determine cell type with a higher degree of specificity than the second level. Additionally or alternatively, the machine learning model hierarchy may include only a single level of machine learning models used to determine the type for the first cell.

[0226] 2C is a flow chart of an exemplary process 250 for determining a type for an event corresponding to an object in a biological sample under measurement by a cytometry platform according to some embodiments of the technology described herein. Process 200 may be implemented on a laptop computer, a desktop computer, one or more servers, a cloud computing environment, by a computing device 108 as described herein with respect to FIG. 1D, a computing device 1000 as described herein with respect to FIG. 10, or in any other suitable manner.

[0227] In operation 252, the processor obtains cytometry data for a biological sample previously obtained from a subject. In some embodiments, obtaining the cytometry data includes obtaining the cytometry data from a subject in a clinical or research setting and / or from a data store storing such information. In some embodiments, the biological sample includes a plurality of objects, including a plurality of cells. In some embodiments, the plurality of objects may additionally include particles, such as debris and / or beads. For example, beads may be added to the biological sample to determine the cellular composition percentages of different cell types in the sample, as described herein, including with respect to at least FIG. 5C and FIG. 6C.

[0228] In some embodiments, the cytometry data includes cytometry measurements obtained during each cytometry event. As described herein, a cytometry event corresponds to an object (e.g., a cell, debris, bead, doublet, or unknown object) under measurement by a cytometry platform (e.g., a flow cytometry platform or a mass cytometry platform). In some embodiments, the cytometry event includes an event subset that corresponds to cells in a biological sample under measurement by the cytometry platform. For example, the event subset may include one, some, or all of the cytometry events. The event subset may include any suitable number of events, as aspects of the technology described herein are not limited in this respect. As non-limiting examples, the event subset may include at least 5,000 events, at least 10,000 events, at least 20,000 events, at least 50,000 events, at least 100,000 events, at least 500,000 events, at least 600,000 events, at least 900,000 events, between 500 events and 1 million events, between 5,000 events and 900,000 events, or between 20,000 events and 700,000 events.

[0229] In some embodiments, measurements obtained during each of the event subsets are included in the cytometry data obtained in process 252. For example, measurements obtained during a first event of the event subset may be included in the cytometry data, where the first event corresponds to a cell in the biological sample under measurement by the cytometry platform. Examples of cytometry data are described herein, including at least in the "Flow Cytometry" and "Mass Cytometry" sections.

[0230] In operation 254, the processor obtains a respective plurality of cell types by identifying cell types in the plurality of cells using a plurality of machine learning models. For example, the plurality of machine learning models includes a first machine learning model and a second machine learning model different from the first machine learning model. In some embodiments, the first machine learning model and the second machine learning model are arranged in a hierarchy of machine learning models. In one non-limiting example, the machine learning model hierarchy includes (a) a first level including a first machine learning model trained to predict an event type for an event and (b) a second level including a second machine learning model trained to predict a cell type for an event. In one embodiment of this example, the first and second machine learning models are both multi-class classifiers. In another embodiment of this example, the first machine learning model is a multi-class classifier and the second machine learning model includes a plurality of binary class classifiers each trained to determine whether an event is a respective cell type.

[0231] In some embodiments, identifying a cell type using multiple machine learning models includes performing operations 254-1, 254-2, 254-3, and 264-4 for each particular event included in the event subset.

[0232] In operation 254-1, the processor obtains cytometry measurements corresponding to a particular event, which in some embodiments includes obtaining the cytometry measurements from the cytometry data obtained in operation 252. For example, the cytometry measurements may include marker values ​​(e.g., fluorescence intensity values, intensities of heavy metal ion tags, etc.) obtained during a particular event.

[0233] In operation 254-2, the processor processes the cytometry event by processing the cytometry measurements corresponding to the particular event using a first machine learning model to determine an event type for the particular event. In some embodiments, the event type indicates whether the particular event corresponds to a cell under measurement by the cytometry platform, debris under measurement by the cytometry platform, or a bead under measurement by the cytometry platform. Additionally or alternatively, the event type may indicate whether the particular event corresponds to multiple cells (e.g., doublets) under measurement by the cytometry platform. Additionally or alternatively, the event type may indicate whether the particular event corresponds to an unknown object in the biological sample.

[0234] The first machine learning model may include any suitable machine learning model configured to predict an event type for an event. For example, the first machine learning model may include a multi-class classifier trained to predict an event type from among multiple event types. Additionally or alternatively, the first machine learning model may include one or more binary class classifiers each trained to predict whether an event is of a particular event type. The first machine learning model may be trained to predict the probability that an event is of a particular event type. The first machine learning model may include any suitable machine learning model, such as, for example, a decision tree classifier, a gradient boosting decision tree classifier, a neural network, or any other suitable type of machine learning model, as aspects of the techniques described herein are not limited in this respect. The first machine learning model may include one or more of the machine learning models described in the "Machine Learning" section.

[0235] In operation 254-2, when the determined event type indicates that a particular event corresponds to a cell being measured by the cytometry platform, the processor determines a type for the cell by processing the cytometry measurements corresponding to the particular event using a second machine learning model. For example, the processor may determine whether the cell includes one or more of the cell types (or cell subtypes) listed in Table 1.

[0236] The second machine learning model may include one or more machine learning models configured to predict a cell type for an event. For example, the second machine learning model may include at least one multi-class classifier trained to predict a cell type for a particular event from among a plurality of cell types. Additionally or alternatively, the second machine learning model may include one or more binary class classifiers each trained to predict whether an event is a particular cell type. The second machine learning model may be trained to predict the probability that an event is a particular cell type. The second machine learning model may include any suitable machine learning model, such as, for example, a decision tree classifier, a gradient boosting decision tree classifier, a neural network, or any other suitable type of machine learning model, as aspects of the technology described herein are not limited in this respect. The second machine learning model may include one or more of the machine learning models described in the "Machine Learning" section.

[0237] In operation 254-4, after determining the cell type for a particular event, the processor determines whether the event subset includes another event. If the event subset includes another event, one or more of operations 254-1, 254-2, 254-3, and 254-4 may be repeated for that event. For example, the processor may use a cytometry measurement corresponding to the next event to determine the event type and / or cell type for the event. If in operation 254-4 the processor determines that the event subset does not include another event, then process 250 ends.

[0238] It should be understood that process 250 may include one or more additional or alternative processes (not shown in FIG. 2C). For example, process 250 may include one or more processes for processing the cytometry data prior to process 254. For example, process 250 may include at least process 204 described herein, including with respect to FIG. 2A, for processing the cytometry data.

[0239] In some embodiments, the cell types identified as a result of process 250 are used to determine a cellular composition percentage of different cell types in the biological sample, as described in more detail herein. For example, determining a first cellular composition percentage for a first cell type may include determining a ratio between the number of cells identified as being of the first type and the total number of cells. Examples of techniques for determining cellular composition percentages are described herein, including with respect to at least Figures 5A-6C.

[0240] [Table 1]

[0241] [Table 2]

[0242] [Table 3]

[0243] [Table 4]

[0244] 3A shows an example diagram of determining the type of an event based on the output of a binary class classifier, according to some embodiments of the techniques described herein. As shown, an event is determined to be of type A in process 310. The technique 300 includes determining whether the event is of type A1 or type A2, both of which are subtypes of type A.

[0245] In some embodiments, the machine learning model A1 302 may be trained to estimate the probability that an event is of type A1. The machine learning model A1 302 may include a decision tree classifier, a gradient boosting decision tree classifier, a neural network, or any other suitable type of machine learning model, as aspects of the technology described herein are not limited in this respect. In some embodiments, the machine learning model A1 302 may include an ensemble of any suitable type of machine learning model. For example, the machine learning model A1 302 may include an ensemble of decision tree classifiers, an ensemble of gradient boosting decision tree classifiers, or an ensemble of neural networks. The machine learning model A1 302 may include one or more of the machine learning models described herein, including in the "Machine Learning" section.

[0246] In operation 304, the cytometry data for the event is processed using machine learning model A1 302 to determine if the event is of type A1. In some embodiments, this may include determining if the probability predicted by the first machine learning model exceeds a threshold. For example, the threshold may be 0.2, 0.3, 0.5, 0.6, 0.7, or any suitable threshold, as aspects of the technology described herein are not limited to any particular threshold.

[0247] If the process 304 determines that the probability does not exceed the threshold, then the event is identified 308 as not being of type A1. In contrast, if the process 304 determines that the probability does exceed the threshold, then the event is identified 306 as being of type A1.

[0248] In some embodiments, machine learning model A2 322 may be trained to estimate the probability that an event is of type A2. Machine learning model A2 322 may include a decision tree classifier, a gradient boosting decision tree classifier, a neural network, or any other suitable type of machine learning model, as aspects of the technology described herein are not limited in this respect. In some embodiments, machine learning model A2 322 may include an ensemble of any suitable type of machine learning model. For example, machine learning model A2 322 may include an ensemble of decision tree classifiers, an ensemble of gradient boosting decision tree classifiers, or an ensemble of neural networks.

[0249] In operation 324, the cytometry data for the event is processed using machine learning model A2 322 to determine if the event is of type A2. In some embodiments, this may include determining if the probability predicted by the first machine learning model exceeds a threshold. For example, the threshold may be 0.2, 0.3, 0.5, 0.6, 0.7, or any suitable threshold, as aspects of the technology described herein are not limited to any particular threshold.

[0250] If the process 324 determines that the probability does not exceed the threshold, then the event is identified 328 as not being of type A2. In contrast, if the process 324 determines that the probability does exceed the threshold, then the event is identified 326 as being of type A2.

[0251] According to some embodiments, machine learning model A1 302 may output 306 that the event is of type A1, while machine learning model A2 322 may output 328 that the event is not of type A2. In this case, type A1 is identified for the event.

[0252] Similarly, in some embodiments, machine learning model A2 322 may output 326 that the event is of type A2, while machine learning model A1 302 may output 308 that the event is not of type A1. In this case, type A2 is identified for the event.

[0253] Alternatively, in some embodiments, machine learning model A1 302 may output 306 that the event is of type A1, and machine learning model A2 322 may output 326 that the event is of type A2. In some embodiments, to determine the type for an event, the type with the highest probability is selected. For example, if machine learning model A1 302 outputs that the event is of type A1 with a probability of 0.8, and machine learning model A2 322 outputs that the event is of type A2 with a probability of 0.9, then type A2 may be identified for the event.

[0254] Alternatively, in some embodiments, machine learning model A1 302 may output 308 that the event is not of type A1, and machine learning model A2 322 may output 328 that the event is not of type A2. In this case, no type is identified for the event.

[0255] 3B shows an example diagram of determining the type of an event based on the output of a multi-class classifier, according to some embodiments of the techniques described herein. As shown, an event is determined to be of type A in process 352. The technique 350 includes determining whether the event is of type A1 or type A2, both of which are subtypes of type A.

[0256] In some embodiments, the machine learning model 354 may be trained to predict whether an event is type A1, type A2, or neither type A1 nor type A2. The machine learning model 352 may include a decision tree classifier, a gradient boosting decision tree classifier, a neural network, or any other suitable type of machine learning model, as aspects of the technology described herein are not limited in this respect. In some embodiments, the machine learning model 352 may include an ensemble of any suitable type of machine learning model. For example, the machine learning model 352 may include an ensemble of decision tree classifiers, an ensemble of gradient boosting decision tree classifiers, or an ensemble of neural networks.

[0257] In process 356, the cytometry data for the event is processed using machine learning model 354 to determine whether the event is of type A1 or type A2. In some embodiments, this may include determining which type the event has a higher probability of being.

[0258] Additionally or alternatively, this may include determining whether the probability is greater than a threshold, for example, the threshold may be 0.2, 0.3, 0.5, 0.6, 0.7, or any suitable threshold, as aspects of the technology described herein are not limited to any particular threshold.

[0259] If process 356 determines that the probability of the event being type A1 exceeds (a) a threshold and (b) the probability of the event being type A2, then type A1 is output 358. If process 356 determines that the probability of the event being type A2 exceeds (a) a threshold and (b) the probability of the event being type A1, then type A2 is output 360. If process 356 determines that neither probability exceeds the threshold, then neither type A1 nor type A2 is output 362.

[0260] FIG. 4 depicts an illustrative example of determining one or more types of an event based on a hierarchy 400 of event types according to some embodiments of the technology described herein.

[0261] In some embodiments, the different event types shown in hierarchy 400 represent possible types for event 402. In some embodiments, the event types may correspond to a machine learning model trained to determine whether event 402 is a particular event type based on cytometry data for event 402. For example, the machine learning model may be trained to determine the probability that event 402 is a white blood cell 404c and whether that probability exceeds a threshold. Here, the machine learning model corresponding to white blood cell 404c outputs that the probability that event 402 is a white blood cell 404c is 0.99, which exceeds an example threshold of 0.5.

[0262] In some embodiments, in addition to processing the cytometry data for the event 402 with a machine learning model trained to determine if the event 402 is a white blood cell 404c, the techniques may also include processing the cytometry data with a machine learning model trained to determine if the event 402 is debris 404a and a machine learning model trained to determine if the event 402 is a bead 404b. Here, the machine learning models output probabilities of 0.01 and 0.05, respectively. Because these probabilities are below the example threshold of 0.5, it is determined that the event 402 is not a bead 404a or debris 404b.

[0263] The probability that the event 402 is a white blood cell 404c is higher than the threshold and any other probability output at level 404, and therefore the event 402 is identified as a white blood cell 404c at this level.

[0264] As shown in this hierarchy, level 404 corresponds to the broadest classification of event 402. Based on the output of level 404, it may be possible to determine a more specific type for event 402. For example, because event 402 is determined to be a white blood cell 404c, it may be possible to determine whether event 402 is a specific subtype of white blood cell 404c. In particular, level 406 corresponds to several event types that are subtypes of white blood cell 404c (e.g., monocyte 406a, granulocyte 406b, lymphocyte 406c, DC 406d, and macrophage 406e).

[0265] To determine whether the event 402 is a subtype of white blood cell 404c, the cytometry data for the event 402 may be processed with a machine learning model trained to determine whether the event 402 is one of the subtypes 406a-e, as was done at level 404. Here, the probability that the event 402 is a lymphocyte 406c (e.g., 0.88) and the probability that the event 402 is a macrophage 406e (e.g., 0.7) each exceed a threshold example (e.g., 0.5). Because the probability that the event 402 is a lymphocyte 406c (e.g., 0.88) is higher than the probability that the event is a macrophage 406e (e.g., 0.7), the event is identified as a lymphocyte 406c.

[0266] Level 408 corresponds to a cell type that is a subtype 408a-c of lymphocyte 406c. Again using the same processing techniques, the probability that event 402 is each of subtypes 408a-c may be determined. Here, T cell 408c is identified as the type for event 402 with a probability of 0.79, which is higher than a threshold example (e.g., 0.5) and higher than both the probability that event 402 is an NK cell 408a (e.g., 0.45) and the probability that event 402 is a B cell 408b (e.g., 0.6).

[0267] Level 410 corresponds to an event type that is a subtype 410a-c of T cell 408c. Again using the same processing techniques, the probability of event 402 being each of subtypes 410a-c may be determined. Here, since none of the determined probabilities (e.g., 0.4, 0.02, or 0.23) are higher than a threshold example (e.g., 0.5), no event type is determined at this level 410 for event 402. Thus, event type determination is complete.

[0268] Here, three types ranging from least specific to most specific have been determined for the event 402: white blood cells, lymphocytes, and T cells. In some embodiments, the technique may output one, some, most, or all of the types determined for the event 402.

[0269] Cellular composition and patient cohorts FIG. 5A illustrates a method for identifying subjects as members of a patient cohort according to some embodiments of the technology described herein. 1 is a flowchart of an example process 500 for identifying a user as a member of a network that is ...

[0270] Process 500 begins with operation 502, in which cytometry data is obtained for a biological sample from a subject, the biological sample including a plurality of cells. In some embodiments, operation 502 may be performed according to techniques described herein, including at least with respect to operation 202 of process 200 and / or operation 252 of process 250.

[0271] In process 504, a respective type is determined for each of at least a portion of the plurality of cells. In some embodiments, process 504 may be performed according to the techniques described herein, including with respect to at least Figures 2A-2C, for determining types for cells in a biological sample. Additionally or alternatively, process 504 may include determining cell types for cells in the biological sample by processing the cytometry data in any suitable manner, as aspects of the techniques described herein are not limited to any particular cytometry data processing technique.

[0272] In process 506, a cellular composition percentage is determined for each of at least some of the determined cell types. In some embodiments, determining the cellular composition percentage of a cell type for a biological sample may include determining a ratio between the number of cells of a particular type and the total number of cells for which cytometry data was obtained. For example, FIG. 6A shows an example of cytometry data for a biological sample in which cells 1-5 are typed. Determining the cellular composition percentage for type A cells may include determining a ratio between the number of type A cells and the total number of cells. In this example, the cellular composition percentage of type A cells may therefore be 2 / 5.

[0273] In some embodiments, cytometric data may be obtained using different marker panels for multiple sub-samples of a biological sample. As described herein, a marker panel may be used to identify a certain cell type or types, but may not be able to distinguish all cell types in the sub-samples. Thus, a first panel may be used to determine the number of type A cells, along with the number of unknown cells, while a second panel may be used to determine the number of type B cells, along with the number of unknown cells. The first panel may include type B cells, but they may not be identified in the cytometric data associated with the panel (e.g., included in the unknown cells). Similarly, the second panel may include type A cells, but they may not be identified in the cytometric data associated with the panel. Thus, it may be a challenge to determine the total number of cells of a certain type (e.g., type A, type B) in a biological sample and the corresponding cellular composition percentage.

[0274] Thus, in some embodiments, process 506 may include performing one or more normalization techniques to account for data from individual panels. Figures 5B and 5C are flow charts illustrating an implementation of process 506.

[0275] FIG. 5B is a flowchart depicting an example implementation 506 of a process 500 for determining a cellular composition percentage based on an estimated cellular composition percentage of common cell types, according to some embodiments of the technology described herein.

[0276] This implementation begins with an operation 522 of estimating a percent cellular composition for each of at least a portion of a first plurality of cell types included in a first sub-sample of the biological specimen. In some embodiments, the first plurality of cell types may be associated with a first panel of cytometry data obtained for the first sub-sample of the biological specimen. For example, as shown in FIG. 6B, this may include estimating a percent cellular composition 608 for cell types associated with panel A 606.

[0277] Process 522a includes estimating a first cellular composition percentage for a first cell type in the first plurality of cells. In some embodiments, estimating the first cellular composition percentage may include determining a ratio between the number of cells of the first type and the total number of cells in the first subsample. In some embodiments, the first cell type may be considered a "reference cell type" and the cellular composition percentage may be considered a "reference cellular composition percentage." For example, as shown in FIG. 6B, this may include determining a cellular composition percentage 608 for cell type A.

[0278] In process 524, the implementation 506 includes estimating a cellular composition for each of at least a portion of a second plurality of cell types included in a second subsample of the biological specimen. In some embodiments, the second plurality of cell types may include cells of a first type (e.g., a reference cell type) and may be associated with a second panel of cytometry data obtained for the second subsample of the biological specimen. For example, as shown in FIG. 6B, panel B 610 includes data for a plurality of cell types, including cell type A. A cellular composition percentage 612 may then be determined for the cell types associated with panel B 610.

[0279] Process 524a includes determining the number of cells in the second subsample that are of the first cell type. For example, with reference to Figure 6, this may include determining that there are three cells of type A associated with panel B 610.

[0280] Process 524b includes estimating a cellular composition percentage for at least a portion of the plurality of cell types based on the first estimated cellular composition percentage (e.g., reference cellular composition percentage) and the number of cells of a first type (e.g., reference cell type) in the second subsample. In some embodiments, this may include normalizing the cellular composition percentage for the cell type based on the first and second estimated cellular composition percentages. For example, a cellular composition percentage (CCP) for cell type N may be estimated according to Equation 2:

number

[0281] The implementation of process 506 then proceeds to process 526, which determines whether there are additional cell types (e.g., a third subtype) associated with another sub-sample (e.g., a third sample). If there are additional cell types, then the implementation 506 returns to process 524, which estimates the cellular composition percentages for the multiple cell types. If there are not additional cell types, then the implementation proceeds to process 528.

[0282] Processing 528 includes determining a single cellular composition percentage for each cell type. For example, if different sub-samples contain cells of the same type and a cellular composition percentage for that cell type has been determined for both sub-samples, then a single cellular composition percentage may be determined for that cell type. In some embodiments, this may include averaging the cellular composition percentages. Additionally or alternatively, this may include selecting one of the cellular composition percentages. As a non-limiting example, consider FIG. 6B. Panel A 606, corresponding to a first sub-sample, and Panel B 610, corresponding to a second sub-sample, each contain cells of type A and cells of type E. A cellular composition percentage 608, 610 is determined for each type (e.g., cell type A1%, cell type A2%, cell type E1%, and cell type E2%). These cellular composition percentages are then averaged 614 to obtain a cellular composition percentage for each cell type for the biological sample.

[0283] FIG. 5C is a flowchart depicting a processing implementation 506 of process 500 for determining cellular composition percentage based on bead percentage according to some embodiments of the technology described herein.

[0284] Processing 542 includes determining the number of beads in a subsample of the biological sample. For example, as shown in FIG. 6C, this may include determining the number of beads (e.g., 2) in the subsample associated with panel A 622.

[0285] Process 544 includes estimating a cellular composition percentage for each of at least some of the cell types contained in the subsample based on the number of beads determined in process 542.

[0286] In some embodiments, this may include determining the number of cells of a first type in the sub-sample in process 544a. For example, as shown in Figure 6C, panel A 622 associated with the first sub-sample contains one cell of type A.

[0287] In operation 544b, the number of cells of the first type (e.g., cell count) is normalized with respect to the number of beads. In some embodiments, the cell count is normalized according to Equation 3:

number

[0288] Process 544c includes estimating the cellular composition percentage for the first cell type, which in some embodiments includes determining a ratio between the normalized cell count and the total number of cells in the biological sample.

[0289] As a non-limiting example of process 544, consider a biological sample partitioned into sub-samples A and B associated with panels A 622 and B 626, respectively, as shown in FIG. 6C. To estimate the percentage cellular composition for cell type C associated with panel A 622, the number of type C cells (e.g., 1 cell) may be normalized with respect to the number of beads (e.g., 2 beads) and bead concentration (e.g., in units of 5,000 beads per million cells) associated with panel A 627 to determine a normalized cell count. To estimate the percentage cellular composition of cell type C for sub-sample A, the ratio between the normalized cell count and the total number of cells in the biological sample (e.g., 7) may be determined.

[0290] The process implementation 506 then proceeds to process 546 to determine whether there is another sub-sample of the biological sample. If there is another sub-sample, then the implementation 506 returns to process 542 to determine the number of beads in the sub-sample. If there are not another multiple cell types, then the implementation proceeds to process 548.

[0291] Process 548 includes determining a single cell composition percentage for each cell type. In some embodiments, the techniques may include at least those described herein, including with respect to process 528.

[0292] Returning now to FIG. 5A, after determining the respective cellular composition percentages for at least some of the cell types of the biological sample, process 500 proceeds to operation 508.

[0293] Processing 508 includes normalizing the cell composition percentages with respect to the hierarchical relationships between cell types. As described above, some techniques include determining the cell composition percentages for different levels of the cell type hierarchy. The cell composition percentage of a more general cell type (e.g., a "parent" cell type) should theoretically be equal to the sum of its "descendant" cell types. For example, B cells, T cells, and NK cells are subtypes of lymphocytes. Therefore, the sum of the cell composition percentages for these types should be equal to the cell composition percentage determined for lymphocytes.

[0294] However, in some embodiments, the sum of the estimated cellular composition percentages of the progeny cell types may exceed the estimated cellular composition percentage of the parent cell type. Accordingly, techniques are described herein, including with respect to FIG. 5D, for normalizing the estimated cellular composition percentages such that the sum of the estimated cellular composition percentages of the progeny cell types does not exceed the cellular composition percentage of the parent cell type.

[0295] In some embodiments, there may be challenges associated with normalizing the cellular composition percentages with respect to the hierarchical relationships between cell types. In particular, such challenges may arise when determining the cellular composition percentages based on data from multiple different sub-samples, as described herein, including with respect to at least Figures 5B-5C. Thus, the techniques described herein, including with respect to Figure 5D, include techniques for normalizing the cellular composition percentages when there are multiple sub-samples.

[0296] For example, consider a first sub-sample containing cells of type A1 and type A2, and a second sub-sample containing cells of type A3 and type A4, where each of cell types A1, A2, A3, and A4 is a subtype of type A. For the first sub-sample, the percentage of cells of type A should be equal to the sum of the percentage of cells of type A1 and the percentage of cells of type A2 (e.g., type A=type A1+type A2). For the second sub-sample, the percentage of cells of type A should be equal to the sum of the percentage of cells of type A3 and the percentage of cells of type A4 (e.g., type A=type A3+type A4). However, when data from different sub-samples are combined to determine the cellular composition percentage for a biological sample, the combined cellular composition percentage of type A is not equal to the sum of the cellular composition percentages of types A1, A2, A3, and A4 (e.g., type A ≠ type A1 + type A2 + type A3 + type A4). Thus, when normalizing according to the techniques described with respect to FIG. 5D, the subtypes of different sub-samples can be treated independently of each other.

[0297] FIG. 5D is a flowchart depicting a processing implementation 508 of the process 500 for normalizing cellular composition percentages with respect to hierarchical relationships between cell types according to some embodiments of the technology described herein.

[0298] The implementation 508 begins with an process 552 of identifying a set of one or more subtypes of a first cell type for which one or more cellular composition percentages have been estimated. When the cellular composition percentages are estimated from a single panel of cytometry data for a biological sample, there may be only one set of cellular subtypes. For example, for white blood cells, the set of subtypes may include monocytes, granulocytes, lymphocytes, and macrophages. Additionally or alternatively, when the cellular composition percentages are estimated from two or more panels of cytometry data corresponding to multiple subsamples of the biological sample, there may be multiple respective sets of cellular subtypes. For example, for white blood cells, a first set may include monocytes and macrophages, while a second set may include lymphocytes and granulocytes.

[0299] The implementation 508 then proceeds to operation 554, where for a first set of identified cellular composition percentage sets, the technique includes determining a sum of the estimated cellular composition percentages for the subtypes included in the first set. This sum is then compared to the cellular composition percentage of the first cell type in operation 556 to determine whether it exceeds the cellular composition percentage of the first cell type. For example, the sum of the cellular composition percentages of granulocytes and lymphocytes may be compared to the cellular composition percentage of white blood cells.

[0300] If the sum does not exceed the cellular composition percentage of the first cell type, then the implementation 508 proceeds to operation 564, which determines whether there is another set (e.g., a second set) of cell subtypes identified in operation 552. If the sum does exceed the cellular composition percentage of the first cell type, then the implementation proceeds to operation 558.

[0301] Process 558 includes determining whether the set of cell subtypes identified in process 559 includes all possible subtypes of the first cell type. In some embodiments, this may include obtaining data indicative of possible cell types for the first cell type from one or more data stores. For example, if the first cell type is a white blood cell, then process 558 may include accessing a data store to identify possible subtypes of white blood cells and comparing them to previously identified subtypes.

[0302] If in operation 558 it is determined that there may be additional subtypes of the first cell type, then the implementation proceeds to operation 564, which determines whether there is another set (e.g., a second set) of cell subtypes identified in operation 552. If it is determined that there are no other subtypes, then the implementation 508 proceeds to operation 560.

[0303] Process 560 includes determining a normalization factor for the first cell type. In some embodiments, determining the normalization factor includes determining a ratio between the cellular composition percentage of the first cell type and the sum of the cellular composition percentages of the subtypes included in the first set. Additionally or alternatively, the normalization factor may be determined in any suitable manner, as aspects of the present technology are not limited to any particular technique for determining the normalization factor.

[0304] Process 562 includes applying the determined normalization factor to at least a portion of one or more estimated cellular composition percentages for the cell subtypes included in the first set, which may include, for example, multiplying the cellular composition percentages by the normalization factor.

[0305] The implementation then proceeds to operation 564, which determines whether there is another subtype set (e.g., a second set) identified for the first cell type in operation 522. If there is another set, then the implementation returns to operation 554, which determines the sum of one or more estimated cellular composition percentages for the cell subtypes included in the next set. If there is not another set, then the implementation proceeds to operation 566.

[0306] In operation 566, the technique includes determining whether there is another cell type (e.g., a second cell type) to which the normalization techniques described herein may be applied. If there is another cell type, then implementation 508 returns to operation 552. Otherwise, implementation 508 ends.

[0307] 5A, after normalizing one or more cellular composition percentages, process 500 proceeds to operations 510 and / or 512. For example, one or both of operations 510 and / or 512 may be implemented as part of process 500.

[0308] Process 510 includes identifying the subject as a member of a patient cohort based on the determined cell composition percentages. In some embodiments, this may include comparing one or more cell composition percentages to those associated with the patient cohort. As a non-limiting example, this may include comparing the subject's percentage of T cells to the average percentage of T cells in patients who have responded positively to a particular treatment. As another example, this may include comparing the subject's percentage of CD39+ cells to the average percentage of CD39+ cells in patients diagnosed with a particular cancer.

[0309] In some embodiments, identifying a subject as a member of a cohort may be useful in determining a diagnosis, developing a treatment plan, identifying effective drugs, and conducting research, however, it should be understood that this is not an exhaustive list.

[0310] Process 512 includes identifying a treatment for the subject based on the determined cellular composition percentage. The determined cellular composition percentage may serve as a biomarker that can be used to identify a treatment for the subject. For example, the cellular composition percentage of peripheral blood mononuclear cells (PBMCs) may serve as a biomarker to identify ipilimumab as a treatment for a subject having HLA-DRlow monocytes. In some embodiments, if the determined cellular composition percentage of PBMCs for a subject is below a threshold, then ipilimumab may be identified as a treatment for the subject. For example, if the cellular composition percentage of PBMCs is below a threshold of 10%, 11%, 12%, 13%, 13.05%, 13.1%, 13.5%, 14%, 15%, 16%, or below a threshold of 10% to 16%, then ipilimumab may be identified as a treatment for the subject. As another example, the cellular percentage of CD8+PD-1+ cells and CD4+PD-1+ cells may serve as a biomarker to identify immune checkpoint blockade therapy for a subject with non-small cell lung cancer (NSCLC). For example, the ratio of the cellular percentage of CD8+PD-1+ cells to CD4+PD-1+ cells may serve as such a biomarker. In some embodiments, if this ratio exceeds a threshold, then immune checkpoint blockade therapy is identified as a treatment for the subject. For example, if this ratio exceeds a threshold of 1.5, 1.6, 1.7, 1.8, 1.85, 1.89, 1.9, 1.91, 1.92, 1.93, 1.95, 2.0, 2.1, 2.2, 2.3, or a threshold between 1.5 and 2.3, then immune checkpoint blockade therapy may be identified as a treatment for the subject.

[0311] Treatment 514 includes administering to the subject a treatment identified in treatment 514. Techniques for administering treatments are described herein, including at least in the "Methods of Treatment" section.

[0312] In some embodiments, one or more visualizations or reports may be generated using the results of the processes described herein, including with respect to at least Figures 2A-2C and 5A-5D. Figures 7A-7D show examples of visualizations and reports that may be generated as a result of these processes.

[0313] 7A, 7B, and 7C are examples of visualizations of cellular composition percentages according to some embodiments of the technology described herein. In particular, they each show a graph that includes nodes and edges. A node represents a cell population of a particular type, while an edge represents a relationship between two nodes.

[0314] As shown, the nodes are organized and connected in a manner that illustrates a hierarchy of cell types, ranging from general cell populations to more specific subtype populations. Each node displays the name of the cell type, and for some nodes where a cellular composition percentage is available, the cellular composition percentage (e.g., a percentage or fraction) is displayed.

[0315] As shown in Figure 7A, the size of some nodes reflects the relative size (e.g., cellularity percentage) of the cell population they represent. As shown, the white blood cell population has the highest cellularity percentage value, which is reflected in its size.

[0316] In some embodiments, the color or shading of the node may reflect information about one or more cohorts identified for the subject. For example, when the subject's cellular composition percentage value is within a certain percentile boundary within the range of the cohort, the node is shaded gray. The node is shaded red if the value is above the upper percentile boundary and blue if it is below the lower boundary.

[0317] Additionally or alternatively, in contrast to Figure 7A, the size of some nodes may represent cell populations with abnormal percentages of cell composition compared to the patient cohort. For example, Figure 7C shows larger nodes representing abnormal cell populations. The example nodes also include fold over or underrepresentation associated with the population.

[0318] FIG. 7D is a screenshot of an example report showing the evaluation of biomarkers based on the determined cell composition percentages, according to some embodiments. Biomarkers are biological measures that may affect the effectiveness of certain treatments for patients with certain diseases. According to some embodiments, the effectiveness of the treatment is estimated using Kaplan-Meier curves, which show statistics about the overall survival of patients along a time axis. Different curves refer to different biomarker value ranges. According to some embodiments, it is possible to predict the survival of a patient by comparing the biomarker values ​​for the patient to the different biomarker value ranges.

[0319] According to some embodiments, the biomarkers may include one or more metrics indicative of a cellular composition percentage. For example, as shown in the example report, the biomarkers evaluated include the ratio of type X cells to type Y cells in the biological sample. As shown in the example, a "patient measure" refers to the numerical value (e.g., biomarker value) of the ratio of type X cells to type Y cells in the biological sample for the patient. A "measure range" refers to different biomarker value ranges according to a Kaplan-Meier curve. As shown, the patient measure falls within a "high" measure range that indicates a high predicted survival rate.

[0320] The example report shown in FIG. 7D also indicates the diagnosis for the patient, study ID, treatment for the diagnosis, and approval and approval phase information related to the treatment.

[0321] 7E is a screenshot of an example report indicating the percentage of cellular composition in a biological sample from a patient according to some embodiments of the technology described herein. As shown, cell types are grouped based on the common cell type or "cell family" that is a descendant thereof. For example, neutrophils, basophils, and eosinophils belong to the granulocyte cell family and are descendants of granulocytes as shown in Table 1. Similarly, CD16- (classical) monocytes and CD16+ (non-classical) monocytes each belong to the monocyte cell family and are descendants of monocytes as shown in Table 1.

[0322] The example table shown in Figure 7E also shows the cellular composition percentages determined for each cell type in the biological sample for the patient. For example, neutrophils make up 71.5% of the biological sample for this patient.

[0323] The table also provides a range (e.g., upper and lower boundaries) of the cellular composition percentage for the reference population (e.g., reference cohort). For example, the cellular composition percentage range of neutrophils in the reference population is 54.3% to 77.1%. According to some embodiments, the reference data is determined based on a statistical analysis of a reference cohort of donors (e.g., healthy donors, donors diagnosed with a particular disease, etc.). For example, the upper and lower boundaries of the range may be calculated based on the 90th and 10th percentiles of the reference cohort. However, it should be understood that the upper and lower boundaries may be determined using any suitable technique, as aspects of the technology described herein are not limited in this respect.

[0324] In some embodiments, the report may provide an indication that the cellular composition percentage for the patient falls outside the boundaries of the provided cellular composition percentage ranges. For example, as shown in FIG. 7E, the entry for eosinophils is shaded gray, indicating that its cellular composition percentage of 1.74% falls outside the reference range of 1.82% to 6.43%.

[0325] Additionally or alternatively, but not shown, the table may include two reference ranges, for example, a first reference range may indicate a cell composition percentage range for a reference cohort of healthy donors and a second range may indicate a cell composition percentage range for a reference cohort of patients who have been diagnosed with a particular disease.

[0326] FIG. 7F is a screenshot of an example report showing the deviation in cellular composition percentages for cell types that are progeny of MAIT cells. The red bar next to the label "MAIT cells" indicates that the cellular composition percentage of MAIT cells in the biological sample exceeds the median of the reference cohort. For example, the reference cohort may include healthy donors or donors diagnosed with a particular disease. The number adjacent to the red bar indicates that the cellular composition percentage exceeds the median of the reference cohort by 2.2 times.

[0327] The example report also includes cell types that are progeny of MAIT cells and have similarly abnormal cellular percentages. For example, the cellular percentages of MAIT CD8+ cells, which are progeny of MAIT cells, and MAIT CD8+ cells for the biological sample are 2.4-fold greater than the median of the respective reference cohorts. Similarly, the cellular percentages of MAIT CD8+ cells and MAIT CD8+CD27+CD45RA-CD56+CD57+ cells, which are progeny of both MAIT CD8+ cells and MAIT cells, are 11.7-fold greater than the median of the respective reference cohorts.

[0328] FIG. 7G is a screenshot of an example report showing the deviation of the cellular composition percentage values ​​for cell types that are progeny of CD4 T helpers. The blue bars indicate that the cellular composition percentage of the listed cell type is lower than the median of the respective reference cohort. For example, the blue bar next to the display name "CD4 central memory CXCR3-CCR4-CCR6-", which is progeny of CD4 T helpers, indicates that the cellular composition percentage of CD4 central memory CXCR3-CCR4-CCR6- is lower than the median of the respective reference cohort. The number adjacent to the blue bar indicates that the cellular composition percentage is 1.1 times lower than the median of the reference cohort.

[0329] Machine learning model training 8 is a flowchart illustrating an example method for training multiple machine learning models according to some embodiments of the techniques described herein. Process 800 may be performed on a laptop computer, a desktop computer, one or more servers, in a cloud computing environment, by a computing device 108 as described herein with respect to FIG. 1D, a computing device 1000 as described herein with respect to FIG. 10, or in any other suitable manner.

[0330] Process 800 begins with operation 802, where cytometry data is obtained for each of a plurality of cells. In some embodiments, the cytometry data may be obtained in any suitable manner, as aspects of the technology described herein are not limited to any particular technique for obtaining the cytometry data. For example, this may include obtaining the cytometry data by processing one or more biological samples. As another example, this may include obtaining the cytometry data from one or more data stores and / or storage devices.

[0331] Process 804 includes obtaining corresponding cell type data for each of at least a portion of the plurality of cells for which cytometry data was obtained. In some embodiments, process 804 includes sub-process 804a and sub-process 804b.

[0332] Sub-processing 804a includes obtaining a first type for a first cell for which cytometry data was obtained. In some embodiments, this includes processing the obtained cytometry data using gating and / or clustering techniques, such as those described herein, including at least with respect to the "training data." In other embodiments, obtaining a first type for the first cell may include accessing data from one or more data stores that store information indicating a type for the cell for which cytometry data was obtained.

[0333] Sub-process 804b includes extracting one or more cell types related to the first type from a cell type hierarchy. In some embodiments, this may include determining one or more subtypes of the first type and one or more "parent" types of the first type. For example, consider a first cell determined in sub-process 804a to be a T cell. Sub-process 804b may include extracting cell types related to the T cell, such as lymphocytes, leukocytes, memory T cells, to name a few. In some embodiments, sub-process 804b may include accessing a cell type hierarchy (e.g., from a data store), such as the cell type hierarchy described with respect to Table 1.

[0334] Process 806 includes training a machine learning model of the plurality of machine learning models to determine whether a cell is of a particular type using the cytometry data and the cell type data. In some embodiments, the obtained data may be split into training, testing, and validation data sets for training, testing, and validation of each machine learning model. In some embodiments, the training data may be processed according to techniques described herein, including at least with respect to process 204 of process 200, before being used to train, test, or validate a machine learning model.

[0335] In some embodiments, process 806 may include training a machine learning model for each of at least some of the cell types obtained in process 804. Sub-process 806a includes training a first machine learning model to determine whether a cell is of a first type using the first cytometry data and the first cell type data, or "first obtained data." For example, sub-process 806a may include training a machine learning model to determine whether a cell is a T cell.

[0336] In some embodiments, the first obtained data may include "positive" data and "negative" data. In some embodiments, the positive data may include data about cells that would be positively identified as being of the first cell type using the first machine learning model, while the negative data may include data about cells that would not be identified as being of the first cell type using the first machine learning model.

[0337] In some embodiments, the positive data may include cytometry data for cells determined to be of a first type and / or for cells determined to be of a type related to the first type. For example, if the first type is a T cell, then the positive data may include cytometry data for cells determined to be T cells. Additionally or alternatively, if the first type is a T cell, then the positive data may include cytometry data for cells determined to be lymphocytes, leukocytes, memory T cells, or any other type related to the first cell type. In this example, the related cell types include parent cell types and / or subtypes of the first cell type. The parent cell types and subtypes of the first cell type may be located at a different level in the cell hierarchy than the first cell type.

[0338] In some embodiments, the negative data may include cytometry data for cells that are types at the same level in the cell hierarchy and / or subtypes of cell types at the same level in the cell hierarchy. For example, if the first cell type is a T cell, then the cell types at the same level in the cell hierarchy may include B cells, while the subtypes of the cell types at the same level may include memory B cells. Additionally or alternatively, the negative data may include cytometry data for cells whose type cannot be identified. In other embodiments, the data may not include any negative training data for the first cell type.

[0339] In some embodiments, the first cytometry data includes values ​​for specific markers obtained for the cells. In some embodiments, the markers may be selected based on the first cell type for which the first machine learning model is trained. For example, if the first machine learning model is trained to determine whether a cell is a T cell, then the first cytometry data may include values ​​for one or more specific markers that are useful for distinguishing T cells. Examples of markers for different cell types are shown in Table 2.

[0340] According to some embodiments, training the first machine learning model on all samples in the training data may consume significant computational resources since all training data is stored in RAM during training. In some cases, RAM may be insufficient, which may lead to interruption of training of the first machine learning model and loss of training results.

[0341] Thus, in some embodiments, the first machine learning model is trained using a batch of training data. The batch may be obtained by dividing the training data into batches of a specified number of samples. For example, the batch may consist of a relatively small number of samples (e.g., 6 samples, 8 samples, 10 samples, 12 samples, 14 samples, etc.). In some embodiments, the batches are non-overlapping and their union is equal to the original sample set.

[0342] In some embodiments, training the first machine learning model using the batches of training data includes training the first machine learning model on each of at least some of the batches, such that only a relatively small number of samples are stored in RAM during training (e.g., the number of samples in the batch), thereby avoiding problems associated with insufficient RAM.

[0343] Operation 806b includes determining whether there is another machine learning model in the plurality of machine learning models that should be trained to determine another cell type. For example, this may include determining whether there is a second machine learning model that should be trained to determine whether the cell is of a second type. If operation 806b determines that there is another machine learning model, then process 800 returns to operation 806a to train the next machine learning model. In some embodiments, any number of machine learning models may be trained according to the techniques described herein, as aspects of the techniques described herein are not limited to any particular number of machine learning models.

[0344] In operation 806b, if there is not another machine learning model to train, then process 800 ends.

[0345] [Table 5]

[0346] [Table 6]

[0347] [Table 7]

[0348] [Table 8]

[0349] [Table 9]

[0350] Training Data As described herein, including with respect to Figure 8, training the plurality of machine learning models includes obtaining training data including cytometry data for cells and corresponding cell types. In some embodiments, obtaining the training data, obtaining cytometry data for one or more biological samples, and manually processing the cytometry data to determine types for cells in the biological samples.

[0351] In some embodiments, processing the cytometry data may include gating the cytometry data. For example, this may include manually gating the cytometry data to separate distinct cell populations based on shared marker expression. In some embodiments, gating may be performed using any suitable gating technique, such as by using FlowJo™ (FlowJo™ Software. Ashland, OR: Beckton, Dickinson and Company; 2021).

[0352] In some embodiments, the gating may result in the creation of a file (e.g., a Workspace (WSP) file) that contains any suitable information regarding the gating, such as gate coordinates, axis transformations, statistics, and layout information.

[0353] In some embodiments, processing the cytometry data may additionally or alternatively include clustering the cytometry data for the sample, which may include computing a two-dimensional t-SNE plot for the sample and computing a FlowSOM for the sample. FlowSOM is described by Van Gassen et al. ("FlowSOM: Using self-organizing maps for visualization and interpretation of cytometry data," in Journal of Quantitative Cell Science, vol. 87, no. 7, pp. 636-645, 2015), which is incorporated herein by reference in its entirety.

[0354] Prior to clustering, in some embodiments, processing the cytometry data may include noise transformation of the cytometry data. This may include transforming the intensities of the markers to reduce the effect of noise on the clustering results. In some embodiments, transforming the intensities of the markers includes reducing the intensities of the markers to be lower than a specified boundary. Such a boundary may be identified based on the results of gating or using a fluorescence minus one control. In some embodiments, the boundary is defined as the boundary between the positive signal from the marker and the intensity of the noise in the marker's channel. Equations 4 and 5 are expressed as the noise-transformed intensity of the marker (I after transform ) to write:

number

[0355] 9A-9B show the difference between the clustered cytometry data before and after noise transformation. As shown in FIG 9B, the clusters become much more clearly distinguishable from each other after noise transformation.

[0356] Figures 9C-D show the difference between the distribution of marker intensities before and after noise transformation. As shown in Figure 9D, the distribution of markers after noise transformation more closely resembles a bimodal distribution.

[0357] Whether or not a noise transformation technique is used, the technique may include plotting a t-SNE multiplot with marker intensity and light scatter after clustering. An example plot is shown in FIG. 9E. In some embodiments, each point may correspond to a cell, particle, or debris value for which cytometry data was obtained. In some embodiments, such a plot may be used to identify different clusters, which may correspond to populations of cells, particles, or debris.

[0358] In some embodiments, a user may manually assign display names to clusters that correspond to cell types. For example, as shown in FIG. 9F, different clusters are assigned display names for different cell types. Points within each named cluster may correspond to specific cells, particles, or debris in the cytometry data.

[0359] In some embodiments, an automatic labeling algorithm is used to assign display names to clusters corresponding to cell types. Display names may be selected based on positive or negative signals from a specified marker. A positive signal from a marker is when the intensity of the marker is higher than or equal to the boundary. A negative signal from a marker is when the intensity of the marker is lower than the boundary. As described above, the boundary for the intensity of the marker may be obtained from gating or from a fluorescence minus one control. The boundary is the boundary between the positive signal from the marker and the intensity of the noise in the marker's channel.

[0360] In some embodiments, the technique may further include discarding some of the identified clusters, for example, clusters corresponding to debris and / or particles may be discarded, and in some embodiments, the steps of computing and plotting the t-SNE plot and assigning display names to the clusters may be repeated without including the discarded clusterings.

[0361] Although various techniques for processing cytometry data have been described, it should be understood that any suitable technique may be used to process such data, as aspects of the technology described herein are not limited in this respect.

[0362] Machine Learning In some embodiments, the machine learning model may include a decision tree classifier, a gradient boosting decision tree classifier, a neural network, a support vector machine classifier, or any other suitable type of machine learning model, as aspects of the technology described herein are not limited in this respect. In some embodiments, the machine learning model may include an ensemble of any suitable type of machine learning model (wherein the machine learning model portion of the ensemble may be referred to as a "weak learner"). For example, the machine learning model may include an ensemble of decision tree classifiers.

[0363] As described above, in some embodiments, the machine learning model may be implemented as a decision tree classifier. Any suitable type of decision tree classifier may be used, and it may be trained using any suitable supervised decision tree learning method. For example, the decision tree classifier may be trained by iterative bisection (e.g., ID3 algorithm, as described, for example, in Quinlan, JR 1986. Induction of Decision Trees. Mach. Learn. 1, 1 (Mar. 1986), 81-106), C4.5 technique (e.g., as described, for example, in Quinlan, JRC4.5: Programs for Machine Learning. Morgan Kaufmann Publishers, 1993), classification and regression tree (CART) method (e.g., as described, for example, in Breiman, Leo; Friedman, JH; Olshen, RA; Stone, CJ (1984). Classification and regression trees. Monterey, CA: Wadsworth&Brooks / Cole Advanced Books&Software). It should be understood that the decision tree classifier may be trained using any other suitable training method, as aspects of the technology described herein are not limited in this respect.

[0364] In some embodiments, a gradient boosted decision tree classifier may be used. A gradient boosted decision tree classifier may be an ensemble of multiple decision tree classifiers (sometimes also referred to as "weak learners"). The prediction (e.g., classification) generated by the gradient boosted decision tree classifier is formed based on the predictions generated by the multiple decision tree parts of the ensemble. The ensemble may be trained using an iterative optimization method that involves computing the gradient of a loss function (hence the name "gradient" boosting). Any suitable supervised training algorithm may be applied to train the gradient boosted decision tree classifier, including, for example, any of the algorithms described in Hastie, T.; Tibshirani, R.; Friedman, JH (2009) "10. Boosting and Additive Trees". The Elements of Statistical Learning (2nd ed.). New York: Springer. pp. 337-384. In some embodiments, the gradient boosting decision tree classifier may be implemented using any suitable publicly available gradient boosting framework, such as XGBoost (e.g., as described in Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 785-794). New York, NY, USA: ACM. The XGBoost software may be available, for example, from http: / / xgboost.ai).Another example framework that may be utilized is LightGBM (e.g., as described in, for example, Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., ... Liu, T.-Y. (2017). Lightgbm: A highly efficient gradient boosting decision tree. Advances in Neural Information Processing Systems, 30, 3146-3154.). The LightGBM software may be obtained, for example, from https: / / lightgbm.readthedocs.io / ).

[0365] In some embodiments, a neural network classifier may be used. The neural network classifier may be trained using any suitable neural network optimization software. The optimization software may be configured to perform the neural network training by steepest descent, stochastic steepest descent, or any other suitable method. In some embodiments, an Adam optimizer (Kingma, D. and Ba, J. (2015) Adam: A Method for Stochastic Optimization. Proceedings of the 3rd International Conference on Learning Representations (ICLR 2015)) may be used.

[0366] Additional or Alternative Embodiments Various embodiments have been described using a hierarchy of machine learning models to identify one or more cell or particle types for an event. However, it should be understood that a multi-class classifier that does not use information about the hierarchy of cells may be used to predict the type for an event. For example, such a multi-class classifier may be based on the use of gradient boosting trees.

[0367] However, training such a multi-class classifier may require a significant increase in computation time and random access memory (RAM) compared to training a machine learning model in a hierarchy of machine learning models. In addition, during model training, all cell populations (e.g., child and parent cell populations) are considered equally. This simplifies the learning process, but it leads to a loss of information that would be captured by using a hierarchy of machine learning models.

[0368] Computer implementation FIG. 10 illustrates an exemplary implementation of a computer system 1000 that may be used in connection with any of the embodiments of the technology described herein (e.g., the methods of FIGS. 2A-2C, 5A-5D, and 8, etc.). The computer system 1000 includes one or more processors 1010 and one or more articles of manufacture that include a non-transitory computer-readable storage medium (e.g., memory 1020 and one or more non-volatile storage media 1030). The processor 1010 may control the writing and reading of data to and from the memory 1020 and the non-volatile storage device 1030 in any suitable manner, as aspects of the technology described herein are not limited to any particular data writing or reading technique. To perform any of the functions described herein, the processor 1010 may execute one or more processor-executable instructions stored in one or more non-transitory computer-readable storage media (e.g., memory 1020), which may act as a non-transitory computer-readable storage medium that stores the processor-executable instructions for execution by the processor 1010.

[0369] Computing device 1000 may also include a network input / output (I / O) interface 1040, through which the computing device may communicate with other computing devices (e.g., over a network), and one or more user I / O interfaces 1050, through which the computing device may provide output to and receive input from a user. The user I / O interfaces may include devices such as a keyboard, a mouse, a microphone, a display device (e.g., a monitor or touch screen), a speaker, a camera, and / or various types of I / O devices.

[0370] The above-described embodiments can be implemented in any of numerous ways. For example, the embodiments may be implemented using hardware, software, or a combination thereof. When implemented in software, the software code can be executed on any suitable processor (e.g., a microprocessor) or group of processors, whether provided on a single computing device or distributed among multiple computing devices. It should be understood that any component or group of components performing the functions described above can be collectively considered as one or more controllers that control the functions described above. The one or more controllers can be implemented in numerous ways, such as by dedicated hardware or by general-purpose hardware (e.g., one or more processors) that are programmed using microcode or software to perform the functions described above.

[0371] In this regard, it should be understood that an implementation of the embodiments described herein includes at least one computer-readable storage medium (e.g., RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device, or other tangible non-transitory computer-readable storage medium) encoded with a computer program (i.e., a plurality of executable instructions) that performs the above-described functions of one or more embodiments when executed on one or more processors. The computer-readable medium may be portable, and the program stored thereon may be loaded into any computing device to perform aspects of the techniques described herein. In addition, it should be understood that when referring to a computer program, it performs any of the above-described functions when executed, and is not limited to an application program running on a host computer. Rather, the terms computer program and software are used herein in a generic sense to refer to any type of computer code (e.g., application software, firmware, microcode, or any other form of computer instructions) that can be utilized to program one or more processors to implement aspects of the techniques described herein.

[0372] The foregoing description of implementations provides illustration and description, but is not intended to be exhaustive or to limit the implementations to the precise form disclosed. Modifications and variations are possible in light of the above teachings or may be acquired from practice of the implementations. In other implementations, the methods depicted in these figures may include fewer operations, different operations, operations in a different order, and / or additional operations. Additionally, independent blocks may be performed in parallel.

[0373] As described above, it will be apparent that example aspects may be implemented in many different forms of software, firmware, and hardware in the implementation illustrated in the figures. Furthermore, certain portions of the implementation may be implemented as a "module" that performs one or more functions. The module may include hardware, such as a processor, an application specific integrated circuit (ASIC), or a field programmable gate array (FPGA), or a combination of hardware and software.

[0374] Biological samples Any of the methods, systems, or other claimed elements may use or be used to analyze a biological sample from a subject. In some embodiments, the biological sample is obtained from a subject who has, is suspected of having, or is at risk of having cancer. The biological sample may be any type of biological sample, including, for example, a bodily fluid (e.g., blood, urine, or cerebrospinal fluid), one or more cells (e.g., from a scraping or brushing, such as a buccal swab or tracheal brushing), a tissue piece (buccal tissue, muscle tissue, lung tissue, heart tissue, brain tissue, or skin tissue), or a biological sample of part or all of an organ (e.g., brain, lung, liver, bladder, kidney, pancreas, intestine, or muscle), or other types of biological samples (e.g., feces or hair).

[0375] In some embodiments, the biological sample is a tumor sample from the subject. In some embodiments, the biological sample is a blood sample from the subject. In some embodiments, the biological sample is a tissue sample from the subject.

[0376] A tumor sample, in some embodiments, refers to a sample that includes cells from a tumor. In some embodiments, a tumor sample includes cells from a benign tumor, e.g., non-cancerous cells. In some embodiments, a tumor sample includes cells from a pre-malignant tumor, e.g., pre-cancerous cells. In some embodiments, a tumor sample includes cells from a malignant tumor, e.g., cancerous cells.

[0377] Examples of tumors include, but are not limited to, adenoma, fibroma, hemangioma, lipoma, cervical dysplasia, pulmonary metaplasia, leukoplakia, carcinoma, sarcoma, germ cell tumor, and blastoma.

[0378] A blood sample, in some embodiments, refers to a sample that includes cells, e.g., cells from a blood sample. In some embodiments, the blood sample includes non-cancerous cells. In some embodiments, the blood sample includes pre-cancerous cells. In some embodiments, the blood sample includes cancerous cells. In some embodiments, the blood sample includes blood cells. In some embodiments, the blood sample includes red blood cells. In some embodiments, the blood sample includes white blood cells. In some embodiments, the blood sample includes platelets. Examples of cancerous blood cells include, but are not limited to, leukemia, lymphoma, and myeloma. In some embodiments, a blood sample is taken to obtain cell-free nucleic acid (e.g., cell-free DNA) in the blood.

[0379] The blood sample may be a whole blood sample or a fractionated blood sample. In some embodiments, the blood sample comprises whole blood. In some embodiments, the blood sample comprises fractionated blood. In some embodiments, the blood sample comprises buffy coat. In some embodiments, the blood sample comprises serum. In some embodiments, the blood sample comprises plasma. In some embodiments, the blood sample comprises a blood clot.

[0380] A tissue sample, in some embodiments, refers to a sample that includes cells from the tissue. In some embodiments, a tumor sample includes non-cancerous cells from the tissue. In some embodiments, a tumor sample includes pre-cancerous cells from the tissue.

[0381] The methods of the present disclosure encompass a variety of tissues, including organ or non-organ tissues, including, but not limited to, muscle tissue, brain tissue, lung tissue, liver tissue, epithelial tissue, connective tissue, and nerve tissue. In some embodiments, the tissue may be normal tissue, or it may be diseased tissue, or it may be tissue suspected of being diseased. In some embodiments, the tissue may be sectioned tissue or whole intact tissue. In some embodiments, the tissue may be animal tissue or human tissue. Animal tissue includes, but is not limited to, tissue obtained from rodents (e.g., rats or mice), primates (e.g., monkeys), dogs, cats, and farm animals.

[0382] The biological sample may be from any source in the subject's body, including, but not limited to, any bodily fluid (such as blood (e.g., whole blood, serum, or plasma), saliva, tears, synovial fluid, cerebrospinal fluid, pleural fluid, pericardial fluid, peritoneal fluid, and / or urine), hair, skin (including portions of the epidermis, dermis, and / or subcutaneous tissue), oropharynx, laryngopharynx, esophagus, stomach, bronchi, salivary glands, tongue, oral cavity, nasal cavity, vaginal cavity, anal cavity, bone, bone marrow, brain, thymus, spleen, small intestine, appendix, colon, rectum, anus, liver, biliary tract, pancreas, kidney, ureter, bladder, urethra, uterus, vagina, vulva, ovaries, cervix, scrotum, penis, prostate, testes, seminal vesicles, and / or any type of tissue (e.g., muscle tissue, epithelial tissue, connective tissue, or nerve tissue).

[0383] Any biological sample described herein may be obtained from a subject using any known technique. For example, see the following publications regarding collection, processing, and storage of biological samples, each of which is incorporated herein by reference in its entirety: Biospecimens and biorepositories: from afterthought to science by Vaught et al. (Cancer Epidemiol Biomarkers Prev. 2012 Feb; 21(2): 253-5), and Biological sample collection, processing, storage and information management by Vaught and Henderson (IARC Sci Publ. 2011; (163): 23-42).

[0384] In some embodiments, the biological sample may be obtained from a surgical procedure (e.g., laparoscopy, microscopic surgery, or endoscopy), a bone marrow biopsy, a punch biopsy, an endoscopic biopsy, or a needle biopsy (e.g., fine needle aspiration, core needle biopsy, aspiration biopsy, or image-guided biopsy).

[0385] In some embodiments, one or more cells (i.e., cell biosamples) may be obtained from a subject using scraping or brushing methods. Cell biosamples may be obtained from any area in or from a subject's body, including, for example, from one or more of the following areas: cervix, esophagus, stomach, bronchus, or oral cavity. In some embodiments, one or more tissue pieces (e.g., tissue biopsy) from a subject may be used. In certain embodiments, a tissue biopsy may include one or more (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, or more than 10) biosamples from one or more tumors or tissues known or suspected to have cancerous cells.

[0386] Any biological sample from a subject described herein may be stored using any method that maintains the stability of the biological sample. In some embodiments, maintaining the stability of the biological sample means preventing the degradation of the components of the biological sample (e.g., DNA, RNA, protein, or tissue structure or morphology) until it is measured, so that at the time of measurement, the measurement corresponds to the state of the sample at the time it is obtained from the subject. In some embodiments, the biological sample is stored in a composition that can permeate it and protect the components of the biological sample (e.g., DNA, RNA, protein, or tissue structure or morphology) from degradation. As used herein, degradation is the change of a component from one form to another, such that the first form is no longer detected at the same level as before degradation.

[0387] In some embodiments, the biological sample (e.g., tissue sample) is fixed. As used herein, a "fixed" sample refers to a sample that has been treated with one or more agents or processes to prevent or reduce decay or degradation, such as autolysis or putrefaction, of the sample. Examples of fixation processes include, but are not limited to, heat fixation, immersion fixation, and perfusion. In some embodiments, the fixed sample is treated with one or more fixatives. Examples of fixatives include, but are not limited to, crosslinkers (e.g., aldehydes such as formaldehyde, formalin, glutaraldehyde, etc.), precipitants (e.g., alcohols such as ethanol, methanol, acetone, xylene, etc.), mercurials (e.g., B-5, Zenker fixative, etc.), picrates, and Hepes-Glutamate Buffer-Mediated Organic Solvent Protective Effect (HOPE) fixatives (fixatuves). In some embodiments, the biological sample (e.g., tissue sample) is treated with a crosslinker. In some embodiments, the crosslinker comprises formalin. In some embodiments, the biological sample that is fixed with formalin is embedded in a solid substrate, for example paraffin wax.In some embodiments, the biological sample is a formalin-fixed paraffin-embedded (FFPE) sample.The preparation method of FFPE sample is known, for example, as described by Li et al.JCO Precis Oncol.2018;2:PO.17.00091.

[0388] In some embodiments, the biological sample is preserved using a cryopreservation method. Non-limiting examples of cryopreservation methods include, but are not limited to, step-down freezing, blast freezing, direct plunge freezing, snap freezing, slow freezing using a programmable freezer, and vitrification. In some embodiments, the biological sample is preserved using freeze-drying. In some embodiments, the biological sample is placed in a container that already contains a protective agent (e.g., RNALater to protect RNA) after collection of the biological sample from the subject, and then frozen (e.g., by snap freezing). In some embodiments, such storage in a frozen state is performed immediately after collection of the biological sample. In some embodiments, the biological sample may be placed in a protective agent or in a buffer solution without a protective agent at either room temperature or 4°C for a period of time (e.g., up to 1 hour, up to 8 hours, or up to 1 day, or several days) before freezing.

[0389] Non-limiting examples of protective agents include formalin solution, formaldehyde solution, RNALater or other equivalent solution, TriZol or other equivalent solution, DNA / RNA Shield or equivalent solution, EDTA (e.g., Buffer AE (10 mM Tris·Cl; 0.5 mM EDTA, pH 9.0)) and other coagulants, and citrate dextrose (e.g., for blood specimens). In some embodiments, special containers may be used to collect and / or store the biological sample. For example, a vacutainer may be used to store blood. In some embodiments, the vacutainer may contain a protective agent (e.g., a coagulant or anticoagulant). In some embodiments, the container in which the biological sample is stored may be placed into a secondary container for better storage or to prevent contamination.

[0390] Any biological sample from a subject described herein may be stored under any condition that maintains the stability of the biological sample. In some embodiments, the biological sample is stored at a temperature that maintains the stability of the biological sample. In some embodiments, the sample is stored at room temperature (e.g., 25°C). In some embodiments, the sample is stored under refrigeration (e.g., 4°C). In some embodiments, the sample is stored under frozen conditions (e.g., -20°C). In some embodiments, the sample is stored under cryogenic conditions (e.g., -50°C to -800°C). In some embodiments, the sample is stored under liquid nitrogen (e.g., -1700°C). In some embodiments, the biological sample is stored at -60°C to -80°C (e.g., -70°C) for up to 5 years (e.g., up to 1 month, up to 2 months, up to 3 months, up to 4 months, up to 5 months, up to 6 months, up to 7 months, up to 8 months, up to 9 months, up to 10 months, up to 11 months, up to 1 year, up to 2 years, up to 3 years, up to 4 years, or up to 5 years). In some embodiments, the biological sample is stored for up to 20 years (e.g., up to 5 years, up to 10 years, up to 15 years, or up to 20 years) as described by any of the methods described herein.

[0391] The method of the present disclosure includes obtaining one or more biological samples from a subject for analysis. In some embodiments, one biological sample is taken from a subject for analysis. In some embodiments, two or more (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more) biological samples are taken from a subject for analysis. In some embodiments, one biological sample from a subject will be analyzed. In some embodiments, two or more (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more) biological samples may be analyzed. If more than one biological sample from a subject is analyzed, the biological samples may be procured at the same time (e.g., two or more biological samples may be taken during the same procedure) or the biological samples may be taken at different time points (e.g., between different procedures, including a procedure at 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 days; 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 weeks; 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 months, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 years, or 10, 20, 30, 40, 50, 60, 70, 80, 90, 100 years after the initial procedure).

[0392] The second or subsequent biological sample may be taken or obtained from the same area (e.g., from the same tumor or tissue area) or a different area (e.g., an area including a different tumor). The second or subsequent biological sample may be taken or obtained from the subject after one or more treatments, and may be taken from the same area or a different area. As a non-limiting example, the second or subsequent biological sample may be useful in determining whether the cancer in each biological sample has different characteristics (e.g., when biological samples are taken from two physically separate tumors in a patient) or whether the cancer has responded to one or more treatments (e.g., in the case of two or more biological samples from the same tumor or different tumors, both pre- and post-treatment). In some embodiments, each of the at least one biological sample is a bodily fluid sample, a cell sample, or a tissue biopsy sample.

[0393] In some embodiments, one or more biological specimens are combined (e.g., placed in the same container for storage) before further processing. For example, a first sample of a first tumor obtained from a subject may be combined with a second sample of a second tumor from the subject, where the first and second tumors may or may not be the same tumor. In some embodiments, the first and second tumors are similar but not the same (e.g., two tumors in the brain of a subject). In some embodiments, the first and second biological samples from a subject are samples of different types of tumors (e.g., tumors in muscle tissue and brain tissue).

[0394] In some embodiments, the sample from which RNA and / or DNA is extracted (e.g., a tumor sample or a blood sample) is sufficiently large to allow at least 2 μg (e.g., at least 2 μg, at least 2.5 μg, at least 3 μg, at least 3.5 μg or more) of RNA to be extracted therefrom. In some embodiments, the sample from which RNA and / or DNA is extracted may be peripheral blood mononuclear cells (PBMCs). In some embodiments, the sample from which RNA and / or DNA is extracted may be any type of cell suspension. In some embodiments, the sample from which RNA and / or DNA is extracted (e.g., a tumor sample or a blood sample) is sufficiently large to allow at least 1.8 μg of RNA to be extracted therefrom. In some embodiments, at least 50 mg (e.g., at least 1 mg, at least 2 mg, at least 3 mg, at least 4 mg, at least 5 mg, at least 10 mg, at least 12 mg, at least 15 mg, at least 18 mg, at least 20 mg, at least 22 mg, at least 25 mg, at least 30 mg, at least 35 mg, at least 40 mg, at least 45 mg, or at least 50 mg) of tissue sample is taken and RNA and / or DNA is extracted therefrom. In some embodiments, at least 20 mg of tissue sample is taken and RNA and / or DNA is extracted therefrom. In some embodiments, at least 30 mg of tissue sample is taken. In some embodiments, at least 10-50 mg (e.g., 10-50 mg, 10-15 mg, 10-30 mg, 10-40 mg, 20-30 mg, 20-40 mg, 20-50 mg, or 30-50 mg) of tissue sample is taken and RNA and / or DNA is extracted therefrom. In some embodiments, at least 30 mg of tissue sample is taken, hi some embodiments, at least 20-30 mg of tissue sample is taken and RNA and / or DNA is extracted therefrom.In some embodiments, the sample from which RNA and / or DNA is extracted (e.g., a tumor sample or a blood sample) is sufficiently large that at least 0.2 μg (e.g., at least 200 ng, at least 300 ng, at least 400 ng, at least 500 ng, at least 600 ng, at least 700 ng, at least 800 ng, at least 900 ng, at least 1 μg, at least 1.1 μg, at least 1.2 μg, at least 1.3 μg, at least 1.4 μg, at least 1.5 μg, at least 1.6 μg, at least 1.7 μg, at least 1.8 μg, at least 1.9 μg, or at least 2 μg) of RNA can be extracted therefrom. In some embodiments, the sample (e.g., a tumor sample or a blood sample) from which RNA and / or DNA is extracted is sufficiently large so that at least 0.1 μg (e.g., at least 100 ng, at least 200 ng, at least 300 ng, at least 400 ng, at least 500 ng, at least 600 ng, at least 700 ng, at least 800 ng, at least 900 ng, at least 1 μg, at least 1.1 μg, at least 1.2 μg, at least 1.3 μg, at least 1.4 μg, at least 1.5 μg, at least 1.6 μg, at least 1.7 μg, at least 1.8 μg, at least 1.9 μg, or at least 2 μg) of RNA can be extracted therefrom.

[0395] subject Aspects of the present disclosure relate to biological samples obtained from a subject. In some embodiments, the subject is a mammal (e.g., a human, mouse, cat, dog, horse, hamster, cow, pig, or other domesticated animal). In some embodiments, the subject is a human. In some embodiments, the subject is an adult human (e.g., 18 years of age or older). In some embodiments, the subject is a child (e.g., under 18 years of age). In some embodiments, the human subject has or has been diagnosed with at least one form of cancer. In some embodiments, the cancer the subject is suffering from is a carcinoma, sarcoma, myeloma, leukemia, lymphoma, or a mixed type of cancer comprising two or more of carcinoma, sarcoma, myeloma, leukemia, and lymphoma. Carcinoma refers to a malignant neoplasm of epithelial origin or a cancer of the lining or lining of the body. Sarcoma refers to a cancer that arises in the supporting and connective tissues, such as bone, tendon, cartilage, muscle, and fat. Myeloma is a cancer that arises in the plasma cells of the bone marrow. Leukemia ("liquid cancer" or "blood cancer") is cancer of the bone marrow (site of blood cell production). Lymphoma develops in the glands or nodes of the lymphatic system, the vascular network, nodes, and organs (particularly the spleen, tonsils, and thymus) that produce the white blood cells, or lymphocytes, that purify the body's fluids and fight infection. Gliomas develop in the brain, particularly in the glial cells that provide physical and metabolic support to neurons. Non-limiting examples of mixed cancers include adenosquamous carcinoma, mixed mesodermal tumor, carcinosarcoma, and teratocarcinoma. In some embodiments, the subject has a tumor. The tumor can be benign or malignant. In some embodiments, the cancer is any one of the following: skin cancer, lung cancer, breast cancer, prostate cancer, colon cancer, rectal cancer, cervical cancer, and uterine cancer. In some embodiments, a subject is at risk for developing cancer because, for example, the subject has one or more genetic risk factors or has been exposed to or is being exposed to one or more carcinogens (e.g., tobacco smoke or chewing tobacco).

[0396] Flow cytometry In some embodiments, a flow cytometry platform may be used to perform flow cytometry interrogation of a flowing sample. The flowing sample may include target particles having specific particle attributes. The flow cytometry interrogation of the flowing sample may provide a flow cytometry result for the flowing sample.

[0397] In some embodiments, the flow sample may be exposed to a stain or dye that, when exposed to interrogating excitation radiation, provides a response radiation that can be measured by the radiation detection system of the flow cytometry platform. In some embodiments, the flow cytometry platform includes multiple light detectors. When a particle passes through the laser beam, time-correlated pulses will occur on the forward scatter (FSC) and side scatter (SSC) detectors, and possibly also on the fluorescence emission detector. This is an "event" and for each event, the magnitude of the detector output for each detector, FSC, SSC and fluorescence detector, is stored. The data obtained includes the measured signal for each of the light scatter parameters and fluorescence emission.

[0398] The flow cytometry platform may further include components for storing the detector output and analyzing the data. For example, data storage and analysis may be performed using a computer connected to the detection electronics. For example, data may be stored logically in a table format, where each row corresponds to the data of one particle (or one event) and columns correspond to each of the measured parameters. A standard file format, such as the "FCS" file format, may be used to store data from the flow cytometer, facilitating analysis of the data using separate programs and / or machines. In some embodiments, data may be displayed in a two-dimensional (2D) plot for ease of visualization, although other methods may be used to visualize multi-dimensional data.

[0399] In some embodiments, parameters measured using a flow cytometer may include FSC, which refers to excitation light scattered generally along the forward direction by a particle, SSC, which refers to excitation light scattered generally to the side by a particle, and light emitted by fluorescent molecules in one or more channels (frequency bands) of the spectrum, designated FL1, FL2, etc., or by the name of the fluorochrome that is primarily emitted in that channel.

[0400] Both flow cytometers and scanning cytometers are commercially available, for example, from BD Biosciences (San Jose, Calif.). Flow cytometry is described, for example, in Landy et al. (eds.), Clinical Flow Cytometry, Annals of the New York Academy of Sciences Volume 677 (1993); Bauer et al. (eds.), Clinical Flow Cytometry: Principles and Applications, Williams & Wilkins (1993); Ormerod (ed.), Flow Cytometry: A Practical Approach, Oxford Univ. Press (1997); Jaroszeski et al. (eds.), Flow Cytometry Protocols, Methods in Molecular Biology No. 91, Humana Press (1997); and Practical Shapiro, Flow Cytometry, 4th ed., Wiley-Liss (2003), all of which are incorporated herein by reference. Fluorescence imaging microscopy is described, for example, in Pawley (ed.), Handbook of Biological Confocal Microscopy, 2nd Edition, Plenum Press (1989), incorporated herein by reference.

[0401] Mass Cytometry In some embodiments, a mass cytometry platform may be used to perform a mass cytometry interrogation of a flowing sample. The flowing sample may include target particles having specific particle attributes. The mass cytometry interrogation of the flowing sample may provide a mass cytometry result for the flowing sample.

[0402] In some embodiments, the flow sample may be exposed to a target-specific antibody labeled with a metal isotope. In some embodiments, elemental mass spectrometry (e.g., inductively coupled plasma mass spectrometry (ICP-MS) and time-of-flight mass spectrometry (TOF-MS)) is used to detect the conjugated antibody. For example, elemental mass spectrometry can distinguish between isotopes of different atomic weights and measure the electrical signal for the isotope associated with each particle or cell. The data obtained for a single cell or particle is considered an "event".

[0403] The mass cytometry platform may further include components for storing the detector output and analyzing the data. For example, data storage and analysis may be performed using a computer connected to the detection elements. The storage of data from the mass cytometry platform may use a standard file format, such as the "FCS" file format, to facilitate analysis of the data using separate programs and / or machines.

[0404] Mass cytometry platforms are commercially available, for example, from Fluidigm (San Francisco, Calif.). Mass cytometry is described, for example, in Bendall et al., A deep profiler's guide to cytometry, Trends in Immunology, 33(7), 323-332 (2012) and Spitzer et al., Mass Cytometry: Single Cells, Many Features, Cell, 165(4), 780-791 (2016), both of which are incorporated herein by reference in their entirety.

[0405] Treatment method In certain methods described herein, an effective amount of an anti-cancer therapy described herein may be administered or recommended for administration by a suitable route (e.g., intravenous administration) to a subject (e.g., a human) in need of treatment.

[0406] The subject to be treated by the methods described herein may be a human patient who has, is suspected of having, or is at risk of having cancer. Examples of cancer include, but are not limited to, melanoma, lung cancer, brain cancer, breast cancer, colorectal cancer, pancreatic cancer, liver cancer, prostate cancer, skin cancer, renal cancer, bladder cancer, or prostate cancer. At the time of diagnosis, cancer may be of unknown primary origin. The subject to be treated by the methods described herein may be a mammal (e.g., may be a human). Mammals include, but are not limited to: farm animals (e.g., livestock), sport animals, laboratory animals, pets, primates, horses, dogs, cats, mice, and rats.

[0407] A subject with cancer may be identified by routine medical examination, such as laboratory tests, biopsies, PET scans, CT scans, or ultrasound. A subject suspected of having cancer may show one or more symptoms of the disorder, such as unexplained weight loss, fever, fatigue, cough, pain, skin changes, abnormal bleeding or discharge, and / or thickening or lumps in body parts. A subject at risk of cancer may be a subject with one or more risk factors for the disorder. For example, risk factors associated with cancer include, but are not limited to, (a) viral infection (e.g., herpes virus infection), (b) age, (c) family history, (d) excessive alcohol intake, (e) obesity, and (f) tobacco use.

[0408] "Effective amount" as used herein refers to the amount of each active agent, either alone or in combination with one or more other active agents, required to confer a therapeutic effect on a subject. As recognized by those skilled in the art, effective amounts vary depending on the detailed condition under treatment, the severity of the condition, the parameters of the individual patient, including age, health, size, sex and weight, duration of treatment, the nature of concomitant treatment (if any), the specific route of administration and similar factors within the knowledge and expertise of a health care practitioner. These factors are well known to those skilled in the art and can be addressed with no more than routine experimentation. In general, it is preferred to use the maximum dose of each component or combination thereof, i.e., the highest safe dose according to sound medical judgment. However, those skilled in the art will understand that for medical reasons, psychological reasons, or virtually any other reason, a patient may require a lower dose or a dose that can be tolerated.

[0409] Empirical considerations such as the half-life of the therapeutic compound generally contribute to the determination of dosage. For example, to prolong the half-life of the antibody and to prevent the antibody from being attacked by the host's immune system, an antibody compatible with the human immune system, such as a humanized antibody or a fully human antibody, may be used. The frequency of administration may be determined and adjusted during the course of therapy, generally (but not necessarily) based on the treatment and / or suppression and / or improvement and / or delay of cancer. Alternatively, a continuous sustained release formulation of the anti-cancer therapeutic agent may be appropriate. Various formulations and devices for achieving sustained release are known in the art.

[0410] In some embodiments, the dosage of the anti-cancer therapeutic agent as described herein may be empirically determined with the individual receiving one or more doses of the anti-cancer therapeutic agent. The individual may receive increasing dosages of the anti-cancer therapeutic agent. To determine the effectiveness of the administered anti-cancer therapeutic agent, one or more aspects of the cancer (e.g., tumor formation, tumor growth, molecular categories identified for the cancer using the techniques described herein) may be analyzed.

[0411] In general, the initial dosage candidate for any administration of the anti-cancer antibodies described herein may be about 2 mg / kg. For purposes of this disclosure, a typical daily dosage may range anywhere from about 0.1 μg / kg to 3 μg / kg to 30 μg / kg to 300 μg / kg to 3 mg / kg, to 30 mg / kg to 100 mg / kg, or more, depending on the factors mentioned above. For repeated administration over several days or more, treatment is continued until a desired suppression or amelioration of symptoms occurs, or until a therapeutic level sufficient to alleviate the cancer, or one or more symptoms thereof, is achieved, depending on the condition. An exemplary titration regimen includes administering an initial dose of about 2 mg / kg, followed by a maintenance dose of about 1 mg / kg of the antibody every week, or a maintenance dose of about 1 mg / kg every other week. However, other dosage regimens may be useful depending on the pharmacokinetic decay pattern the practitioner (e.g., physician) wishes to achieve. For example, titration 1-4 times per week is contemplated. In some embodiments, a dose ranging from about 3 μg / mg to about 2 mg / kg (such as about 3 μg / mg, about 10 μg / mg, about 30 μg / mg, about 100 μg / mg, about 300 μg / mg, about 1 mg / kg, and about 2 mg / kg) may be used. In some embodiments, the dose frequency is once every week, every 2 weeks, every 4 weeks, every 5 weeks, every 6 weeks, every 7 weeks, every 8 weeks, every 9 weeks, or every 10 weeks; or once every month, every 2 months, or every 3 months, or more. The progress of this therapy may be monitored by conventional techniques and assays. The dose regimen (including the treatment used) may vary over time.

[0412] When the anti-cancer therapeutic agent is not an antibody, it may be administered at a rate of about 0.1-300 mg / kg of patient body weight in 1-3 divided doses, or as disclosed herein. In some embodiments, for a normal weight adult patient, a dose in the range of about 0.3-5.00 mg / kg may be administered. The particular dosage regimen, e.g., dose, timing, and / or repetition, will depend on the particular subject and their individual medical history, as well as the characteristics of the individual agent (such as the half-life of the agent, and other considerations well known in the art).

[0413] For purposes of this disclosure, the appropriate dosage of an anti-cancer therapeutic will depend on the particular anti-cancer therapeutic agent or agents (or compositions thereof) utilized, the type and severity of the cancer, whether the anti-cancer therapeutic agent is administered for prophylactic or therapeutic purposes, previous therapy, the patient's medical history and response to the anti-cancer therapeutic agent, and the discretion of the attending physician. Typically, a clinician will administer an anti-cancer therapeutic agent, such as an antibody, until a dosage is reached that achieves the desired result.

[0414] Administration of anti-cancer therapeutic agents can be continuous or intermittent, depending, for example, on the physiological condition of the recipient, whether the purpose of administration is therapeutic or prophylactic, and other factors known to those of skill in the art. Administration of anti-cancer therapeutic agents (e.g., anti-cancer antibodies) can be essentially continuous over a preselected period of time, or can be in a series of spaced doses, for example, either before, during, or after the onset of cancer.

[0415] As used herein, the term "treating" refers to the application or administration of a composition containing one or more active agents to a subject having cancer, a symptom of cancer, or a predisposition to cancer, for the purpose of curing, curing, mitigating, ameliorating, altering, repairing, ameliorating, reversing, or affecting the cancer or one or more symptoms of cancer, or the predisposition to cancer.

[0416] Alleviating cancer includes delaying the onset or progression of disease or reducing disease severity. Alleviating disease does not necessarily require a curative outcome. As used herein, "delaying" the onset of disease (e.g., cancer) means delaying, preventing, slowing, slowing, stabilizing, and / or postponing the progression of disease. This delay may be of various lengths of time depending on the course of the disease and / or individual under treatment. A method of "delaying" or alleviating the onset of disease, or delaying the onset of disease, is a method that reduces the likelihood of developing one or more symptoms of the disease in a given period of time and / or reduces the severity of symptoms in a given time frame, compared to not using the method. Such comparisons are typically based on clinical trials using a sufficient number of subjects to produce statistically significant results.

[0417] "Onset" or "progression" of a disease refers to the first symptoms and / or subsequent progression of the disease. Onset of a disease can be detected and determined using clinical techniques known in the art. However, onset also refers to progression, which may be undetectable. For purposes of this disclosure, onset or progression refers to the biological course of a condition. "Onset" includes onset, recurrence, and onset. As used herein, "onset" or "onset" of cancer includes initial onset and / or recurrence.

[0418] In some embodiments, the anti-cancer therapeutics (e.g., antibodies) described herein are administered to a subject in need of treatment in an amount sufficient to reduce cancer (e.g., tumor) growth by at least 10% (e.g., 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% or more). In some embodiments, the anti-cancer therapeutics (e.g., antibodies) described herein are administered to a subject in need of treatment in an amount sufficient to reduce the number of cancer cells or tumor size by at least 10% (e.g., 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% or more). In other embodiments, the anti-cancer therapeutics are administered in an amount effective to alter the cancer type. Alternatively, the anti-cancer therapeutics are administered in an amount effective to reduce tumor formation or metastasis.

[0419] Depending on the type of disease or site of disease to be treated, the anti-cancer therapeutic agent may be administered to the subject using conventional methods known to those skilled in the medical field. The anti-cancer therapeutic agent may also be administered by other conventional routes, for example, orally, parenterally, by inhalation spray, topically, rectally, nasally, bucally, vaginally, or from an implanted reservoir. The term "parenteral" as used herein includes subcutaneous, intradermal, intravenous, intramuscular, intra-articular, intra-arterial, intrasynovial, intrasternal, intrathecal, intralesional, and intracranial injection or infusion techniques. In addition, the anti-cancer therapeutic agent may be administered to the subject by injectable depot administration routes, such as using 1-month, 3-month, or 6-month depot injectable or biodegradable materials and methods.

[0420] Injectable compositions may contain a variety of carriers, such as vegetable oils, dimethylactamide, dimethylformamide, ethyl lactate, ethyl carbonate, isopropyl myristate, ethanol, and polyols (e.g., glycerol, propylene glycol, liquid polyethylene glycol, etc.). For intravenous injections, the water-soluble anti-cancer therapeutic agent may be administered by drip infusion, resulting in the infusion of a pharmaceutical formulation containing the antibody and a physiologically acceptable excipient. The physiologically acceptable excipient may include, for example, 5% dextrose, 0.9% saline, Ringer's solution, and / or other suitable excipients. For intramuscular preparations, for example, a sterile formulation of the anti-cancer therapeutic agent in a suitable soluble salt form may be administered dissolved in a pharmaceutical excipient, such as water for injection, 0.9% saline, and / or 5% glucose solution.

[0421] In one embodiment, anti-cancer therapeutic agents are administered by site-specific or targeted local delivery techniques. Examples of site-specific or targeted local delivery techniques include various implantable drug depot sources or local delivery catheters such as infusion catheters, indwelling catheters, or needle catheters, synthetic grafts, adventitial wrapping, shunts and stents or other implantable devices, site-specific carriers, direct injection, or direct application. See, for example, WO 00 / 53211 and U.S. Patent No. 5,981,568 (the contents of each of which are incorporated herein by reference for this purpose).

[0422] Targeted delivery of therapeutic compositions containing antisense polynucleotides, expression vectors, or subgenomic polynucleotides can also be used. Receptor-mediated DNA delivery techniques are described, for example, in Findeis et al., Trends Biotechnol. (1993) 11:202; Chiou et al., Gene Therapeutics: Methods and Applications Of Direct Gene Transfer (JA Wolff, ed.) (1994); Wu et al., J. Biol. Chem. (1988) 263:621; Wu et al., J. Biol. Chem. (1994) 269:542; Zenke et al., Proc. Natl. Acad. Sci. USA (1990) 87:3655; Wu et al., J. Biol. Chem. (1991) 266:338. The contents of each of the foregoing are incorporated herein by reference for this purpose.

[0423] Therapeutic compositions containing polynucleotides may be administered in the range of about 100 ng to about 200 mg of DNA for local administration in gene therapy protocols. In some embodiments, concentration ranges of about 500 ng to about 50 mg, about 1 μg to about 2 mg, about 5 μg to about 500 μg, and about 20 μg to about 100 μg of DNA or more may also be used in gene therapy protocols.

[0424] Therapeutic polynucleotides and polypeptides can be delivered using gene delivery vehicles. Gene delivery vehicles can be of viral or non-viral origin (e.g., Jolly, Cancer Gene Therapy (1994) 1:51; Kimura, Human Gene Therapy (1994) 5:845; Connelly, Human Gene Therapy (1995) 1:185; and Kaplitt, Nature Genetics (1994) 6:148), the contents of each of the foregoing being incorporated herein by reference for this purpose. Expression of such coding sequences can be induced using endogenous mammalian or heterologous promoters and / or enhancers. Expression of the coding sequences can be either constitutive or regulated.

[0425] Viral-based vectors for delivery of a desired polynucleotide and expression in a desired cell are well known in the art. Exemplary virus-based vehicles include, but are not limited to, recombinant retroviruses (see, e.g., WO 90 / 07936; WO 94 / 03622; WO 93 / 25698; WO 93 / 25234; WO 93 / 11230; WO 93 / 10218; WO 91 / 02805; U.S. Pat. Nos. 5,219,740 and 4,777,127; British Patent No. 2,200,651; and European Patent No. 0 345 242), alphavirus-based vectors (e.g., Sindbis virus vectors, Semliki Forest virus (ATCC VR-67; ATCC VR-1247), Ross River virus (ATCC VR-373; ATCC VR-1246), and Venezuelan equine encephalitis virus (ATCC VR-923; ATCC VR-1247). VR-1250; ATCC VR 1249; ATCC VR-532), and adeno-associated virus (AAV) vectors (see, e.g., WO 94 / 12649, WO 93 / 03769; WO 93 / 19191; WO 94 / 28938; WO 95 / 11984 and WO 95 / 00655). Also, administration of DNA linked to killed adenovirus as described in Curiel, Hum. Gene Ther. (1992) 3:147 can be utilized. The contents of each of the foregoing are incorporated herein by reference for this purpose.

[0426] Non-viral delivery vehicles and methods can also be used, including, but not limited to, polycation-condensed DNA, either linked or not linked to killed adenovirus alone (see, e.g., Curiel, Hum. Gene Ther. (1992) 3:147); ligand-linked DNA (see, e.g., Wu, J. Biol. Chem. (1989) 264:16985); eukaryotic cell delivery vehicles (see, e.g., U.S. Pat. No. 5,814,482; WO 95 / 07994; WO 96 / 17072; WO 95 / 30763; and WO 97 / 42338) and nuclear charge neutralization or fusion with cell membranes. Naked DNA can also be used. Exemplary naked DNA transfer methods are described in WO 90 / 11092 and U.S. Pat. No. 5,580,859. Liposomes that can serve as gene delivery vehicles are described in U.S. Patent No. 5,422,120; WO 95 / 13796; WO 94 / 23697; WO 91 / 14445; and EP 0524968. Further techniques are described in Philip, Mol. Cell. Biol. (1994) 14:2411, and Woffendin, Proc. Natl. Acad. Sci. (1994) 91:1581, the contents of each of the foregoing being incorporated herein by reference for this purpose.

[0427] It is also clear that expression vectors can be used to direct the expression of any protein-based anti-cancer therapeutic agent, such as an anti-cancer antibody. For example, peptide inhibitors are known in the art that have the ability to block (partially to completely) the biological activity caused by cancer.

[0428] In some embodiments, two or more anti-cancer therapeutic agents, such as antibodies and small molecule inhibitor compounds, may be administered to a subject in need of treatment. These agents may be of the same type or different types. At least one, at least two, at least three, at least four, or at least five different agents may be co-administered. In general, the anti-cancer agents for administration have complementary activities that do not adversely affect each other. Anti-cancer therapeutic agents may also be used in conjunction with other agents that act to enhance and / or complement the effectiveness of the agent.

[0429] The efficacy of treatment may be determined by methods well known in the art, for example, by monitoring tumor growth or formation in a patient undergoing treatment. Alternatively, or in addition, the efficacy of treatment may be determined by monitoring tumor type over the course of treatment (e.g., before, during, and after treatment).

[0430] A subject with cancer may be treated with any combination of anti-cancer therapeutic agents or one or more anti-cancer therapeutic agents and one or more additional therapies (e.g., surgery and / or radiation therapy). The term combination therapy, as used herein, includes administration of two or more therapies (e.g., an antibody and a small molecule or an antibody and radiation therapy) in a sequential manner, i.e., where each therapeutic agent is administered at a different time, as well as administration of the therapeutic agents, or at least two of the agents or therapies in a substantially simultaneous manner.

[0431] Sequential or substantially simultaneous administration of each agent or therapy may be effected by any suitable route, including, but not limited to, oral, intravenous, intramuscular, subcutaneous, and direct absorption through mucosal tissue. The agents or therapies may be administered by the same route or by different routes. For example, a first agent (e.g., a small molecule) may be administered orally and a second agent (e.g., an antibody) may be administered intravenously.

[0432] As used herein, the term "sequential" means characterized by a set order or sequence, unless otherwise specified; for example, if a dosage regimen includes administration of an antibody and a small molecule, a sequential dosage regimen may include administration of the antibody prior to, simultaneously with, substantially simultaneously with, or after administration of the small molecule, but both agents will be administered in a set order or sequence. The term "separate" means separate from one another, unless otherwise specified. The term "concurrently" means occurring or taking place at the same time, i.e., the agents are administered at the same time, unless otherwise specified. The term "substantially simultaneously" means that the agents are administered within minutes of each other (e.g., within 10 minutes of each other) and is intended to encompass co-administration as well as sequential administration, but where administration is sequential, it is only separated in time by a short period of time (e.g., the time it would take a medical practitioner to administer the two agents separately). As used herein, parallel administration and substantially simultaneous administration are used interchangeably. Sequential administration refers to the administration of the agents or therapies described herein separated in time.

[0433] Combination therapy can also include administration of anti-cancer therapeutic agents (e.g., antibodies) in further combination with other biologically active ingredients (e.g., vitamins) and non-pharmacological therapies (e.g., surgery or radiation therapy).

[0434] It should be understood that any combination of anti-cancer therapeutic agents may be used in any order to treat cancer. The combinations described herein may be selected based on a number of factors, including, but not limited to, efficacy in reducing tumor formation or tumor growth, and / or alleviating at least one symptom associated with cancer, or mitigating the side effects of the other agent in the combination. For example, combination therapy as provided herein may reduce any of the side effects associated with each individual member of the combination, such as the side effects associated with the anti-cancer agent administered.

[0435] In some embodiments, the anti-cancer therapeutic agent is an antibody, immunotherapy, radiation therapy, surgical therapy, and / or chemotherapy.

[0436] Examples of antibody anti-cancer agents include, but are not limited to, alemtuzumab (Campath), trastuzumab (Herceptin), ibritumomab tiuxetan (Zevalin), brentuximab vedotin (Adcetris), Ado-trastuzumab emtansine (Kadcyla), blinatumomab (Blincyto), bevacizumab (Avastin), cetuximab (Erbitux), ipilimumab (Yervoy), nivolumab (Opdivo), pembrolizumab (Keytruda), atezolizumab (Tecentriq), avelumab (Bavencio), durvalumab (Imfinzi), and panitumumab (Vectibix).

[0437] Examples of immunotherapies include, but are not limited to, PD-1 or PD-L1 inhibitors, CTLA-4 inhibitors, adoptive cell transfer, therapeutic cancer vaccines, oncolytic virotherapy, T cell therapy, and immune checkpoint inhibitors.

[0438] Examples of radiation therapy include, but are not limited to, ionizing radiation, gamma irradiation, neutron beam radiotherapy, electron beam radiotherapy, proton therapy, brachytherapy, systemic radioisotopes, and radiosensitizers.

[0439] Examples of surgical therapies include, but are not limited to, curative surgery (eg, tumor resection), preventative surgery, laparoscopic surgery, and laser surgery.

[0440] Examples of chemotherapeutic agents include, but are not limited to, carboplatin or cisplatin, docetaxel, gemcitabine, Nab-paclitaxel, paclitaxel, pemetrexed, and vinorelbine.

[0441] Further examples of chemotherapy include, but are not limited to, platinum agents such as carboplatin, oxaliplatin, cisplatin, nedaplatin, satraplatin, lobaplatin, triplatin, tetranitrate, picoplatin, Prolindac, Aroplatin and other derivatives; topoisomerase I inhibitors such as camptothecin, topotecan, irinotecan / SN38, rubitecan, belotecan and other derivatives; etoposide (VP-16), daunorubicin, doxorubicin-based agents (e.g., doxorubicin, doxorubicin hydrochloride, Topoisomerase II inhibitors such as cyclosporine, doxorubicin analogs, or liposomal doxorubicin and its salts or analogs, mitoxantrone, aclarubicin, epirubicin, idarubicin, amrubicin, amsacrine, pirarubicin, valrubicin, zorubicin, teniposide and other derivatives; antimetabolites such as the folate family (methotrexate, pemetrexed, raltitrexed, aminopterin, and its relatives or derivatives); purine antagonists (thioguanine, fludarabine, cladribine, 6-methylbutyric acid, methyl-2 ... cyclophosphamide, melphalan, chlorambucil, mechlorethamine, ifosfamide, trofosfamide, prednisone, cyclosporine, cyclobutene, cyclobutene-1, cyclobutene-2 ​​... alkylating agents such as benzodiazepine, bendamustine, uramustine, estramustine, and their analogs or derivatives; nitrosoureas (e.g., carmustine, lomustine, semustine, fotemustine, nimustine, ranimustine, streptozocin, and their analogs or derivatives); triazene groups (e.g., dacarbazine, altretamine, temozolomide, and their analogs or derivatives); alkyl sulfonates (e.g., busulfan, mannosulfan, treosulfan, and their analogs or derivatives); procarbazine;Antibiotics such as mitobronitol, and aziridines (e.g., carboquone, triaziquone, thiotepa, triethylenemalamine, and its analogs or derivatives); hydroxyurea, anthracyclines (e.g., doxorubicin, daunorubicin, epirubicin, and its analogs or derivatives); anthracenediones (e.g., mitoxantrone and its analogs or derivatives); Streptomyces family antibiotics (e.g., bleomycin, mitomycin C, actinomycin, and plicamycin); and ultraviolet light;

[0442] Although several aspects and embodiments of the technology presented in this disclosure have thus been described, it should be understood that various changes, modifications, and improvements will readily occur to those skilled in the art. Such changes, modifications, and improvements are intended to be within the spirit and scope of the technology described herein. For example, one skilled in the art could readily envision various other means and / or structures for performing the functions and / or achieving one or more of the results and / or advantages described herein, and each of such variations and / or modifications is deemed to be within the scope of the embodiments described herein. Those skilled in the art will recognize or be able to ascertain using no more than routine experimentation many equivalents to the specific embodiments described herein. Thus, it should be understood that the foregoing embodiments are provided by way of example only, and that within the scope of the appended claims and their equivalents, the embodiments of the invention may be implemented in other ways than as specifically disclosed. In addition, any combination of two or more of the features, systems, articles, materials, kits, and / or methods described herein is encompassed within the scope of the present disclosure, where such features, systems, articles, materials, kits, and / or methods are not mutually inconsistent.

[0443] The above-described embodiments may be implemented in any of numerous ways. One or more aspects and embodiments of the present disclosure involving the implementation of a process or method may utilize program instructions executable by a device (e.g., a computer, a processor, or other device) to perform or control the implementation of the process or method. In this regard, various inventive concepts may be embodied as a computer-readable storage medium (or multiple computer-readable storage media) (e.g., a computer memory, one or more floppy disks, compact disks, optical disks, magnetic tapes, flash memories, circuitry in field programmable gate arrays or other semiconductor devices, or other tangible computer storage media) encoded with one or more programs that, when executed on one or more computers or other processors, perform methods that implement one or more of the various embodiments described above. The one or more computer-readable media may be portable, such that the one or more programs stored thereon can be loaded into one or more different computers or other processors to implement various of the aspects described above. In some embodiments, the computer-readable medium may be a non-transitory medium.

[0444] The terms "program" or "software" are used herein in a generic sense to refer to any type of computer code or set of computer-executable instructions that can be utilized to program a computer or other processor to implement various aspects as described above. Additionally, according to one aspect, it should be understood that one or more computer programs that, when executed, perform the methods of the present disclosure need not reside on a single computer or processor, but may be distributed in a modular manner among several different computers or processors to implement various aspects of the present disclosure.

[0445] Computer-executable instructions may be in many forms, such as program modules, executed by one or more computers or other devices. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically the functionality of the program modules may be combined or distributed as desired in various embodiments.

[0446] Also, the data structures may be stored in any suitable form on the computer-readable medium. For ease of explanation, the data structures may be depicted as having fields that are related through their location in the data structure. Such relationships may equally be achieved by allocating storage to the fields with locations in the computer-readable medium that convey the relationship between the fields. However, any suitable mechanism may be used to establish relationships between information in the fields of the data structures, including using pointers, tags, or other mechanisms for establishing relationships between data elements.

[0447] When implemented in software, the software code may be executed on any suitable processor or collection of processors, whether provided on a single computer or across multiple computers.

[0448] Further, it should be understood that a computer may be embodied in any of a number of forms, such as, by way of non-limiting examples, a rack-mounted computer, a desktop computer, a laptop computer, or a tablet computer. In addition, a computer may be embedded in devices that are not generally considered computers, but have suitable processing capabilities, including personal digital assistants (PDAs), smartphones, tablets, or any other suitable portable or fixed electronic device.

[0449] A computer may also have one or more input and output devices. Such devices may be used, among other things, to present a user interface. Examples of output devices that may be used to provide a user interface include a printer or display screen for visual presentation of output and a speaker or other sound generating device for audible presentation of output. Examples of input devices that may be used in a user interface include keyboards and pointing devices such as mice, touch pads, and digitizing tablets. As another example, a computer may receive input information through voice recognition or in other audible forms.

[0450] Such computers may be interconnected by one or more networks of any suitable type, including local area networks or wide area networks, enterprise networks, and intelligent networks (IN) or the Internet, etc. Such networks may be based on any suitable technology and operate according to any suitable protocol, and may include wireless networks, wired networks or fiber optic networks.

[0451] Also, as described, some aspects may be embodied as one or more methods. Operations performed as part of the methods may be ordered in any suitable manner. Accordingly, embodiments may be constructed in which operations are performed in an order different from that shown as illustrated, including performing some operations simultaneously even though an example embodiment shows operations as sequential.

[0452] All definitions defined and as used herein should be understood to take precedence over dictionary definitions, definitions in documents incorporated by reference, and / or ordinary meanings of the defined terms.

[0453] The indefinite articles "a" and "an," as used herein in the specification and claims, unless expressly indicated otherwise, should be understood to mean "at least one."

[0454] The term "and / or," as used herein in the specification and claims, should be understood to mean "either or both" of the elements so conjoined, i.e., elements that are conjunctive in some cases and disjunctive in other cases. Multiple elements listed with "and / or" should be construed in the same manner, i.e., "one or more" of the elements so conjoined. Other elements may optionally be present beyond the elements specifically identified by the "and / or" clause, whether related or unrelated to those specifically identified elements. Thus, as a non-limiting example, a reference to "A and / or B," when used in conjunction with open-ended language such as "comprising," may refer in one embodiment to A only (optionally including elements other than B); in another embodiment to B only (optionally including elements other than A); in yet another embodiment to both A and B (optionally including other elements), etc.

[0455] As used herein in the specification and claims, the phrase "at least one" in reference to a list of one or more elements should be understood to mean that the at least one element is selected from any one or more of the elements in the list of elements, but may not necessarily include at least one of each and every element specifically listed in the list of elements, and may not necessarily exclude any combination of elements in the list of elements. This definition also allows for elements to be optionally present other than those elements specifically identified in the list of elements to which the phrase "at least one" refers, whether related or unrelated to those elements specifically identified. Thus, as a non-limiting example, "at least one of A and B" (or, equivalently, "at least one of A or B" or, equivalently, "at least one of A and / or B") can refer, in one embodiment, to at least one, and optionally including more than one, A, with no B (and optionally including elements other than B); in another embodiment, to at least one, and optionally including more than one, B, with no A (and optionally including elements other than A); in yet another embodiment, to at least one, and optionally including more than one, A, and at least one, and optionally including more than one, B (and optionally including other elements), etc.

[0456] In the claims, as well as in the above specification, all transitional phrases such as "comprising," "including," "carrying," "having," "containing," "involving," "holding," "composed of," and the like, shall be understood to be open-ended, i.e., meaning including, but not limited to. Only the transitional phrases "consisting of" and "consisting essentially of" shall be closed-ended or semi-closed-ended transitional phrases, respectively.

[0457] The terms "approximately," "substantially," and "about" may be used in some embodiments to mean within ±20% of a target value, in some embodiments within ±10% of a target value, in some embodiments within ±5% of a target value, and in some embodiments within ±2% of a target value. The terms "approximately," "substantially," and "about" may include the target value.

Claims

1. 1. A method for identifying types of cells present in a biological sample using cytometry and a plurality of machine learning models, comprising: Using at least one computer hardware processor, obtaining cytometry data for a biological sample previously obtained from a subject, the biological sample comprising a plurality of cells, the cytometry data including cytometry measurements obtained during respective cytometry events, the cytometry events corresponding to particular objects in the biological sample under measurement by a cytometry platform, the cytometry events including a subset of events corresponding to cells in the biological sample under measurement by the cytometry platform; obtaining a respective plurality of cell types by identifying a type of cell in the plurality of cells using the plurality of machine learning models, the plurality of machine learning models including a first machine learning model and a second machine learning model different from the first machine learning model, and the identifying includes, for each particular event in the event subset: obtaining from said cytometry data a cytometry measurement corresponding to said particular event; determining an event type for the particular event by processing the cytometry measurements corresponding to the particular event with the first machine learning model, indicating whether the event type for the particular event corresponds to a cell under measurement by the cytometry platform, debris under measurement by the cytometry platform, or beads under measurement by the cytometry platform; and determining a type of the cell by processing the cytometry measurements corresponding to the particular event with the second machine learning model when the determined event type indicates that the particular event corresponds to the cell being measured by the cytometry platform. Including To carry out A method comprising:

2. The method of claim 1 , wherein the event subset comprises at least 10,000 events.

3. The method of claim 1 , wherein the event subset comprises at least 100,000 events.

4. the first machine learning model includes a first multi-class classifier; the second machine learning model includes a second multi-class classifier. The method according to claim 1, 2 or 3.

5. the first machine learning model comprises a first decision tree classifier, a first gradient boosting decision tree classifier, or a first neural network; the second machine learning model comprises a second decision tree classifier, a second gradient boosting decision tree classifier, or a second neural network; The method according to any one of claims 1 or 2 to 4.

6. determining a percentage of the cellular composition of different cell types in said biological sample based on said identified plurality of cell types; The method of any one of claims 1 or 2 to 5, further comprising:

7. Determining the cellular composition percentage determining a first cellular composition percentage for said first cell type by determining a ratio between a number of cells identified as being of a first type in said plurality of cells and a total number of said cells in said plurality of cells. The method according to claim 6 or any one of claims 1 to 5, comprising:

8. The subject has, is suspected of having, or is at risk of having cancer, and the method comprises: Identifying a treatment for said subject based on said determined cellular composition percentage. The method of claim 6 or any one of claims 1 to 5 or 7, further comprising:

9. 8. The method of claim 8 or any one of claims 1 to 7, further comprising administering the identified treatment to the subject.

10. identifying the treatment for the subject based on the determined cellular composition percentage; Identifying ipilimumab for the subject when the cellular composition percentage of peripheral blood mononuclear cells (PBMCs) is below a threshold value. The method of claim 8 or any one of claims 1 to 7 or 9, comprising:

11. identifying the treatment for the subject based on the determined cellular composition percentage; Determining the ratio between the cellular percentage of CD8+PD-1+ cells and the cellular percentage of CD4+PD-1; and identifying an immune checkpoint blockade therapy for said subject when said determined ratio is above a threshold. The method of any one of claims 8 or 1-7 or 9-10, comprising:

12. comparing a cellular composition percentage of the determined cellular composition percentages to a cellular composition percentage range associated with a patient cohort; and identifying said subject as a member of said patient cohort based on the results of said comparison.

12. The method of claim 6, or any one of claims 1 to 5, or 7 to 11, further comprising:

13. 12. The method of claim 11, wherein the patient cohort comprises a healthy cohort, a cohort of diseased patients, or a cohort of treated patients.

14. comparing a cellular composition percentage of the determined cellular composition percentages with a cellular composition percentage range relevant to a study, the study evaluating the efficacy of one or more treatments in treating a disease; and and identifying a treatment for said subject based on the results of said comparison.

14. The method of claim 6, or any one of claims 1 to 5, or 7 to 13, further comprising:

15. the event subset corresponding to the cells in the biological sample under measurement by the cytometry platform comprises a first event subset; The cytometry event, a second subset of events corresponding to beads in the biological sample under measurement by the cytometry platform; and a third subset of events corresponding to debris in the biological sample under measurement by the cytometry platform. Further included are:

15. The method according to any one of claims 1 or 2 to 14.

16. The method of any one of claims 1 or 2 to 15, wherein the cytometric measurements corresponding to the particular event comprise fluorescence intensity values ​​for at least a portion of a plurality of markers.

17. the plurality of events includes a first plurality of events and a second plurality of events; the cytometry data includes first cytometry data for the first plurality of events and second cytometry data for the second plurality of events, the first cytometry data includes measurements obtained for a first marker of a plurality of markers during each of at least a portion of the first plurality of events, the second cytometry data includes measurements obtained for a second marker of the plurality of markers during each of at least a portion of the second plurality of events, and the first marker of the plurality of markers and the second marker of the plurality of markers are different; 17. The method according to any one of claims 1 or 2 to 16.

18. the first cytometry data comprises data from a first panel; the second cytometry data comprises data from a second panel different from the first panel.

17. The method according to claim 1 or any one of claims 1 to 16.

19. obtaining cytometry data for the biological sample includes obtaining flow cytometry data for the biological sample; the cytometry measurements obtained during each of the cytometry events include flow cytometry measurements obtained during each of the flow cytometry events.

19. The method according to any one of claims 1 or 2 to 18.

20. At least one non-transitory computer readable storage medium storing processor executable instructions which, when executed by at least one computer hardware processor, cause the at least one computer hardware processor to perform the method of any one of claims 1 to 19.

21. At least one computer hardware processor; At least one non-transitory computer readable storage medium storing processor executable instructions which, when executed by said at least one computer hardware processor, cause said at least one computer hardware processor to perform a method according to any one of claims 1 to 19; A system including:

22. Using at least one computer hardware processor, obtaining flow cytometry data for a biological sample from a subject, the biological sample including a plurality of cells including a first cell, and the flow cytometry data including first flow cytometry data for the first cell; determining a respective type for each of at least a portion of the plurality of cells using a machine learning model hierarchy corresponding to a cell type hierarchy, said determining including determining a first type for the first cell by processing the first flow cytometry data using a first subset of the machine learning model hierarchy; To carry out A method comprising:

23. The plurality of cells includes a second cell of a different type than the first cell, and the flow cytometry data includes second flow cytometry data for the second cell, and the determining includes: processing the second flow cytometry data using a second subset of the machine learning model hierarchy, the second subset of the machine learning model hierarchy being different from the first subset of the machine learning model hierarchy.

23. The method of claim 22, further comprising:

24. 24. The method of claim 22, or any one of claims 1 to 21 or 23, wherein said determining is performed on each of at least 10,000 cells from among said plurality of cells.

25. processing the first flow cytometry data using the first subset of the hierarchy of machine learning models; processing the first flow cytometry data using a first machine learning model in the first subset of the hierarchy of machine learning models; identifying a second machine learning model in the first subset of the hierarchy of machine learning models based on a first output of the first machine learning model; and processing the first flow cytometry data with the second machine learning model to obtain a second output.

25. The method of claim 22, or any one of claims 1 to 21, or 23 to 24, comprising:

26. 25. The method of claim 25, or any one of claims 1 to 24, wherein the first output indicates a first type for the first cell and the second output indicates a subtype of the first type for the first cell.

27. 26. The method of claim 26 or any one of claims 1 to 25, wherein said first type comprises white blood cells and said subtype of said first type comprises granulocytes.

28. 28. The method of claim 26 or any one of claims 1 to 25 or 27, wherein said first type comprises lymphocytes and said subtype of said first type comprises B cells.

29. 29. The method of any one of claims 26 or 1-25 or 27-28, wherein said first type comprises T helper cells and said subtype of said first type comprises memory T helper cells.

30. processing the first flow cytometry data using the first subset of the hierarchy of machine learning models; identifying a third machine learning model in the first subset of the hierarchy of machine learning models based on the first output of the first machine learning model; and Processing the first flow cytometry data using the third machine learning model.

30. The method of claim 25, or any one of claims 1-24, or 26-29, further comprising:

31. processing the first flow cytometry data using the first subset of the hierarchy of machine learning models; identifying a third machine learning model in the first subset of the hierarchy of machine learning models based on the second output of the second machine learning model; and Processing the first flow cytometry data using the third machine learning model.

31. The method of claim 25, or any one of claims 1-24, or 26-30, further comprising:

32. 32. The method of claim 22, or any one of claims 1 to 21, or 23 to 31, wherein a machine learning model in the hierarchy of machine learning models comprises a decision tree classifier, a gradient boosting decision tree classifier, or a neural network.

33. 33. The method of claim 22 or any one of 1 to 21 or 23 to 32, wherein the machine learning models in the machine learning model hierarchy comprise an ensemble of machine learning models.

34. determining a respective cellular composition percentage for each of said at least some of said determined cell types, said determining further comprising: determining a first cellular composition percentage for said first cell type; 34. The method of claim 22, or any one of claims 1 to 21, or 23 to 33, comprising:

35. 34. The method of claim 34, or any one of claims 1-33, wherein determining the first cellular composition percentage comprises determining a ratio between a number of cells of the first type and a total cell number of the at least some of the plurality of cells.

36. comparing the first cellular composition percentage to a cellular composition percentage range associated with a patient cohort; and identifying said subject as a member of said patient cohort based on the results of said comparison.

36. The method of claim 34 or any one of claims 1 to 33 or 35, further comprising:

37. 36. The method of claim 36 or any one of claims 1 to 35, wherein the patient cohorts comprise a healthy cohort, a cohort of diseased patients, and / or a cohort of treated patients.

38. generating a visualization of the determined cellular composition percentage, the visualization indicating the comparison of the first cellular composition percentage to the cellular composition percentage range associated with the patient cohort.

38. The method of claim 36, or any one of claims 1 to 35, or 37, further comprising:

39. comparing the first cellular composition percentage to a cellular composition percentage range associated with a study, the study evaluating the effectiveness of one or more treatments in treating a disease; and and identifying a treatment for said subject based on the results of said comparison.

39. The method of claim 34, or any one of claims 1-33, or 35-38, further comprising:

40. generating a visualization of the determined cellular composition percentages using the cell type hierarchy; 40. The method of claim 34, or any one of claims 1-33, or 35-39, further comprising:

41. 40. The method of claim 1, wherein the visualization comprises an indication of a cell type subset of the cell type hierarchy, and wherein the determined cell type subset cell composition percentages comprise aberrant cell composition percentages compared to a reference cell composition percentage.

42. 42. The method of claim 40, or any one of claims 1 to 39, or 41, wherein the visualization includes a plurality of nodes organized in a hierarchy with at least some pairs of nodes connected by respective edges, the plurality of nodes including a first node representing a first cell type and a second node representing a subtype of the first cell type, and the edges including a first edge connecting the first node and the second node.

43. 42. The method of claim 41, wherein the visualization further comprises, for the first cell type, a first number indicating a percentage of cells in the biological sample having the first cell type, the first number being shown in the visualization adjacent to the first node.

44. 44. The method of any one of claims 34 or 1-33 or 35-43, wherein determining the respective cellular composition percentages further comprises determining one or more second cellular composition percentages for one or more second cell types, wherein the one or more second cell types are subtypes of the first cell type.

45. comparing a sum of the one or more second cellular composition percentages to the first cellular composition percentage; determining a normalization factor based on a result of the comparison; and applying said normalization factor to said one or more second cellular composition percentages.

44. The method of claim 44 or any one of claims 1 to 43, further comprising:

46. determining whether the one or more second cell types include all subtypes of the first cell type; and applying the normalization factor to the one or more second cell composition percentages when the one or more second cell types include all of the subtypes of the first cell type.

45. The method of claim 45 or any one of claims 1 to 44, further comprising:

47. 47. The method of claim 45, or any one of claims 1-44 or 46, wherein the normalization factor comprises a ratio between the first cellular composition percentage and the sum of the one or more second cellular composition percentages.

48. comparing a sum of the one or more second cellular composition percentages to the first cellular composition percentage; determining a type-specific normalization coefficient for each of the one or more second cell types based on a result of the comparison; and applying said type-specific normalization factor to said respective one or more cellular composition percentages.

48. The method of claim 44, or any one of claims 1-43, or 45-47, further comprising:

49. 29. The method of claim 22, or any one of claims 1 to 21, or 23 to 48, wherein the obtained flow cytometry data comprises noise transformed flow cytometry data.

50. 50. The method of claim 22, or any one of claims 1-21, or 23-49, wherein the flow cytometry data comprises a fluorescence intensity value for each of at least a portion of a plurality of markers for each of at least a portion of the plurality of cells.

51. the plurality of cells includes a first plurality of cells and a second plurality of cells; the flow cytometry data includes a first flow cytometry data subset for the first plurality of cells and a second flow cytometry data subset for the second plurality of cells, the first flow cytometry data subset including fluorescence intensity values ​​for a first subset of the plurality of markers for each of at least a portion of the first plurality of cells, and the second flow cytometry data subset including fluorescence intensity values ​​for a second subset of the plurality of markers for each of at least a portion of the second plurality of cells, and the first subset of the plurality of markers and the second subset of the plurality of markers are different; 50. The method of claim 50 or any one of claims 1 to 49.

52. the first flow cytometry data subset comprises data from a first panel; the second flow cytometry data subset comprises data from a second panel.

51. The method of claim 51 or any one of claims 1 to 50.

53. 53. The method of claim 51 or any one of 1-50 or 52, wherein determining the respective types for each of the at least some of the plurality of cells comprises determining a first plurality of respective types for the first plurality of cells and determining a second plurality of respective types for the second plurality of cells, The method further comprising: determining a first plurality of cellular composition percentages using the first plurality of cell types, said determining including determining a respective cellular composition percentage for each of at least a portion of the first plurality of cell types; and determining a second plurality of cellular composition percentages using the second plurality of cell types, said determining including determining a respective cellular composition percentage for each of at least a portion of the second plurality of cell types. The method further comprises:

54. determining a respective cellular composition percentage for each of said at least some of said determined cell types, said determining including combining at least a portion of said first plurality of cellular composition percentages with at least a portion of said second plurality of cellular composition percentages.

53. The method of claim 53 or any one of claims 1 to 52, further comprising:

55. combining said at least a portion of said first plurality of cellular composition percentages with said at least a portion of said second plurality of cellular composition percentages based on estimated composition percentages of cell types included in said first plurality of cell types and said second plurality of cell types.

54. The method of claim 54 or any one of claims 1 to 53, further comprising:

56. combining said at least a portion of said first plurality of cellular composition percentages with said at least a portion of said second plurality of cellular composition percentages based on data obtained using beads contained in said biological sample.

56. The method of claim 54 or any one of claims 1 to 53 or 55, further comprising:

57. the biological sample includes a plurality of particles; the flow cytometry data includes flow cytometry data for a first particle of the plurality of particles, the first particle being a debris, a cell doublet, or a bead; 27. The method of any one of claims 22 or 1 to 21 or 23 to 56.

58. determining a respective particle type for each of at least a portion of the plurality of particles using the machine learning model hierarchy, said determining including determining a first particle type for the first particle using the machine learning model hierarchy.

57. The method of claim 57 or any one of claims 1 to 56, further comprising:

59. 59. The method of claim 57, or any one of claims 1-56 or 58, further comprising determining whether the first particle comprises the bead, the debris, the cell doublet, or the first particle is unidentifiable.

60. 60. The method of claim 22, or any one of claims 1-21, or 23-59, wherein obtaining said flow cytometry data comprises processing said biological sample using a flow cytometry platform.

61. 61. The method of claim 22 or any one of 1-21 or 23-60, wherein the machine learning model hierarchy comprises at least 250 machine learning models.

62. 62. The method of claim 22, or any one of 1-21, or 23-61, wherein the machine learning model hierarchy comprises at least 50 machine learning models.

63. 63. The method of claim 22 or any one of claims 1 to 21 or 23 to 62, wherein the first machine learning model comprises 20 hyperparameters.

64. At least one computer hardware processor; When executed by the at least one computer hardware processor, obtaining flow cytometry data for a biological sample from a subject, the biological sample including a plurality of cells including a first cell, the flow cytometry data including first flow cytometry data for the first cell; and determining a respective type for each of at least a portion of the plurality of cells using a machine learning model hierarchy corresponding to a cell type hierarchy, said determining including determining a first type for the first cell by processing the first flow cytometry data using a first subset of the machine learning model hierarchy. at least one non-transitory computer-readable storage medium storing processor-executable instructions for causing said at least one computer hardware processor to implement the method of the present invention; Including, the system.

65. At least one non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by at least one computer hardware processor, obtaining flow cytometry data for a biological sample from a subject, the biological sample including a plurality of cells including a first cell, the flow cytometry data including first flow cytometry data for the first cell; and determining a respective type for each of at least a portion of the plurality of cells using a machine learning model hierarchy corresponding to a cell type hierarchy, said determining including determining a first type for the first cell by processing the first flow cytometry data using a first subset of the machine learning model hierarchy. a non-transitory computer-readable storage medium for causing said at least one computer hardware processor to implement said

66. Using at least one computer hardware processor, obtaining flow cytometry data for a biological sample from a subject, the biological sample comprising a plurality of cells including a first cell and a second cell of a different type than the first cell, the flow cytometry data including first flow cytometry data for the first cell and second flow cytometry data for the second cell; and determining a respective type for each of at least a portion of the plurality of cells using a plurality of machine learning models, said determining comprising: determining a first type for the first cell by processing the first flow cytometry data using a first subset of the plurality of machine learning models, the first subset including a first machine learning model; and determining a second type for the second cell by processing the second flow cytometry data using a second subset of the plurality of machine learning models, the second subset including a second machine learning model that is not in the first subset of the plurality of machine learning models. containing, To carry out A method comprising:

67. 66. The method of claim 66, wherein determining the respective type for each of at least a portion of the plurality of cells is performed for each of at least 10,000 cells from among the plurality of cells.

68. the first machine learning model is a first machine learning model in the first subset; Determining the first type for the first cell comprises: processing the first flow cytometry data using the first machine learning model; identifying a second machine learning model in the first subset of the plurality of machine learning models based on an output of the first machine learning model; and processing the first flow cytometry data using the second machine learning model on the first subset to obtain a second output. Including, 68. The method of claim 66 or any one of claims 1 to 65 or 67.

69. 68. The method of claim 68 or any one of claims 1 to 67, wherein the first output indicates a type for the first cell, and the second output indicates a subtype of the type for the first cell.

70. 69. The method of claim 69 or any one of claims 1 to 68, wherein the type for the first cell comprises a leukocyte and the subtype of the type comprises a granulocyte.

71. 71. The method of claim 69 or any one of claims 1 to 68 or 70, wherein the type for the first cell comprises a lymphocyte and the subtype of said type comprises a B cell.

72. 72. The method of any one of claims 69 or 1-68 or 70-71, wherein said first type comprises T helper cells and said subtype of said type comprises memory T helper cells.

73. 73. The method of claim 66, or any one of claims 1-65, or 67-72, wherein a machine learning model in the hierarchy of machine learning models comprises a decision tree classifier, a gradient boosting decision tree classifier, or a neural network.

74. 74. The method of claim 66, or any one of 1-65, or 67-73, wherein the machine learning models in the machine learning model hierarchy comprise an ensemble of machine learning models.

75. determining a respective cellular composition percentage for each of said at least some of said determined cell types, said determining including: determining a first cellular composition percentage for said first cell type; Contains 75. The method of claim 66, or any one of claims 1-65, or 67-74, further comprising:

76. 75. The method of any one of claims 1-74, wherein determining the first cellular composition percentage comprises determining a ratio between a number of cells of the first type and a total cell number of the at least some of the plurality of cells.

77. comparing the first cellular composition percentage to a cellular composition percentage range associated with a patient cohort; and identifying said subject as a member of said patient cohort based on the results of said comparison.

77. The method of claim 75, or any one of claims 1 to 74, or 76, further comprising:

78. 77. The method of claim 77 or any one of claims 1 to 76, wherein the patient cohorts comprise a healthy cohort, a cohort of diseased patients, and / or a cohort of treated patients.

79. generating a visualization of the determined cellular composition percentage, the visualization indicating the comparison of the first cellular composition percentage to the cellular composition percentage range associated with the patient cohort.

80. The method of claim 77, or any one of claims 1-76 or 78, further comprising:

80. comparing the first cellular composition percentage to a cellular composition percentage range associated with a study, the study evaluating the effectiveness of one or more treatments in treating a disease; and and identifying a treatment for said subject based on the results of said comparison.

80. The method of claim 75, or any one of claims 1-74, or 76-79, further comprising:

81. generating a visualization of said determined cellular composition percentages using a cell type hierarchy; 81. The method of claim 75, or any one of claims 1-74, or 76-80, further comprising:

82. 81. The method of claim 81 or any one of claims 1 to 80, wherein the visualization comprises an indication of a cell type subset of the cell type hierarchy, and wherein the determined cell type subset cell composition percentages comprise aberrant cell composition percentages compared to a reference cell composition percentage.

83. 83. A method according to claim 81 or any one of 1 to 80 or 82, wherein the visualization includes a plurality of nodes organized hierarchically with at least some pairs of nodes connected by respective edges, the plurality of nodes including a first node representing a first cell type and a second node representing a subtype of the first cell type, and the edges including a first edge connecting the first node and the second node.

84. 83. The method of claim 83 or any one of claims 1 to 82, wherein the visualization further comprises, for the first cell type, a first number indicating a percentage of cells in the biological sample having the first cell type, the first number shown in the visualization adjacent to the first node.

85. 85. The method of any one of claims 75 or 1-74 or 76-84, wherein said determining further comprises determining one or more cell subtype composition percentages for one or more cell subtypes of said first cell type.

86. comparing the sum of the one or more cell subtype percent composition to the first cell subtype percent composition; determining a normalization factor based on a result of the comparison; and applying said normalization factor to said one or more cell subtype composition percentages.

85. The method of claim 85 or any one of claims 1 to 84, further comprising:

87. determining whether the one or more cell subtypes include all subtypes of the first cell type; and applying said normalization factor to said one or more cell subtype percentage composition when said one or more cell subtypes include all of said subtypes of said first cell type.

86. The method of claim 86 or any one of claims 1 to 85, further comprising:

88. 88. The method of any one of claims 86 or 1-85 or 87, wherein the normalization factor comprises a ratio between the first cell composition percentage and the sum of the one or more cell subtype composition percentages.

89. comparing the sum of the one or more cell subtype percent composition to the first cell subtype percent composition; determining a type-specific normalization coefficient for each of the one or more cell subtypes based on the results of the comparison; and applying said type-specific normalization factor to each of said one or more cell subtype percentage composition.

89. The method of claim 85, or any one of claims 1-84, or 86-88, further comprising:

90. 90. The method of any one of claims 66 or 1-65 or 67-89, wherein the obtained flow cytometry data comprises noise transformed flow cytometry data.

91. 91. The method of any one of claims 66 or 1-65 or 67-90, wherein the flow cytometry data comprises a fluorescence intensity value for each of at least a portion of a plurality of markers for each of at least a portion of the plurality of cells.

92. the plurality of cells includes a first plurality of cells and a second plurality of cells; the flow cytometry data includes a first flow cytometry data subset for the first plurality of cells and a second flow cytometry data subset for the second plurality of cells, the first flow cytometry data subset including fluorescence intensity values ​​for a first subset of the plurality of markers for each of at least a portion of the first plurality of cells, and the second flow cytometry data subset including fluorescence intensity values ​​for a second subset of the plurality of markers for each of at least a portion of the second plurality of cells, and the first subset of the plurality of markers and the second subset of the plurality of markers are different.

91. The method of claim 91 or any one of claims 1 to 90.

93. the first flow cytometry data subset comprises data from a first panel; the second flow cytometry data subset comprises data from a second panel.

92. The method of claim 92 or any one of claims 1 to 91.

94. 94. The method of claim 92 or any one of 1-91 or 93, wherein said determining comprises determining a first plurality of respective types for the first plurality of cells and determining a second plurality of respective types for the second plurality of cells, The method further comprising: determining a first plurality of cellular composition percentages using the first plurality of cell types, said determining including determining a respective cellular composition percentage for each of at least a portion of the first plurality of cell types; and determining a second plurality of cellular composition percentages using the second plurality of cell types, said determining including determining a respective cellular composition percentage for each of at least a portion of the second plurality of cell types. The method further comprises:

95. determining a respective cellular composition percentage for each of said at least some of said determined cell types, said determining including combining at least a portion of said first plurality of cellular composition percentages with at least a portion of said second plurality of cellular composition percentages.

94. The method of claim 94 or any one of claims 1 to 93, further comprising:

96. combining said at least a portion of said first plurality of cellular composition percentages with said at least a portion of said second plurality of cellular composition percentages based on estimated composition percentages of cell types included in said first plurality of cell types and said second plurality of cell types.

95. The method of claim 95 or any one of claims 1 to 94, further comprising:

97. combining said at least a portion of said first plurality of cellular composition percentages with said at least a portion of said second plurality of cellular composition percentages based on data obtained using beads contained in said biological sample.

97. The method of claim 95 or any one of claims 1 to 94 or 96, further comprising:

98. the biological sample includes a plurality of particles; the flow cytometry data includes flow cytometry data for a first particle of the plurality of particles, the first particle being a debris, a cell doublet, or a bead; 98. The method of any one of claims 66 or 1 to 65 or 67 to 97.

99. determining a respective particle type for each of at least a portion of the plurality of particles using the plurality of machine learning models, the determining including determining a first particle type for the first particle using the plurality of machine learning models.

98. The method of claim 98 or any one of claims 1 to 97, further comprising:

100. 100. The method of claim 98 or any one of claims 1-97 or 99, further comprising determining whether the first particle comprises the bead, the debris, the cell doublet, or the first particle is unidentifiable.

101. 100. The method of any one of claims 66 or 1-65 or 67-100, wherein obtaining the flow cytometry data comprises processing the biological sample using a flow cytometry platform.

102. 102. The method of claim 66, or any one of 1-65, or 67-101, wherein the machine learning model hierarchy comprises at least 250 machine learning models.

103. 103. The method of claim 66, or any one of 1-65, or 67-102, wherein the machine learning model hierarchy comprises at least 50 machine learning models.

104. 102. The method of claim 66, or any one of claims 1 to 65, or 67 to 103, wherein the first machine learning model comprises 20 hyperparameters.

Images