Detection of Early Lung Cancer in Sputum Using Automated Flow Cytometry and Machine Learning

Abstract

11. A system and method for analyzing a sputum sample from a subject suspected of having lung cancer, comprising: obtaining a plurality of cells from a sputum sample from the subject; marking the plurality of cells with i) a plurality of cell lineage-specific marker compositions, ii) a cell viability composition, and iii) a tetra(4-carboxyphenyl)porphyrin (TCPP) composition; analyzing the plurality of cells marked with i-iii in a flow cytometer to obtain a cell size selected subpopulation from the plurality of cells based on an automatically selected bead size exclusion gate; selecting a viable singlet population of cells from the cell size selected subpopulation using an automated non-debris gate and an automated singlet gate; obtaining a flow cytometer value from the viable singlet population of cells based on the plurality of cell lineage-specific marker compositions, the viability marker, and the TCPP marker; applying a trained classifier to metadata from the subject and the obtained flow cytometry values; and generating a classification of the sputum sample based on the application of the trained classifier, wherein the classification is selected from a plurality of classification options including cancer and non-cancer.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to and the benefit of U.S. Provisional Application No. 63 / 291,247, entitled "Detection of Early-Stage Lung Cancer in Sputum Using Automated Flow Cytometry and Machine Learning," filed December 17, 2021, and U.S. Provisional Application No. 63 / 357,994, entitled "Sputum Analysis by Flow Cytometry; an Effective Platform to Analyze the Lung Environment," filed July 1, 2022, and U.S. Provisional Application No. 63 / 390,826, entitled "Detection of Early-Stage Lung Cancer in Sputum Using Automated Flow Cytometry and Machine Learning," filed July 20, 2022. The present specification and claims thereto are incorporated herein by reference. STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] Not applicable. INCORPORATION-BY-REFERENCE OF MATERIAL SUBMITTED ON A COMPACT DISC

[0003] Not applicable. Statement of Prior Disclosure by Inventor or Co-Inventors

[0004] Not applicable. copyrighted work

[0005] Not applicable. [Background technology]

[0006] It should be noted that the following discussion concerns several publications by the authors and publication years, and that certain publications should not be considered prior art with respect to the present invention because of their recent publication dates. The discussion of such publications herein is provided for more complete context and should not be construed as an admission that such publications are prior art for purposes of patentability determinations.

[0007] Lung cancer accounted for an estimated 1.8 million deaths worldwide in 2020. An estimated 130,180 people will die from lung cancer in 2022 in the United States alone. Overall, the 5-year survival rate for lung cancer remains low at 22.9%, as most patients present with advanced disease. The American National Lung Screening Trial (NLST) demonstrated that LDCT screening detects 93.8% of lung cancers among high-risk individuals (i.e., people aged 55-74 years who have smoked for more than 30 pack-years and who are currently smoking or have quit smoking within the past 15 years). Low-dose computed tomography (LDCT) is the standard of care for lung cancer screening in the United States (US). Although LDCT has a sensitivity of 93.8%, its specificity of 73.4% results in potentially harmful follow-up procedures in patients who do not have lung cancer. Thus, there is a need for additional assays with high accuracy that can be used as an adjunct to LDCT to diagnose lung cancer. Low-dose spiral computed tomography may not yield a clear treatment pathway if the identified nodules are small.

[0008] The NLST showed that LDCT screening resulted in an overall 20% reduction in lung cancer-specific mortality compared with screening by chest radiography. Unfortunately, in this study, 96.4% of positive LDCT scans were false positives, and approximately 90% of patients with a positive LDCT underwent additional procedures to determine whether the nodules observed on the LDCT scan were cancerous. These procedures, including imaging, biopsy, and surgical resection, can cause severe adverse effects, including death. New guidelines for interpreting LDCT scans and models that estimate the probability that a nodule is cancerous have improved the false positive rate (FPR). Still, only a fraction of eligible patients undergo LDCT screening. Failure to communicate about the benefits and potential harms of screening (whether due to lack of knowledge or time), costs associated with LDCT, lack of LDCT access, and repeated radiation exposure from serial LDCT scans may all contribute to low adoption rates of screening.

[0009] A simple, non-invasive, radiation-free, cost-effective test that would assist physicians in making or ruling out a lung cancer diagnosis with more certainty could decrease unnecessary follow-up procedures and increase lung cancer screening. Sputum is an easily obtained body fluid that has long been part of lung cancer diagnosis. The PAP sputum cytology test, developed by Papanicolaou and optimized by Saccomanno, was the first lung cancer diagnosis by 1960. For this test, two sputum smear slides are labeled with PAP stain and read by a pathologist specialized in lung cytology. Although the sensitivity of sputum cytology is highly variable, its specificity is very high. A review of 16 published studies on sputum cytology involving more than 28,000 patients reported sensitivities ranging from 42% to 97%, with an average sensitivity of 66%, but specificities averaging 99%.

[0010] The low sensitivity of sputum cytology is due in part to insufficient samples and only a small portion of the sample being analyzed. This may be insufficient because the sample generated is saliva or because mucus / debris / red blood cells in the smear obscures the cellular components necessary for accurate analysis. Over time, modifications to the original sputum cytology test have improved its sensitivity. Nebulizers and auxiliary devices such as the Acapella and Lung Flute, as well as patient adherence to proper instructions on how to generate lung sputum samples, have been shown to improve patients' ability to generate sputum. Liquid cytology tests and automated slide preparation devices can reduce background contamination of sputum smears and therefore increase slide quality. Increasing the number of samples read has also been shown to increase the chances of finding abnormal cells indicative of lung cancer.

[0011] Porphyrins such as TCPP are currently used as diagnostic reagents in surgery to identify bladder cancer and the margins of cancerous tissue. Using microscopy, we have shown that by labeling sputum cells with the fluorescent porphyrin TCPP, study participants with lung cancer can be distinguished from those without disease with high accuracy using slide-based assays (cytology-based methods) and human graders. Cytology-based methods are of limited utility because reading slides is time-consuming and requires highly specialized personnel. Furthermore, extensive debris or too many squamous epithelial cells (SECs), or buccal cells, often result in samples that are inappropriate for diagnosis. Because slide-based assays are time-consuming and often preclude analysis of the entire sample, thereby potentially missing important events, and because human graders introduce subjective bias (also known as operator bias) that is not consistent between different human graders, alternative methods for analyzing sputum would be useful to determine the likelihood of cancer.

[0012] The feasibility of using a flow cytometry platform to analyze whole sputum samples to identify significant differences between samples obtained from individuals diagnosed with lung cancer and those without the disease, without clogging the instrument, is illustrated in accordance with one embodiment of the present invention. Summary of the Invention [Problem to be solved by the invention]

[0013] Early detection of lung cancer through screening may increase survival and decrease morbidity. Certain regions of the United States and the United Kingdom now advocate annual low-dose computed tomography (LDCT) screening for high-risk individuals. Thus, a positive LDCT result requires follow-up testing to determine whether the nodule is benign or malignant. These medical procedures carry inherent morbidity and mortality risks. 6 , which can impose a significant burden on screening participants and their families, and the associated costs represent a large financial burden for patients and society.

[0014] Therefore, efforts have been directed towards the development of non-invasive tests that can be used in conjunction with LDCT or as stand-alone tests to identify people who are at high risk of having lung cancer and should undergo LDCT. In either case, the goal of these tests is to identify lung cancer patients at an early stage while eliminating unnecessary medical procedures for low-risk patients. One of the easily accessible materials from the lung is sputum, which contains a variety of blood cells and exfoliated bronchial epithelial cells, including precancerous and malignant cells in patients with lung cancer. 7We previously reported a slide-based assay that could classify cancer and non-cancer patients from sputum stained with tetra(4-carboxyphenyl)porphyrin (TCPP). Although it was 81% accurate, reading the labeled slides was time-consuming, subject to observer bias, and could miss important low-frequency events due to undersampling. A high-throughput approach using automated flow cytometry (FCM) for sputum sample analysis could improve the shortcomings of slide-based analysis.

[0015] The disclosed embodiments combine flow cytometry and machine learning to develop a sputum-based test that can assist physician decision-making in such cases. [Means for solving the problem]

[0016] Summary of the Invention One embodiment of the present invention provides a flow cytometer method for analyzing a sputum sample from a subject suspected of having lung cancer. A plurality of cells from a sputum sample from a subject suspected of having lung cancer, such as a single cell suspension of a sputum sample, is obtained. The plurality of cells are marked with i) a plurality of cell lineage-specific marker compositions, ii) a cell viability composition, and iii) a tetra(4-carboxyphenyl)porphyrin (TCPP) composition. For example, i) comprises at least three, or at least four, or at least five, or at least six of CD206, CD3, CD19, CD66b, CD45, EpCAM, PanCK, and any combination thereof, and any combination may explicitly exclude any of CD206, CD3, CD19, CD66b, CD45, EpCAM, PanCK. In a further example, ii) the cell viability composition preferentially labels dead cells over live cells and may comprise FVS510. The plurality of cells marked with i-iii are analyzed with a flow cytometer to obtain a subpopulation selected for cell size from the plurality of cells based on an automatically selected bead size exclusion gate. For example, the bead size exclusion gate is set to 5 μm to about 30 μm, and events less than about 5 μm and greater than about 30 μm are not further analyzed. For example, the analyzing step includes obtaining flow cytometry values ​​for side scatter, forward scatter, fluorescence from TCPP, fluorescence from a cell viability composition, and fluorescence from a plurality of cell lineage specific marker compositions from the plurality of cells. From the subpopulation selected for cell size, a viable singlet population of cells is selected using an automated non-debris gate (e.g., the automated non-debris gate excludes most dead cells from the non-debris population) and an automated singlet gate (e.g., the automated singlet gate is applied to the population of cells selected with the automated non-debris gate). From the viable singlet population of cells, flow cytometer values ​​are obtained based on the plurality of cell lineage specific marker compositions, the viability marker, and the TCPP marker. The trained classifier is applied to metadata from the subject (e.g., age) and the resulting flow cytometry values.Based on application of the trained classifier, a classification of the sputum sample is generated, where the classification is selected from a plurality of classification options including cancer and non-cancer.

number

[0017] In one embodiment, the multiple lineage specific marker composition comprises fluorescent anti-CD206, fluorescent anti-CD3, fluorescent anti-CD19 and fluorescent anti-CD66b. In another embodiment, the classifier comprises a coefficient b determined by fitting a classifier model to the particular set of samples used to construct the classifier. 0 ~b 5 is the equation of a straight line containing

[0018] Another embodiment provides a system for automated analysis of flow cytometry data, the system comprising a computer processor in communication with a memory storing flow cytometry data from a plurality of markers in a plurality of cells from a sputum sample of a subject, the plurality of markers comprising i) a plurality of cell lineage-specific marker compositions, ii) a cell viability composition, and iii) a tetra(4-carboxyphenyl)porphyrin (TCPP) composition. For example, i) comprises at least three, or at least four, or at least five, or at least six of CD206, CD3, CD19, CD66b, CD45, EpCAM, PanCK, and any combination thereof, any combination may explicitly exclude any of CD206, CD3, CD19, CD66b, CD45, EpCAM, PanCK. In a further example, ii) the cell viability composition preferentially labels dead cells over live cells and may comprise FVS510. The system further provides a computer program product embodied in a non-transitory computer readable medium, the computer program product comprising instructions for causing a computer processor to: Flow cytometry data obtained from a plurality of cells from a sputum sample is received. From the plurality of cells in the sputum sample, a subpopulation of cells is selected automatically based on application of an automatic gate selected from a bead size exclusion gate, a viability gate, and a singlet gate. For example, the bead size exclusion gate is set between 5 μm and about 30 μm, and events less than about 5 μm and greater than about 30 μm are not further analyzed, for example, an automatic non-debris gate excludes most dead cells from the non-debris population, and for example, an automatic singlet gate is applied to the population of cells selected with the automatic non-debris gate. From the subpopulation, flow cytometry values ​​of interest for a plurality of cell lineage-specific marker compositions, a viability marker, and a TCPP marker are determined. A classifier is applied to the flow cytometry values ​​of interest and the metadata of interest, for example, the trained classifier is

number

[0019] Another embodiment of the present invention provides a non-transitory computer readable medium comprising program code that, when executed, causes a processing circuit to: Obtain flow cytometer values ​​(e.g., side scatter, forward scatter, fluorescence from TCPP, fluorescence from the cell viability composition, and fluorescence from the plurality of cell lineage specific marker compositions) for a viable singlet population of sputum cells of the subject based on a plurality of cell lineage specific marker compositions (e.g., the plurality of cell lineage specific marker compositions are selected from fluorescent anti-CD206, fluorescent anti-CD3, fluorescent anti-CD19, and fluorescent anti-CD66b), a viability marker (e.g., but not limited to, FVS510), and a TCPP marker; Apply a trained classifier to the metadata from the subject and the obtained flow cytometry values; Create a classification of the sputum sample based on the application of the trained classifier, where the classification is selected from a plurality of classification options including cancer and non-cancer; The method of claim 1, wherein the cell viability composition preferentially labels dead cells over live cells. In one embodiment, a bead size exclusion gate is set to exclude events less than about 5 μm and greater than about 30 μm, and a subpopulation of cells is selected from the sputum sample having a cell size that is not excluded from further analysis by the bead size exclusion gate, and from this population, a viable singlet population is selected. In another embodiment, the viable singlet population is selected using an automated non-debris gate that excludes most dead cells from the non-debris population for further analysis. In one embodiment, the trained classifier is

number

[0020] One aspect of the present invention provides a flow cytometry method, e.g., automated FCM, for analyzing sputum from a subject suspected of having lung cancer, the method comprising one or more of the following: 1) extracting from the analyzed sputum sample, e.g., bead standards and viability dyes; The method includes the following steps: 1) excluding contaminants, both debris and squamous epithelial cells (SEC, a common contaminant from the oral cavity), using a gating strategy defined by a 100% fluorochrome dye (100% fluorochrome dye), to generate a population of singlet cells; 2) including quality control parameters to detect alveolar macrophages in the samples, thereby verifying the pulmonary origin of each sputum sample; 3) defining a numerical cutoff for the population of cells of interest for the adequacy of the sample to provide a reliable analysis; 4) obtaining optical properties from sputum-derived cells labeled with one or more of a fluorescent specific antibody or its fragment and a leukocyte and / or epithelial cell lineage specific marker, such as TCPP, to identify significant differences between samples obtained from individuals diagnosed with lung cancer and those without the disease; 5) obtaining subject metadata, such as information on age and / or years of smoking; 6) applying a classifier based on the properties selected from the output of items 1-5; and 7) determining whether the analyzed sputum sample is higher or lower than the cancer likelihood numerical value. If cancer or likelihood of cancer is identified, the subject is processed for further testing.

[0021] One embodiment of the present invention provides a computer-implemented method for classifying a lung sputum sample from a subject at risk for lung cancer, comprising receiving data from the subject on at least one processor. The at least one processor is used to evaluate the data using a classifier that is an electronic representation of a classification system, the classifier being trained using a plurality of electronically stored training data sets, each of the plurality of training data sets representing a separate training data set, each of the separate training data sets representing an individual subject and data of each subject, each training data set further comprising a determination of a lung cancer characteristic, if present, for each subject, and the classification system comprises an identification of the lung cancer sputum sample as cancer or non-cancer. The at least one processor is used to evaluate the classification of the test sputum sample from the test subject based on the evaluating step. In one embodiment, the data comprises flow cytometry data, subject metadata, or a combination thereof. For example, the flow cytometry data is obtained from a sputum sample labeled with a plurality of markers in a plurality of cells from the sputum sample of the subject, the plurality of markers including i) a plurality of cell lineage-specific marker compositions, ii) a cell viability composition, and iii) a tetra(4-carboxyphenyl)porphyrin (TCPP) composition. The subject metadata includes one or more of gender, age, genetic information, biomarker data, smoking status, medical history, or combinations thereof. In a further embodiment, a non-transitory computer-readable medium storing an executable program includes instructions for performing a computer-implemented classification method. [Brief description of the drawings]

[0022] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate one or more embodiments of the invention and, together with the description, serve to explain the principles of the invention. The drawings are only for the purpose of illustrating one or more embodiments of the invention and are not to be construed as limiting the invention. In the following drawings:

[0023] [Figure 1]1 illustrates a flow chart of a sample identified according to one embodiment of the present invention.

[0024] [Diagram 2] Figures 2A-C illustrate automated gating of FCM data according to one embodiment of the present invention. Figure 2A illustrates bead size exclusion (BSE) gate parameters set on the entire sputum sample. Figure 2B illustrates a BSE gate threshold set based on observation of both forward (FSC) and side (SSC) scatter areas, 2.5x105 events above which are dead cells. Figure 2C illustrates a dot plot of NIST beads from an automated flowClust analysis and how it is used to set the lower left threshold of the BSE gate.

[0025] [Diagram 3]3A-3I illustrate automated gating of FCM data according to an embodiment of the present invention. FIG. 3A illustrates that in some samples, events appear in the lower right corner ("debris") of the forward scatter height (FSC-H) versus side scatter height (SSC-H) plot. FIG. 3B illustrates that the events identified by debris in FIG. 3A have very low optical SSC-A characteristics that are highly unusual for human cell populations. FIG. 3C-3I are histograms with density specified on the Y-axis and fluorescence intensity specified on the X-axis. FIG. 3C illustrates that the events identified by debris illustrated in FIG. 3A are identified as live because they lack staining with a viability dye (FVS510), a dye that preferentially stains dead cells. FIG. 3D illustrates that the debris identified in FIG. 3A expresses low levels of TCPP when labeled with TCPP. FIG. 3E illustrates that the debris events identified in FIG. 3A are negative for CD45. FIG. 3F illustrates that the debris events identified in FIG. 3A are negative for CD66b / CD3 / C19 when incubated with a compound that specifically recognizes CD66b, such as an antibody, a compound that specifically recognizes CD3, such as an antibody, and a compound that specifically recognizes CD19, such as an antibody. FIG. 3H illustrates that the debris events identified in FIG. 3A are negative for CD206 when incubated with a compound that specifically recognizes CD206, such as an antibody. FIG. 3G illustrates that the debris events identified in FIG. 3A are negative for Pan-CK when incubated with a compound that specifically recognizes Pan-CK. FIG. 3I illustrates that the debris events identified in FIG. 3A are negative for EpCAM when incubated with a compound that specifically recognizes EpCAM. The debris events identified by the profile in FIG. 3A are gated out to avoid false positive results according to one embodiment of the present invention.

[0026] [Figure 4]Figures 4A-4E illustrate a dot plot analysis and Figure 4F illustrates a histogram analysis of the non-debris events identified as illustrated in Figure 3A according to one embodiment of the present invention.

[0027] [Figure 5A] 2A and 3B illustrate dot plots of identified events where debris is excluded as in FIG. 3A. [Figure 5B] 2A and 3B illustrate dot plots of identified events in which debris is excluded as in FIG. 3A. [Figure 5C] 2A and 3B illustrate dot plots of identified events in which debris is excluded as in FIG. 3A. [Figure 5D] 2A and 3B illustrate dot plots of identified events where debris is excluded as in FIG. 3A. [Figure 5E] 2A and 3B illustrate dot plots of identified events where debris is excluded as in FIG. 3A. [Figure 5F] 2A and 3B illustrate histograms of identified events where debris is excluded as in FIG. [Figure 5G] 2A and 3B illustrate dot plots of identified events in which debris is excluded as in FIG. 3A. [Figure 5H] 2A and 3B illustrate dot plots of identified events in which debris is excluded as in FIG. 3A.

[0028] [Figure 6A] 1 illustrates a histogram showing cells stained with TCPP, with the x-axis being TCPP / log10 SSC and the y-axis being density. [Figure 6B] 1 illustrates a histogram showing cells stained with TCPP, with the x-axis being TCPP / log10 SSC and the y-axis being density. [Figure 6C] Illustrates a histogram showing cells stained with FVS510 (a viability dye), with the x-axis being FVS510 / log10 FSC and the y-axis being density. [Figure 6D] Illustrates a histogram showing cells stained with FVS510 (a viability dye), with the x-axis being FVS510 / log10 FSC and the y-axis being density. [Figure 6E] Illustrated is a dot plot with CD206 on the x-axis and CD66b / CD3 / CD19 on the y-axis. [Figure 6F] Illustrated is a dot plot with CD206 on the x-axis and CD66b / CD3 / CD19 on the y-axis.

[0029] [Figure 7] 1 illustrates a flow chart of data analysis for a sputum-based pulmonary assay according to one embodiment of the present invention.

[0030] [Figure 8A] 1 illustrates a receiver operator c (ROC) graph of the calculated false positive rate versus the true positive rate for varying the model response threshold. [Figure 8B] 8B is a graph of the model value (likelihood of cancer) based on the ROC curve of FIG. 8A.

[0031] [Figure 9] Illustrated is a correlation graph of age on the x-axis versus model value on the y-axis when samples were analyzed on an LSRII flow cytometer or a NaviosEX cytometer, respectively.

[0032] [Figure 10] 1 illustrates an exemplary system according to one embodiment of the systems disclosed herein.

[0033] [Figure 11] 1 illustrates one embodiment of classifier development according to one embodiment of the present invention.

[0034] Before describing the present invention in more detail, it is to be understood that this invention is not limited to particular embodiments described, as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the present invention will be limited only by the appended claims. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0035] Sputum is a biological fluid that can be obtained non-invasively and dissociated to release its cellular contents, providing a snapshot of the lung environment. Sputum can be made into a single cell suspension and stained with both TCPP-conjugated and fluorochrome-conjugated antibodies for manual and automated flow cytometry (FCM) analysis. Automated FCM allows for analysis of sputum samples for cancer or cancer-related cells using TCPP while providing information about the lung environment from the subject providing the sputum sample to interrogate the sputum sample using a panel of compounds specific for cell lineage markers and to capture predictive features of the sputum sample using trained classifiers. In one embodiment, the sputum sample analyzed by flow cytometry contains about 10,000 cells in a subpopulation that is selected by automated gating and analyzed and classified in real time by the systems and methods disclosed herein. In other embodiments, the total number of cells in the subpopulation can be about 1,000-5,000, about 5,000-10,000, about 10,000-100,000. Data acquired by automated FCM may be stored and analyzed at a later time rather than being analyzed in real time as the data is acquired.

[0036] Over the past two decades, the field of automated FCM analysis has spawned powerful software tools to identify cell populations that correlate with clinical outcomes and manage increasingly complex FCM datasets. Much effort has focused on replicating expert analysis of FCM data to automate cell population identification, for example in human immune profiling. Such data-driven algorithms can match or exceed human expertise and can fully automate the analysis of acquired FCM data, thereby eliminating potential operator bias. However, the application of automated flow cytometry analysis of sputum samples to determine lung health has proven challenging due to the complex nature of the lung environment, the interplay between inflammatory markers and disease, and the high frequency occurrence of cells of interest in sputum samples. Complicating factors in analyzing sputum samples for lung health include the variable size of sputum samples, as samples that are too small contain too few cells for analysis and samples that are too large may attenuate events of interest if the events are rare. Cytological workup of sputum samples on slides is limited to only the specimen fragments, thereby reducing the sensitivity of the slide assay, which is dependent on the skill of the observer to see significant events on the slide. Smears on the slide and overlapping of cells on the slide reduce the usefulness of cytology to provide a meaningful analysis when looking for rare events in the sample. Further cytology does not provide a complete picture of which non-malignant cells may be present in the sample and their incidence in the sample, and what that information means for lung health as it relates to lung cancer or other diseases of the lung.

[0037] Another aspect of one embodiment of the present invention provides a supervised learning approach to develop an assay to isolate viable single cell events in combination with automated FCM data acquisition from induced sputum and classify patient samples as cancer or non-cancer using machine learning techniques.

[0038] In one embodiment of the present invention, the developed lung cancer / non-cancer classifier performs well with a sensitivity of 82% and a specificity of 88%. Furthermore, the classifier achieves comparable sensitivity and specificity when applied to an independent set of samples collected using a different flow cytometer platform (Navios EX) than that used for assay development (LSRII). One aspect of the automated FCM lung assay system and method according to one embodiment of the present invention is that the assay is accurate even when early stage (I and II) and lung nodules are small (diameter <20 mm). The system and method of one embodiment of the present invention is robust to differences in sample handling and processing and captures important predictors of early lung carcinogenesis.

[0039] In one embodiment of the present invention, sputum was obtained from current and past smokers identified as having lung cancer or at high risk of developing the disease, e.g., with a smoking history of 20 pack years. The isolated sputum cells were counted, viability determined, and labeled with a panel of markers to determine cell type, e.g., anti-CD45 to separate leukocytes from non-leukocytes, although other markers are possible as discussed herein. After excluding debris and dead cells, including squamous epithelial cells, we identified reproducible population signatures and confirmed the pulmonary origin of the samples. For example, in addition to labeling sputum samples with leukocyte- and epithelial-specific fluorescent antibodies, fluorescent meso-tetra(4-carboxyphenyl)porphyrin (TCPP), known to preferentially stain cancer (associated) cells, was also used to label sputum samples. Differences in cellular characteristics, population size, and fluorescence intensity were identified that are useful for distinguishing cancer samples from high-risk samples.

[0040] In one embodiment of the present invention, an analytical pipeline was developed that combines automated flow cytometry processing with machine learning to distinguish cancer from non-cancerous cells in sputum samples. Flow data and patient characteristics were evaluated to identify predictors of lung cancer. The training set was used to fit the model, and the remaining samples were used for independent validation (test set). This approach was further validated in a second set of samples processed on a different flow cytometry platform.

[0041] Referring now to FIG. 1, the utilization of sputum samples is illustrated in a flow chart. Of the 171 samples run on the LSRII flow cytometer that were originally considered (136 non-cancer, 31 cancer, 4 with unidentified health status), 168 samples were used for model building and analysis pipeline development. This included 4 samples for which we did not have a definitive disease status, for which the addition of unlabeled samples would have been useful for model building. Additionally, 14 samples that were flagged as disqualified based on cell count (see below) were also used in model development to better capture the distribution of the underlying data and help make the model generalization more robust to sample noise. Three samples could not be used at all due to technical issues during acquisition.

[0042] 150 samples were finally used for the model validation phase (122 non-cancer, 28 cancer). Eighteen of the 168 samples were omitted, 13 contained too few cells for accurate analysis, one contained too few alveolar macrophages to be confirmed as a lung sample, and four samples were excluded because their cohort status could not be confirmed. An independent validation of the automated analysis was performed on 32 new samples. Participants adhered to the same enrollment criteria and samples were processed with the same protocol as the previous sample set. A second set of samples was run using a different flow cytometer (Navios EX), but the same model and coefficients were used to analyze the data from both instruments. Of the 171 samples run on the LSRII, originally considered (136 high risk, 31 confirmed cancer, 4 unconfirmed), 150 samples were finally used for the automated analysis pipeline development (122 high risk, 28 confirmed cancer). Of the 21 omitted samples, 13 contained too few cells for accurate analysis, 1 contained too few alveolar macrophages to be confirmed as a lung sample, and 3 indicated technical issues encountered during acquisition. Four further samples were excluded because their cohort status could not be confirmed. Of the 45 samples processed on Navios EX, 7 were excluded due to too few cells, 1 was excluded due to too few macrophages, and 5 were excluded due to technical issues (1 during processing and 4 by the flow cytometer). The remaining samples consisted of 26 high-risk and 6 cancer samples.

[0043] Referring now to Figure 2A, bead size exclusion (BSE) gate parameters are set for the entire sputum sample. Figure 2B shows the BSE gate threshold set at 2.5 x 10 based on observation of both forward (FSC) and side (SSC) scatter areas. 5Figure 2C illustrates a dot plot of NIST beads from the automated flowClust analysis and how it is used to set the lower left threshold of the BSE gate. The lower left threshold of the BSE gate is derived from the automated NIST beads flowClust analysis. Single cell suspensions from 3-day sputum samples were labeled with viability dyes to exclude dead cells, antibodies to distinguish cell types, and porphyrins to label cancer-associated cells and run on a flow cytometer.

[0044] One step towards the development of a computationally automated lung cancer assay is the automated flow cytometer (FCM) identification of viable single cells, which involves the sequence of events illustrated by Figures 2A, 3A, 4A, 4B according to one embodiment of the present invention. In one embodiment of the assay, the sample preparation component of the method consists of multiple assay tubes, e.g., one assay tube may be labeled with a blood cell marker ("blood tube") and another assay tube may be labeled with an epithelial cell marker ("epithelial tube"). For example, both tubes may contain a fluorescent anti-CD45 antibody that selectively binds to blood cells, as well as a viability dye to facilitate viability gating (see viability gate in Figure 4A) to exclude a population of dead cells, including squamous epithelial cells (SECs), from further analysis and TCPP to identify cancer or cancer-associated cells. Cells in the blood tube may be labeled with anti-CD206 (a marker for lung macrophages) and / or anti-CD3 (a marker for T cells) in addition to or instead of CD45, and / or anti-CD206 (a marker for lung macrophages) in addition to or instead of CD45, and / or anti-CD3 (a marker for T cells). may be stained with anti-CD19 (a marker for B cells) and / or anti-CD66b (a marker for granulocytes). Cells in the epithelial tube may be stained with pan-cytokeratin (panCK) antibodies and / or antibodies against epithelial cell adhesion molecule (EpCAM). Viability dyes, TCPP cell markers and blood cell markers and epithelial cell markers may be distinguished from each other based on their optical properties, e.g., fluorescence. In one embodiment of the present invention, the fluorescence intensity, forward scatter and side scatter of cell subpopulations selected from the more comprehensive cell population based on selection criteria including one or more of bead size exclusion gating, singlet gating, viability dye gating, cell lineage marker gating and TCPP gating of flow cytometry sputum data were used for downstream numerical analysis.

[0045] Additionally, in one embodiment, control samples may include one or more of the following: polystyrene beads of known approximate diameter (5-30 μm NIST beads), compensation samples for each fluorochrome channel used, an unstained sputum sample, and a cell lineage / type marker sample (e.g., sputum control for antibody isotype). Each sample tube corresponds to a single flow cytometry standard (fcs) file, which contains sample metadata and per-event values ​​for each optical and fluorescent channel acquired, as well as time parameters recorded when the contents of the sample tube are interrogated and / or acquired by the flow cytometer.

[0046] SECs are highly autofluorescent and may result in false positive events when sputum samples are interrogated and / or analyzed by flow cytometry. Therefore, the exclusion of SECs from the population of cells analyzed during the analysis of sputum samples is a step in one embodiment of the lung assay. We confirmed that neither physical exclusion of SECs by filtration before analysis nor negative size selection during analysis resulted in the exclusion of SEC cells. A live / dead cell discriminator (FVS510) to exclude SECs from the population of cells from the sputum sample being analyzed was one solution to exclude SECs from the population of events acquired by the flow cytometer and / or analyzed in downstream analysis.

[0047] Separated sputum cells within the size parameter of 5-30 μm were analyzed for viability if the sputum cells of interest and SECs fell within this gate area. The cutoff for FVS510 positivity was based on unstained controls. Backgating dead cells on the light scatter profile of sputum showed that these cells generally had a high SSC, which may be expected with SECs. To confirm that SECs were dead, sputum cells were sorted into dead and live cell populations. Aliquots of the pre-sorted samples and sorted populations were transferred to cytospins and stained with Wright-Giemsa. These slides showed that SECs were primarily present among the dead cells, whereas live cells sorted from the same samples contained hematopoietic and non-hematopoietic cells, as well as a small number of contaminating SECs. It was therefore determined that sputum samples could be analyzed by flow cytometry while excluding contaminating SECs using a viability gate.

[0048] Figure 2A shows the steps to limit the events analyzed downstream based on forward scatter area (FSC-A) and side scatter area (SSC-A), both of which are reasonable surrogates for cell size. A two-dimensional cluster gate was used to find the main peak of 5 μm NIST beads in FSC-A vs. SSC-A (Figure 2C). The lower FSC-A limit of the bead clusters was set as the minimum sample FSC-A to exclude small particles and debris. Events above these thresholds were considered to be death (FVS510 + ) cells (Figures 2A and 2B). 5 An upper limit of was set for both FSC-A and SSC-A.

[0049] 3A-3B illustrate dot plot data setup according to one embodiment of the present invention. Referring now to FIG. 3A, in some samples, events appear in the lower right corner of the FSC-H plot vs. SSC-H plot. These events are gated from further analysis to avoid inclusion of the events / cells as live marker-negative TCPP low cells, as shown in FIG. 3C-3I. FIG. 3A illustrates that a temporary flowClust gate (non-debris gate) is set on non-debris events in FSC-H vs. SSC-H to retain mostly live cells for final FVS510 tail gating ("core viability gate"). FIG. 3B illustrates dot plots of low SSC-A and debris with low FSC-A, showing the identification of additional debris (FIG. 3A) and showing that this debris exhibits a light scattering profile that is not typical for the cell population. Additional analysis (Figures 3A-I) indicates that this debris can be mistaken for live cells (Figure 3C) that are low or negative for other markers (Figures 3D-I), so it needs to be excluded.

[0050] Events were then restricted within a bead size exclusion (BSE) gate to eliminate populations with abnormal FSC and SSC height profiles (Figure 3A) and staining profiles that may result in inappropriate inclusion in downstream analyses. Exclusion of non-normal appearing populations is generally justified.

[0051] Referring now to Figure 4A, a viability gate (solid rectangle) is set on the remaining events based on FVS510-A fluorescence. Cells within the viability gate (set at the threshold for FVS510 positivity) were retained (Figure 4A). In some samples, setting the viability gate was difficult due to patient-to-patient variability in the cellular composition of sputum. In those cases, a heuristic-guided viability gate was used (Figure 4C-F).

[0052] Now referring to Figure 4B, a singlet gate (solid polygon) is set for all viable events (events inside the viability gate in Figure 4A). The identified (viable) singlets are used for downstream numerical analysis. In some samples, setting the singlet gate was difficult due to patient-to-patient variability in the cellular composition of sputum. In those cases, a heuristic-guided viability gate setting was used (Figure 5C-H).

[0053] Figures 4C-E illustrate dot plots of heuristic-guided viability gate setting. The subpopulation most likely to contain viable singlet cells (i.e., cells with relatively small area and height in the light scatter channel) is used to guide the positioning of the viability gate (Figures 4C and 4D). Figure 4C illustrates a sample with a low percentage of high SSC / FSC cells. A temporary flowClust gate set on non-debris events in FSC-H vs. SSC-H to retain mostly viable cells for final FVS510 tail gating ("core viability gate"). Figure 4D illustrates flowClust was re-run more inclusively by increasing the "quantile" parameter to 0.99 for samples with <10% events within the core viability gate. Figure 4E illustrates a temporary singlet gate set on core viability events, with the top right point at 2.5 x 10 on both axes. 5 Force capture of the upper diagonal by setting it to . Figure 4F illustrates a histogram of FVS510 staining (viability dye) of events identified in the transient singlet gate of Figure 4E. The viability cutoff is automatically set to the core viability singlet (black histogram). The line identified as the "blue curve" is a complete non-debris FVS510 profile for comparison. The vertical line identified as the "red bar" indicates the viability gate cutoff. Viable events are to the left of the viability gate cutoff. Once the viability cutoff has been determined, remove all transient gates. The viability gate cutoff thus determined is used to set a viability gate on the entire sample (shown in Figure 4C and Figure 4D).

[0054] Referring now to Figures 5A-5D, heuristic-guided singlet gating is illustrated. In some cases, cells with high SSC-A (see, for example, Figure 3A) prevent singlet gating in the complete viable cell population. Figure 5A shows a dot plot with FSC-A on the x-axis and FSC-W on the y-axis, a singlet gate (solid polygon) that was automatically assigned to a sample with cells with too high SSC-A. A wide singlet gate is an incorrect gate. Such a gate can be corrected by fitting the gate to a transient subpopulation of viable cells that excludes high SSC-A cells (Figure 5B). Figure 5B shows a dot plot with FVS510-A on the x-axis and SSC-A on the y-axis, with approximately 5 × 10 4 Figure 5C is an example with a temporal gate setting that excludes events exceeding 2.5 × 10 in both axes (dashed polygons) that was automatically fitted to a restricted population (solid polygon) with the upper right corner at 2.5 × 10 in both axes (dashed polygons). 5 Figure 5D illustrates the refined singlet gate (using cells identified in the get shown in Figure 5B) adjusted to include the upper diagonal by setting the gate to 0. Figure 5D illustrates the refined singlet gate (solid polygon) applied to the complete live population when the temporary gate is removed (compare the refined polygon in Figure 5D with the singlet gate (solid polygon) in Figure 5A).

[0055] Referring now to Figures 5E-H, heuristic-guided singlet gating is illustrated for different types of cases that prevent the setting of the singlet gate. In these cases, the population that accounts for approximately >10% of singlets is between approximately 2.5 (logicle scale) and the viability threshold that prevents the setting of the viability gate and the singlet gate. This can be corrected by resetting the viability gate and fitting the singlet gate to the more restricted viable population. Figure 5E shows a dot plot of such a difficult case with FVS510-A on the x-axis and FSC-A on the y-axis. The solid oval represents >10% of singlets and the dashed line represents the viability threshold. Figure 5F illustrates the population mixture analysis exemplified in a density histogram of FVS510 staining. The analysis highlights the difference in signal distribution of the right-most population identified by the solid ellipse in Figure 5E (identified as the blue curve) relative to the majority of events located to the left of the ellipse in Figure 5E (identified as the black curve), suggesting that the natural cutoff (dashed line) for these aberrant cases is 2.5. Figure 5G illustrates a dot plot with an adjusted viability gate (solid box) replacing the one found by the automated tailgating (dashed line). Figure 5H illustrates a dot plot with FSC-A on the x-axis and FSC-W on the y-axis, along with a new singlet gate (solid polygon) calculated for the refined viable cell population identified with the "adjusted viability gate shown in Figure 5G".

[0056] A "singlet" gate (FSC area vs. FSC width) was used to exclude cell doublets or small aggregates (Figure 5A). Due to sample variability from patient to patient, it was difficult to automatically set the singlet gate. In some samples, high SSC-A cells were included in the viable cell population, skewing the results of the singlet gating algorithm. This can be corrected by fitting the gate to a transient subpopulation that excludes most of these high SSC-A events (Figure 5B). In other samples, two populations were found within the viable cell gate, one with low FVS510 staining and the other just below the viability threshold and with a high side scatter profile. Correction in this case involves resetting the viability gate and fitting the singlet gate to the more restricted viable population. Light scatter and fluorescence signal values ​​were recorded for each single event and used for downstream model development and validation.

[0057] Based on our previous slide-based assay results, we expected that smoking history (or correlated factors such as age) and TCPP signal density (as opposed to fluorescence intensity itself) would be important predictors. Therefore, we scaled the fluorescence signals in all channels as log 10 FSC-A or log 10 SSC-A partitioned the resulting density distribution into three regions (R1 = approx. <0.25, R2 = approx. 0.25-0.6, R3 = approx. >0.6, Figure 6A-Figure 6D). Two such density signals proved to be informative for the classifier: TCPP / log 10 SSC-A (Figure 6A, region 3 [R3]) and FVS510-A / log 10 FSC-A (Figure 6C, region 2 [R2]). TCPP / log 10 The predictive value of SSC-A signal density was not imposed in the stepwise regression but emerged spontaneously. The fact that FVS510-A signal density was also found to be informative is interesting and may be related to the fact that apoptotic cells can take up this dye at intermediate levels.

[0058] Figure 6A illustrates a histogram of a live single cell stained with TCPP showing sections R1, R2 and R3, with the R3 section identified by a rectangular solid box. Region R3 (shaded) represents a viable singlet with high TCPP signal relative to side scatter (log10 transformed). Figure 6B illustrates a histogram of a live single cell stained with TCPP showing sections R1, R2 and R3, with the R3 section identified by a rectangular solid box. Region R3 (shaded) represents a viable singlet with high TCPP signal relative to side scatter (log10 transformed). 10 FIG. 6C illustrates a histogram in which the y-axis is density, cells without TCPP (unstained control) are displayed, and region R3 is event-free. FIG. 6D illustrates a histogram in which the x-axis is FVS510 / log 10 Figure 6D illustrates histograms of viable single cells, with FSC, density on the y-axis, and cells negative for FVS510 (R1) or stained with low levels of FVS510 (R2; solid rectangular shaded box). While cells in R1 and R2 are all below the FVS510 positivity cutoff, cells in R2 nevertheless show a relatively high FVS510 signal relative to forward scatter (log10 transformed) compared to the unstained control sample in Figure 6D. ... 10 Shown is a histogram with FSC, y-axis being density, and illustrating cells that were not stained with FVS510 (unstained control).

[0059] Combinations of lineage markers can identify subpopulations that cannot be captured by a single lineage marker alone. Careful investigation of patient sputum samples by FCM revealed complex patterns of lineage marker expression in blood and epithelial tracts, but it was not clear what information was revealed about lung health.

[0060] FIG. 6E illustrates a dot plot with CD206 on the x-axis and CD66b / CD3 / CD19 on the y-axis. The solid shaded box on the low quadrant of CD206 identifies non-macrophages, while the solid boxes on the mid to high quadrants of CD206 identify pulmonary macrophages. Anti-CD206 is a macrophage cell marker specific for a macrophage population present in lung tissue and not found in the blood circulation. Anti-CD66b is a granulocyte cell marker, anti-CD3 is a T cell marker, and anti-CD19 is a B cell marker. FIG. 6F illustrates a dot plot with CD206 on the x-axis and CD66b / CD3 / CD19 on the y-axis, where the solid shaded box on the low quadrant of CD206 identifies the absence of non-macrophages, while the solid boxes on the high quadrant of CD206 identify the absence of pulmonary macrophages from the control sample. FIG. 6E illustrates a dot plot with CD206 on the x-axis and CD66b / CD3 / CD19 on the y-axis, where the solid shaded box on the low quadrant of CD206 identifies the absence of non-macrophages, while the solid boxes on the high quadrant of CD206 identify the absence of pulmonary macrophages from the control sample. 低 ) White blood cells (CD66b / CD3 / CD19 中 The shaded area (CD206) is an example of a tan shaded area containing 中 / 高 CD66b / CD3 / CD19 低~中 ) contain pulmonary macrophages, which are used as a marker for samples deemed to adequately sample the lung environment. Compare with the unstained control sample in Figure 6F, where quadrant labeling also applies to Figure 6E. All panels are from the same exemplary sample.

[0061] Analysis of paired cell markers by distributing fluorescence based on signal distribution in blood (Figure 6E-F) and epithelial tubes (data not shown) was analyzed. Logicle scale signal intensity was quantified into low (<1.5), low-medium (1.5-2.5), medium (2.5-3), and high (>3) windows. Events per 10,000 were tallied for each area of ​​the resulting 4x4 grid of CD206 vs. CD3 / CD19 / CD66b (blood tubes) or EpCAM vs. panCK (epithelial tubes not shown). The area of ​​the blood signal intensity grid proved to be informative for the lung assay model of the predictive classifier (tan shaded rectangle; low levels for CD206 and medium levels for CD3 / CD19 / CD66b, Figure 6E). This population may indicate the presence of immune or inflammatory processes in the lung.

[0062] The development of automated processing of flow cytometer data features of sputum samples acquired on an LSRII flow cytometer in accordance with one embodiment of the present invention provides one or more setup and / or quality control steps prior to automated analysis of patient tubes / samples as follows: For each sample tube, remove outlying events using, for example, the time vs. fluorescence channel, e.g., using [flowCut] The compensation tube is used to automatically derive the spillover matrix, for example using [flowStats]. Note that on the Navios flow cytometer, an automated process is initiated with the operator spill matrix from the unstained sample, the alignment of the medians in the "off" channel (i.e., the non-FITC channel, if FITC is being compensated) is checked and fine-tuned if necessary based on the expectation that the means of the positive and negative populations (as defined in the channel being compensated, e.g., FITC) should be aligned in the off channel (e.g., the non-FITC channel). · The fluorescence signal is compensated and converted to a logic scale, e.g. using [flowWorkspace]; Note that Navious's automated processing converts the 20-bit signal (0-1,048,575) into a 60-year period to roughly match the LSRII range. Automatically find the smallest NIST bead peak and set a lower FSC-A threshold (low) to exclude bulk debris, e.g., using [openCyto]. For example, 10 if using an LSRII flow cytometer. 4 (low)~5×10 4 (High) window is the area of ​​interest; For Navio: 10 4 (low)~10 5 Note that you are looking at the (High) window. Further automated analysis of patient tubes / samples according to one embodiment of the present invention results in: Patient Tube: Events within the rectangular gate are retained for further analysis, “BSE” (FSC-A: low ~2.5×10 5 , SSC-A: 0 to 2.5 × 10 5 ). [flowWorkspace] Navios: "BSE" (FSC-A: Low to 10 6 ,SSC-A:5x10 3 ~10 6 ) A secondary "NON-DEBRIS" gate was applied to the maintained BSE population to identify cells with high FSC-H (>2×10 5 ), moderate / high SSC-H (>1 × 10 5 ) to filter out occasional abnormal cell populations. [flowWorkspace]; In Navios: The equivalent of the LSRII NON-DEBRIS gate is not the FSC-H vs. SSC-H exclusion gate, but the SSC-W vs. SSC-A (bottom left = 50,50; top left = 50,10 6 ;Bottom right=200,50;Top right=10 3 ,10 6), the selection polygon "cleanSSC" defined on the channel. The choice of channel depends on how well the debris can be visualized and excluded. This gate is applied at the end of Navios processing on the retained live singlets. A subset of cells most likely to contain the live singlet population of interest (dummy gating set) is utilized to set the viability gate coordinates. This helps reduce artifacts and distortions caused by contaminating squamous epithelial cells (SECs) and other dead cells. For LSRII: FSC-H vs. SSC-H, set up an automatic flowClust gate for NON-DEBRIS events using "hxh" to exclude the majority of contaminating SEC, and ensure both channels are <1×10 5 and use a 0.99 quantile cutoff. [openCyto, flowClust] For some samples, if <10% of events are retained in hxh (low.viable=true), set the channel limit to 1.5×10 5 , we relax the quantile to 0.9. · Automatic singletGate "singlets" are configured in hxh using FSC-A vs. FSC-W channels and "wider_gate=TRUE" setting. [openCyto, flowStats]; · The top right FSC-A coordinate is forced to be the same as the bottom right one to avoid the case-persistent SEC distorting the gate downwards. [flowWorkspace]; · Automatically set tailGate "VIABLE" for singlets using FVS510-A channel, limiting to min=2, max=3 (min and max in logic scale), tolerance=0.1. [openCyto] If low.viable=true, first filter singlet events with SSC-A<5×10 4 and set smoothing parameter adjust=1. [flowCore] If not, first filter singlet events to FSC-A < 5 × 104 and set smoothing parameter adjust=1.2. [flowCore] In the case of Navios, note that the BV510-A signal is better resolved in Navios than in LSRII, which results in somewhat different processing steps. Set a rectangular gate on NON-DEBRIS events using BV510-A[0,3.5] vs. SSC-W[50,200] “quickLive” to exclude most of the contaminating SEC. ·Automatically set singletGate "SINGLETS" on quickLive using FSC-A vs. FSC-W channels and "wider_gate=FALSE" setting. Force the rightmost FSC-A coordinate to the BSE limit (10 6 ), the maximum FSC-W limit is 2×10 5 to counter instances of persistent SEC distorting the gate downwards. · Set automatic tailGate “VIABLE” for BSE using FVS510-A channel with arguments (side="left", max=3.5, tolerance=0.1). Note that the BV510-A cutoff seen in most samples is reasonable and close to 4, but occasional heuristic "tweaks" are necessary to correct for cases where the signal does not fall into two well-resolved populations. Tweaks are made to the VIABLE tailGate parameter. Note that the isotype sample does not have TCPP resulting in a higher BV510-A signal. · Threshold fine tuning: reason.cut=4, high.cut=5 (patient samples excluding isotype) reason.cut=4.5, high.cut=5.5 (isotype) If (BV510-A># of events where VIABLE cutoff)>1.5x(# of events where BV510-A>reason.cut): The VIABLE cutoff is too low and needs to be moved to the right. Set the tailGate "min" argument to the larger of reason.cut and mean BV510-A expression, and increase by an additional 0.5 if >25% of events >reason.cut are also >high.cut. Ff(number of events where BV510-A > VIABLE cutoff) > 2x (number of events where BV510-A > reason.cut): Set tailGate tolerance=0.2, adjust=0.5. If (BV510-A># of events where VIABLE cutoff)>1.5x(# of events where BV510-A>reason.cut): The VIABLE cutoff is too high and needs to be moved to the left. · Set the tailGate "max" argument to the smaller of the high.cut and the average of BV510-A expression, and increase by an additional 0.5 if >25% of events have BV510-A>high.cut. Set the tailGate "min" argument to 3.5, and increase by an additional 0.5 if >25% of events have BV510-A>high.cut. If (# of events where BV510-A > VIABLE cutoff) is 2.25x < (# of events where BV510-A > reason.cut): There is an overshift to the left, so increase tailGate “min” by another 0.5 and set adjust=2. ·Create a temporary gating set for both BSE and SINGLETS gate events and automatically recalculate tailGate VIABLE using tweaked arguments. - Apply VIABLE gates determined from the dummy gating set to NON-DEBRIS events in the complete gating set. [flowWorkspace] · The automatic singletGate set on VIABLE events, "SINGLETS", gives good results in most cases, but some samples still contain confounding events at this point. SEC contamination can map close to larger viable cells, resulting in bottom left coordinate < 0 or bottom right coordinate < top left coordinate. In either case, set VIABLE to SSC-A < 5 x 10 before automatically setting singletGate with "wider_gate=FALSE". 4 [openCyto,flowStats] In all cases, we make the top right FSC-A coordinate the same as the bottom right FSC-A coordinate to avoid the SEC in which the case persists skewing the gate downwards. [flowWorkspace] · Apply the SINGLETS gate to VIABLE events. [flowWorkspace] When viable cells are relatively scarce, the proportion of events retained in the SINGLETS gate in the gap between FVS510-A>2.5 (logic scale) and the viability cutoff can be substantial. If the "gap" population is >10% of the SINGLETS, the survival cutoff will be FVS501-A<2.5. -Remove "VIABLE" from the complete gating set. [flowWorkspace] For NON-DEBRIS with FVS510-A<2.5, set to "VIABLE" rectangleGate. Automatically recalculate singletGate "SINGLETS" as above. [openCyto,flowStats,flowWorkspace] -Apply the SINGLETS gate to the VIABLE events of the complete gating set. [flowWorkspace] Note that in the case of Navios, the intermediate gates are removed from the patient samples to retain only BSE. ·VIABLE gate is applied to BSE. ·SINGLETS gate is applied to VIABLE. ·CleanSSC gates are applied to SINGLETS. For each patient tube gate, the full matrix of events retained by SINGLETS (LSRII) or cleanSSC (Navios) is written out. These values, along with patient metadata, are the input to the cancer / non-cancer classifier. Note that the retained events are digitized to create model variable values. Due to differences in the output values ​​of the LSRII and Navios detectors, the ranges are not identical. See 5.5 Grid Settings below.

[0063] The analytical process of an LSRII Flow Cytometer with Embedded Navios Details [R package] is demonstrated according to one embodiment of the present invention. Note 1: The process for the Navios EX flow cytometer is essentially the same but adjusted for different file formats (i.e., LMD files for Navios EX instead of FCS files for LSRII) as well as detector sensitivity and dynamic range (LSRII = 18 bit, Navios = 20 bit). 1.1: LSRII utilizes two configuration files. One is to match file names with fluorescent controls for compensation (matchfile.csv). The other is to identify the patient tube (or tubes in some cases) and flow cytometer channel names (samplematch.csv). See 5 below. For an example, Additional Materials "LSRII Configuration Files" (5.1 and 5.2). 1.2: A controlled tube naming convention has been introduced for Navios sample collection rather than using a configuration file. See 5.3 Controlled Tube Naming Protocol for Navios below. 1.3 The names and order of the fluorescent channels in the flow cytometer files are different but can be mapped between LSRII and Navios, see 5.4 Channel Equivalence and Detector Maxima below. Note 2: Gate names are shown in BOLD UPPERCASE. NOTE 3: Temporary gate names are shown in bold lower case. Note 4: Calculated thresholds are in bold italics. Note 5: The main packages used in the current process are shown in [square brackets]. Note 6: Some heuristic adjustments were necessary to handle the wide range of sample compositions and viabilities.

[0064] An embodiment of a sample processing pipeline according to an embodiment of the present invention is provided below: Analysis tube: ·NIST Beads Compensation tubes (one per fluorescence channel) Patient single cell suspension (4 tubes): ·PE-anti-CD45 and TCPP, no staining as control Isotype controls stained with FVS510 (viability), PE-anti-CD45, PE-CF594 isotype, A488 isotype, and FITC isotype "Blood" stained with TCPP, FVS510, PE-anti-CD45, PE-CF594-anti-CD206, A488-anti-CD3, FITC-anti-CD66b, A488-anti-CD19 (Note: A488 and FITC are read in the same channel, giving a combined signal of CD66b / CD19 / CD3) "Epithelium" stained with TCPP, FVS510, PE-anti-CD45, PE-CF594-anti-EpCAM (epithelial cell adhesion molecule), and A488-anti-PanCK (pan-cytokeratin) 5.5 Grid Settings JPEG2025505917000005.jpg124156

[0065] 5. Additional Materials as Examples of Data Inputs to the System and Method 5.1 Sample matchfile.csv configuration file for LSRII File name, channel Specimen_001_BL-18-148 Unstained_011.fcs, unstained Specimen_001_A549 FVS510_007.fcs, BV510-A Specimen_001_PE comp beads_003.fcs, PE-A Specimen_001_PE-TxRed comp beads_005.fcs, PE-Texas Red-A Specimen_001_FITC comp beads_004.fcs, FITC-A Specimen_001_A549 TCPP_008.fcs, APC-A 5.2 Samplematch.csv configuration file for LSRII / / Number of multiple collection tubes separated by Variables, details Blood, s_BL-19-151_Blood_016.fcs / / s_BL-19-151_Blood-tube2_017.fcs Epithelial, s_BL-19-151_Epithelial_018.fcs Isotype, s_BL-19-151_Isotype_015.fcs stub, BL-19-151 fitc_name, FITC-A nist, s_NIST_001.fcs 5.3 Navios Control Tube Naming Protocol 01_NIST 03_FITC 04_PE 05_PECF594(Texas Red) 06_APCpos (must be combined with pre-compensation of tube 13) 07_FVS510 08_Accession#_Unstained (patient sample) 09_Accession#_Isotype (patient sample) 10_Accession#_Blood (patient sample) 11_Accession#_Epithelial (patient sample) 13_APCneg (must be combined with pre-compensation of tube 6)

[0066] 5.4 Channel Equivalence and Detector Maxima examples of different instruments used to transform the data. Illustrated in the grid below. JPEG2025505917000006.jpg94151 *As seen in FCS (LSRII) or LMD (Navios) metadata.

[0067] We identified a reduced list of lineage markers as potential predictors for the most promising and queried pairwise interactions between them for classifiers that yield high sensitivity and selectivity in assays to identify cells with or without cancer potential. 10 It was unexpectedly discovered that adding a negative value proportional to the "number of FSC-A R2 events" improved the performance of the classifier. One interpretation of this interaction term is that it helps to mitigate the possible age-related accumulation of stressed cells in high-risk patient populations as a result of smoking or health history.

[0068] After developing the two stages of the Lung assay, we assembled a complete pipeline including quality control steps, determination of predictor variable values, and sample classification (Figure 7).

[0069] Referring now to FIG. 7, a lung sputum data processing pipeline according to one embodiment of the present invention is illustrated. The schematic depicts the following major elements: Quality control (QC) measurements 701 and 703 are performed on the data and the instrument. The data acquisition files are reviewed for adequacy. Sputum samples with about 10,000 live singlet events and about 10 or more lung macrophages in the sample are determined to be suitable. A search of subject-specific data, such as the information in Table 1, is obtained and combined with sputum sample characteristics obtained from the FCM-acquired population 705 from the FCM. A classifier 706 is generated based on coefficients b , which are determined by fitting the classifier model of 706 to the particular set of samples used to build the classifier. 0 ~b 5 The classifier 706 is applied to a new sample to determine a cancer likelihood value based on the cutoff value 707. The classifier model is applied to the population obtained by FCM from the sputum sample and the subject-specific data input into the classifier model to determine whether the sample is likely to be cancer 708 or not (lower diamond) 709. The coefficients b derived from fitting the classifier 706 to the particular set of samples used to build the classifier are 0 ~b 5 is as follows: b 0 :53.515414 b 1 0.701153 b 2 :0.001545 b 3 :58.071495 b 4 :4.258709 b 5 :0.784982 One skilled in the art will appreciate that when the classifier 706 is fitted to a different set of sample values, the values ​​of the coefficients will change, but the model itself will not change.

[0070] In one embodiment of the present invention, data acquisition validity 701 and sample acceptability assessment 703 begin by ensuring that each collection tube's data file is readable and its encoded data matrix is ​​complete. Time signatures are used to interrogate the fluorescence channels of each tube to remove flow anomalies resulting from air bubbles or clogging during sample acquisition. A fluorescence compensation tube is then used to derive a compensation matrix de novo (as opposed to using a compensation matrix encoded in the sample file metadata). Fluorescence signals are compensated and transformed to logicle scale to generate a sample data matrix that is used by automated FCM gating to isolate viable singlet events. To provide confidence in downstream numerical analysis, samples containing a threshold number of singlets, e.g., at least 10,000 viable singlets, were analyzed when the number of events in some analysis windows was potentially low. The threshold was set at a threshold value of pulmonary macrophages (CD206) to confirm that the sputum sample originated from the lung. 中および高 CD3 / CD19 / CD66b 低~中 ) was set to have at least 10 cells in the shaded area of ​​Figure 6E.

[0071] A further step in the assay pipeline of Figure 7 is to feed years and flow-based values ​​from the survival singlets to a classifier model. In one embodiment of the present invention, a classifier for determining the likelihood of having cancer is determined using four variables: i) age, ii) TCPP / log in region 3 (Figure 6A, R3). 10 Number of events per 10,000 live singlets (per 10K) for SSC-A, iii) FVS510-A / log in region 2 (Figure 6B, R2) 10 Events per 10K for FSC-A, and iv) CD20 6 Low CD3 / CD19CD66b 中Number of events per 10K in the sector (Figure 6C, shaded box). In addition, the classifier includes a negative term for the interaction between age and the FVS510 density variable, as well as an "intercept" term (b 0 ) and the coefficient (b 1 , b 2 , b 3 , b 4 , and b 5 The value of ) depends on the training set used for model fitting and gives weight to the variables. We did not normalize the data given to the model to make it easier to interpret the model. For example, the model formula teaches us that increasing the number of events with high TCPP density increases the likelihood of cancer, consistent with previous results.

[0072] The decision step in the pipeline in Figure 7 is to make a cancer / non-cancer assignment. The model returns a value in the range [0,1]. Whether a given sample is classified as cancer depends on the model returning a value greater than a given cutoff. If the value is less than or equal to the cutoff, the sample is classified as non-cancer. A reasonable cutoff can be chosen by running a cutoff value between 0 and 1 through the steps and measuring the true and false positive calls at each step by comparing it to the known group categories.

[0073] The development of a GLM classifier according to one embodiment of the present invention is described below in further iterations of Figure 11. The input to the process is illustrated by the output of the FCM step (see, for example, 1101-1102 in Figure 11). Evaluate combinations of potential predictors (1103-1105 in FIG. 11). 1) Combinations of potential predictors were evaluated (see, e.g., 1103 in FIG. 11 ) to develop classifiers using generalized linear models (GLM) for semi-supervised machine learning as follows: 1.1 Clinical parameters available for all samples 1.2. Quantified measurements from patient blood and epithelial samples (light and fluorescence), individually and in paired combinations, using equally or heuristically placed cuts (3x3 and 4x4 grids per examined channel) 1.3. Subtract background counts from quantified measurements as in number 2 above 1.4. Quantified Fluorescence / log10(Light Scattering) 1.5. Frequencies of specific subpopulations identified by prior manual analysis 2) Parameter inclusion of iterative forward and backward steps in the GLM using separate training and testing sets (roughly 2 / 3 training, 1 / 3 testing, randomly selected and non-overlapping (see, e.g., 1104-1105 in FIG. 11 ); 3) Evaluate the GLM based on the Akaike Information Criterion (AIC), which estimates prediction error and provides a comparator of model quality; (see, e.g., 1105 in FIG. 11 ); 4) repeatedly holding parameters in different iterations of model testing, with different combinations of potential predictor variables tested individually and in combination to detect potential interactions (see, e.g., 1104-1105 in Figure 11 ); 5) Controlling overfitting of the final model by repeated random sampling (n=10) of training / test sets to validate prediction robustness independent of the specific samples used to train the model; (see, e.g., 1106 in FIG. 11 ); and 6) Validate the processing pipeline and classifier predictions on samples not used in model building and testing (see, e.g., 1107-1110 in FIG. 11). Once a classifier is developed, it is applied to new, previously unanalyzed samples whose cancer / non-cancer classification is unknown (see 1101 and 1111 in FIG. 11).

[0074] Referring now to Figure 8A, the results of this process are shown as a receiver operating characteristic (ROC) curve with an AUC of 0.89. The assay achieved its best performance in discriminating cancer from non-cancer at a threshold of 0.28 (Figure 8B, solid vertical line).

[0075] Referring now to FIG. 8A, Receiver Operating Characteristic (ROC) curves are illustrated showing the false positive rate vs. true positive rate calculated as lines marked with asterisks for varying model response thresholds. For comparison, the ROC curve of a previous version of the slide-based lung assay is shown as an inverted triangular curve. The dashed lines indicate 80% and 90% sensitivity and specificity (1-false positive rate). FIG. 8B illustrates a model based on the ROC curve of FIG. 8A. The model value (likelihood of cancer) threshold was 0.28 (solid vertical line), which corresponds to a sensitivity of 82.1% and 87.7%. The dashed lines indicate the level at which cancer prediction is very unlikely (leftmost dashed line) or very likely (rightmost dashed line). Filled circles represent cancer and open circles represent high risk. CyPath Lung performance for LSRII samples is shown.

[0076] In one embodiment of the present invention, automated analysis of samples by flow cytometry combined with machine learning resulted in a predictive model with sensitivity (82%) and specificity (88%) that was robust to differences in sample handling and disease stage. Importantly, the test was 92% sensitive and 87% specific in difficult to manage cases with no nodes or nodules 20 mm in diameter or less.

[0077] Referring now to Figures 9A-9B, correlation graphs illustrate age on the x-axis versus modeled values ​​on the y-axis when samples were analyzed on an LSRII flow cytometer or a NaviosEX cytometer, respectively.

[0078] Referring now to FIG. 10, an exemplary system and method 1000 for receiving input from a flow cytometer 1002 based on a sample 1001 analyzed by the flow cytometer and subject metadata information 1020 is illustrated. A computer-implemented device 1003 includes a processor 1010 (e.g., processing circuitry) operable to execute program instructions or software, causing the computer to perform various methods or tasks, such as performing techniques for generating and / or using analytical models of classifier models described herein. The processor 1010 is connected via a bus (not shown) to a memory 1030, which is used to store information such as program instructions and / or other data during operation of the computer. A storage device 1040, such as a hard disk drive non-volatile memory, or other non-transitory storage device, stores information such as program instructions, data files of multidimensional data and reduced data sets, and other information. The computer also includes various input / output elements 1050, including parallel or serial ports, USB, Firewire or IEEE 1394, Ethernet, and other such ports for connecting the computer to external devices such as printers, video cameras, display devices, medical imaging devices, monitoring equipment, etc. Other input / output elements include wireless communication interfaces such as Bluetooth, Wi-Fi, and cellular data networks. The computer itself may be a traditional personal computer, a rack-mounted or business computer or server, or any other type of computerized system. The computer may, in further examples, include less than all of the elements listed above, such as a thin client or mobile device having only some of the elements shown. In another example, the computer is distributed among multiple computer systems, such as a distributed server having many computers working together to provide various functions.

[0079] Referring now to FIG. 11, an exemplary process flow chart 1100 is illustrated for use in semi-supervised machine learning of a GLM classifier for cancer / non-cancer status and subsequent classification of biological samples such as sputum, according to an embodiment of the present invention. A sputum sample processed for processing in a flow cytometer is provided to the flow cytometer for analysis 1100 as outlined herein. Flow cytometer data 1101 from multiple sputum samples is processed with an automated FCM to extract a data matrix of viable singlet populations from each sample 1102. Numeric values ​​of potential predictive variables associated with the collected flow cytometry data matrix and associated clinical data are calculated for all samples 1103. The multiple samples are randomized into a training set and a test set 1104. A GLM classifier is constructed using the values ​​of the potential predictors from the training set along with the known cancer / non-cancer status of the training set. The resulting classifier is evaluated on the test set for its ability to correctly classify the test set into cancer / non-cancer 1105. Step 1105 is repeated with different combinations of potential predictor variables, retaining only the truly predictor variables from each iteration. The GLM classifier fits only the retained predictor variables to generate a final model 1106. A single sputum sample of unknown status to be classified is processed with an automated FCM 1101 (single) to obtain the final predictor variable values ​​from sputum and subject metadata (e.g., age) 1111, which are input into the final model to generate a classification value 1107. Based on the classification value cutoff determined by the ROC analysis, an output classification call is generated 1108 for the sputum sample of cancer 1109 or non-cancer 1110.

[0080] According to one aspect of the invention, the system and method accurately classify study participants as cancer or non-cancer with high accuracy, including participants with nodules smaller than 20 mm at different disease stages, thus the test has the potential to improve the process of early lung cancer diagnosis.

[0081] material and method Clinical Trials and Collection Sites The minimal-risk study was registered with ClinicalTrials.gov, reviewed and approved by the Sterling Institutional Review Board (Atlanta, GA), and conducted in accordance with the ethical principles of the Declaration of Helsinki (v 1996) and Good Clinical Practice guidelines. Sample collection was performed at five study centers: Atlantic Health System, NJ; Mt. Sinai Hospital, NY; Radiology Associates of Albuquerque, NM; South Texas Veterans Healthcare System; and Waterbury Pulmonary Associates, CT. Each site had institutional approval to participate in the study. Each potential participant was presented with an informed consent form, and only those who signed it were enrolled.

[0082] Participant Information: Participants (male and female) were eligible for enrollment in one of two groups. The non-cancer group included participants (ages 50–80 years) who were either current smokers with at least a 20 pack-year smoking history who had quit smoking within the past 15 years, or current non-smokers with at least a 30 pack-year smoking history. The exceptions were two patients who had quit smoking 25 and 26 years prior. Most participants in the non-cancer group had LDCT results or other forms of imaging that were not suspicious for cancer and were invited to return for LDCT screening in 12 months. In a few cases, participants initially placed in the non-cancer group underwent follow-up LDCT, PET / CT or biopsy. These participants were followed until their health status was confirmed. If participants were diagnosed with lung cancer, they were switched to the cancer group.

[0083] Each participant in the cancer group had been assessed by a physician as likely to have lung cancer, based on medical history and LDCT or other imaging results. The diagnosis was confirmed by biopsy after a sputum sample was provided. The exception was a patient who developed a new nodule measuring 24 mm and was too fragile to undergo a biopsy. If the biopsy did not show cancer, the participant was switched to the non-cancer group. There were no age or smoking history restrictions for enrollment in the cancer group.

[0084] For each participant, we collected the following demographic data: sex (male or female); age (years); ethnicity (Hispanic / non-Hispanic Latino / Latino); and race (American Indian / Alaska Native; Asian; Black / African American; Native Hawaiian / Other Pacific Islander; White; Other). Data on smoking history was collected, as well as data on comorbidities (asthma, COPD, emphysema, chronic bronchitis) and previous cancer history. All participants were required to provide contact information for their primary care physician and consent to releasing medical information if requested. Exclusion criteria included the presence of severe obstructive pulmonary disease and the inability to cough with sufficient exertion to generate a sputum sample, angina with minimal exertion, and pregnancy.

[0085] Exclusion criteria included the presence of severe obstructive pulmonary disease and inability to cough with sufficient exertion to produce a sputum sample, angina with minimal exertion, and pregnancy.

[0086] Sputum samples: Sample donors were trained on how to use the Acapella Assist Device (Smiths Medical, St. Paul, MN) and repeated this procedure at home on three consecutive days, storing the specimen cups in a cool, dark place or in the refrigerator. Within 1 day after collection was completed, samples were transported overnight to the bioAffinity laboratory, where further processing and FCM analysis were performed by people blinded to the origin of the samples.

[0087] Sputum processing: Sputum was isolated and labeled. For example, sputum samples were incubated with a mixture of 0.1% dithiothreitol and 0.5% N-acetyl-L-cysteine ​​at room temperature for 15 min and neutralized with Hank's balanced salt solution. Cells were then filtered through a 100 micron nylon strainer, washed, and resuspended in HBSS. Total cell yield was determined using the trypan blue exclusion method. Sputum was liquefied using pre-warmed 0.1% dithiothreitol (DTT) at a ratio of 1:4 (w / v) to the weight of sputum and pre-warmed 0.5% N-acetyl-L-cysteine ​​(NAC) at a ratio of 1:1 (w / v). The resulting cell suspension was filtered through a 100 μm nylon cell strainer (Falcon, Corning Inc.) to remove larger debris while minimizing cell loss. Cells were collected in 50 mL conical tubes, washed, and centrifuged at 800 × g for 10 min. The separated sputum pellets were combined into 15 mL conical tubes for each sputum sample. Total cell yield and viability were determined with a Neubauer hemocytometer using the trypan blue exclusion method.

[0088] In one embodiment, a small aliquot of cells was reserved for use as a control, and the majority was split into two tubes for the main analysis. For example, both tubes were labeled with Fixable Viability Stain 510 (FVS510) and CD45-PE. One tube, the so-called "blood tube", contained CD66b-FITC, CD3-Alexa-Fluor-488, CD19-Alexa-Fluor-488 and CD206-PE-CF594. In the other "epithelial tube", cells were labeled with pan-cytokeratin-Alexa-Fluor-488 and EpCAM-PE-CF594. Cells were incubated on ice for 35 minutes. After washing with HBSS, cells were fixed and kept on ice until the next day, when TCPP solution (20 μg / mL) was added (3.3×106 cells / ml; 1:1 v / v). After incubation on ice for 1 hour, cells were washed twice with cold HBSS and kept on ice until analysis.

[0089] In another embodiment, the sample is divided into at least two tubes: white blood cells (CD45+ ) one tube containing a marker to interrogate the cellular compartment and the other containing the epithelial (CD45 - Cell labeling was performed by splitting the sputum into one tube for the cellular compartment, and the other for the cellular compartment. Each tube contained anti-CD45 antibody, FVS510 (to exclude dead cells, including SECs), and porphyrin TCPP (to identify cancer (associated) cells). To identify the leukocyte population, anti-CD206 antibody was added to label macrophages, as well as a cocktail of antibodies to label granulocytes (anti-CD66b) and lymphocytes (anti-CD3 and anti-CD19). For epithelial cell recognition, we used anti-cytokeratin (panCK) and anti-EpCAM. No permeabilization step for cytokeratin labeling was performed, since the initial DTT and NAC treatment for sputum processing was sufficient for intracellular cytokeratin staining.

[0090] Isolated sputum cells were incubated with antibodies and FVS510 for 35 min. After washing once with cold HBSS, cells were fixed with paraformaldehyde on ice for 1 h, after which cells were washed again and stored on ice until TCPP labeling the next day. TCPP was added to cells for 1 h. After incubation, cells were washed twice with cold HBSS and then stored on ice until flow cytometry analysis. Cells were kept on ice and protected from light throughout the labeling procedure until analysis. For reagent details, see Table 7.

[0091] Flow cytometry: Sputum samples were acquired on a BD LSR II flow cytometer (BD Biosciences) equipped with four lasers (404 nm, 488 nm, 561 nm, 633 nm) or a Navios EX (Beckman Coulter Life Sciences) equipped with three lasers (405 nm, 488 nm, and 638 nm). Post-collection data analysis was performed using FlowJo software (Tree Star, Inc. Ashland, OR).

[0092] Sample characteristics: Of the 171 patient LSRII samples collected, 150 were suitable for analysis by the full assay pipeline, consisting of 122 high-risk patients without cancer and 28 patients with lung cancer (Table 1). An additional 4 samples that did not have the final disease status were included in the pipeline development stage, as the addition of unlabeled samples had been shown to be useful for model building. In addition, 14 samples that were flagged as ineligible based on counts (see below in the implementation of the lung assay pipeline) were also used in the model fitting stage to better capture the distribution of the underlying data and help make the model generalization more robust to sample noise. Due to issues during acquisition, only 3 samples could be used.

[0093] Table 1. Patient characteristics of LSRII Lung assay validation samples [Table 1]

[0094] Traditionally, the presence of "large numbers" of macrophages in a sputum smear indicates a sample of pulmonary origin. A quality control measure using the cell surface antigen CD206, specific for macrophage populations present in lung tissue and not found in the blood circulation, was utilized in the FCM lung assay according to one embodiment of the present invention. In one embodiment, sputum cells were stained with i) CD45-specific cell markers, e.g., antibodies against CD45 (to identify white blood cells), ii) CD206-specific cell markers, iii) CD66b-specific cell markers, iv) CD3-specific cell markers, and v) CD19-specific cell markers. In one embodiment, any combination of i)-v) can be combined and added to a sputum sample, e.g., a cocktail of antibodies composed of anti-CD66b, anti-CD3, and anti-CD19 compounds, e.g., where the compounds are antibodies or fragments thereof, to further separate macrophages from other hematopoietic cells. In one embodiment, FVS510 was used as a viability dye to exclude dead cells. A portion of viable sputum cells specifically expresses CD45, as evidenced by anti-CD45PE signal. Cytospins of sorted CD45+ sputum cells confirmed their hematopoietic origin.

[0095] Further analysis of sputum samples analyzed by FCM reveals that sputum-derived leukocytes contain distinct subpopulations of macrophages. Cells selected by size-exclusion and live cell gates as well as doublet discrimination gates illustrate the representative light scatter profile of unstained single live sputum cells, defining both CD45+ and CD45- gates for sorting and further analysis. Cells that fell within the gate of live, single sputum cells and were stained with a blood panel of antibodies were further analyzed.

[0096] Sputum-derived leukocyte profiles of FVS510-CD45+ cells from different samples stained with a blood panel of antibodies were further analyzed. Gates were used to identify lymphocytes / granulocytes (gate 1) as well as alveolar macrophages (gate 2) and interstitial macrophages (gate 3) based on lineage-specific markers (fluorescence) and / or optical properties of specific cell types (FFS / SSC).

[0097] In this embodiment, fluorescence minus one (FMO) controls were analyzed utilizing the same gates for leukocyte subpopulations defined by the blood panel of antibodies. All FMO controls included the viability dyes, CD45 and TCPP.

[0098] Sputum cells stained with a leukocyte antibody panel excluding CD66b, CD3, and CD19 antibodies were analyzed. Sputum cells stained with a leukocyte antibody panel excluding CD206 antibodies were also analyzed. Wright-Giemsa stained cytospins from sorted CD45+ gate 2 and gate 3 populations were reviewed under a microscope for visual identification. Scale bars indicate 20 μm for gate 2 cell population and 10 μm for gate 3 cell population.

[0099] Cell type was confirmed by a pathologist. Cell size measurements of sorted macrophage populations in gate 2 and gate 3. At least 100 cells were measured for each population. The mean cell size in gate 2 is 16um + / - standard deviation approximately 3-5um (****p<0.0001).

[0100] FCM profiles were acquired for CD45+ sputum cells labeled with anti-CD206 antibody and a cocktail of anti-CD66b, anti-CD3 and anti-CD19 antibodies. The isotype control shows higher background staining than unstained or fluorescence minus one (FMO) controls. Since the use of isotype control antibodies comes with its own set of problems, we used FMO controls to identify the main subpopulations of sputum. By comparing the FMO control of the CD66b / CD3 / CD19 cocktail to the stained sample containing all antibodies, gate 1 can be set to identify a combination of lymphocytes and granulocytes. Similarly, by comparing the FMO control for the CD206 antibody, two populations of CD206 positive cells can be identified.

[0101] After sorting cells from gates 2 and 3, cytological analysis revealed cell populations with morphology consistent with that of macrophages. However, cells sorted from gate 3 were significantly smaller in size compared to cells sorted from gate 2. Our calculated sizes for the alveolar and interstitial macrophage populations are consistent with previously reported size ranges.

[0102] Alveolar macrophages were identified as strongly positive for CD206 and autofluorescence in FITC channel gate 2, while interstitial lung macrophages were smaller in size and lower in CD206 expression gate 3.

[0103] The mean background staining in the CD206 FMO control was 0.0023% (+ / -SD 0.0021%) in both gates combined. The positivity threshold based on 2 standard deviations (SD) above the mean background staining is set at 0.0065%, or approximately 6 macrophages per 100,000 cells for both gates combined. Due to concerns that a lower threshold would not fall within the linear detection range of the PE-CF594 fluorochrome, we instead chose an arbitrary threshold of 0.05%, which includes alveolar and interstitial macrophages. This threshold could not be based solely on interstitial macrophages. The 0.05% threshold was well within the linear range of detection of the flow cytometer and met the "large number of macrophages" criteria for suitable samples set by the Papanicolaou Society.

[0104] 179 samples were analyzed for macrophage content. Based on the criteria outlined above, 15 samples were found to have inadequate macrophage numbers. However, six of these samples (3.4%) had less than 1000 CD45+ events for analysis, which is too few cells for adequate analysis based on the limits set according to one embodiment of the present invention. Five of the six samples had less than 1.5×10 6The remaining nine samples (5.0%) had more than 10,000 CD45+ cells (range 11648–463382), all with a mean of 1.7 × 10 6 sputum cells. Additionally, only four of 164 eligible samples showed fewer than 1.5 × 106 cells at the start of the antibody staining process. Although all of these samples contained robust macrophage numbers, three of the four showed fewer than 10,000 CD45+ cells (range 1327–2908). This data supports the conclusion that the 1.5 × 10 6 This suggests that sputum samples with fewer than 10 cells are too small for reliable diagnostic flow cytometry analysis.

[0105] A calculation of the total number of sputum cells (excluding SECs) for each individual sample before antibody labeling was calculated. All suitable samples (n=164) revealed >0.05% macrophages (alveolar and interstitial combined). Of the suitable samples, 18 had cell counts greater than 50 million cells. The median cell count for suitable samples was 14.6 x 10 6 Unsuitable samples (n=15) showed no alveolar macrophages or a combined event in the alveolar and interstitial macrophage gate of <0.05%. The median cell count in unsuitable samples was 6.9×10 6 A subset of unsuitable samples contained "too few cells" for a reliable profile (<1000 CD45+ events), while the remainder contained enough cells but did not meet the QC macrophage criteria to consider them suitable samples. The median cell number for the too few cells was 1.1 × 10 6 Each cell is a single cell.

[0106] One hundred and sixty-four suitable sputum samples were further analyzed for differences between cancer and high risk. This set included 32 samples obtained from individuals diagnosed with lung cancer and 132 samples from high-risk individuals without cancer. The cancer group included 40.6% current smokers and 44.7% high-risk. There was no significant difference in pack-years smoked between the groups. There was also no significant difference in the average number of years former smokers had myositis. The proportion of women in the cancer group was smaller than in the high-risk group (21.9% vs. 54.5%, respectively). The mean age of participants in the cancer group was 69.8 years, compared to 64.8 years in the high-risk group (p<0.0002).

[0107] In the first phase of the analysis, we examined the percentage of CD45+ cells versus CD45- cells and various subpopulations in each compartment without the TCPP marker. We found that the percentage of CD45+ cells in sputum samples from cancer patients was significantly higher than in sputum from high-risk patients without the disease (49.64% vs. 38.95%; p=0.0099). Different subpopulations of the CD45+ compartment were discernible in all samples, but the relative contribution of each population differed between samples and groups. By comparing the relative size of each CD45+ subpopulation between cancer and high-risk samples, we found that cancer samples contained significantly more granulocytes / lymphocytes (gate 1 p=0.0378) and interstitial macrophages (gate 3 p=0.0031).

[0108] The CD45- compartment contained cells of epithelial origin, which was confirmed by the presence of goblet cells and ciliated epithelial cells when the CD45- cells were sorted and their morphology visualized by cytospin. By using antibodies against EpCAM and cytokeratin, we were able to further delineate the CD45- cells by flow cytometry. FMO controls show that the background of each of the antibodies used is low. The relative contribution of the various CD45 subpopulations varied from sample to sample, and no significant differences were observed between the cancer and high-risk groups.

[0109] Live single CD45- sputum cells from different samples were stained with epithelial cell markers, e.g. PanCK and EpCAM (epithelial profile) / epithelial antibody panel. Fluorescence minus one (FMO) controls of the profile were obtained by FCM. FMO controls were obtained including viability dye, CD45, and TCPP. Sputum-derived epithelial profiles of FVS510-CD45 cells stained with isotype controls for the antibodies used were obtained (unstained sputum cells). FMO control FVS510-CD45- cells stained with EpCAM but without panCK antibody profile were obtained. FMO control FVS510-CD45- cells stained with panCK without EpCAM antibody were obtained.

[0110] The second phase of FCM analysis observed TCPP fluorescence. Single live cells were separated into three cell subsets based on TCPP staining intensity: TCPP-HIGH, TCPP-INTERMEDIATE (IM) and TCPP-LOW cells. The relative ratios of these cell subsets did not differ between the high-risk and cancer groups. Each of these three subsets was then further interrogated for their content of CD45+ leukocyte populations and CD45- epithelial cell populations. TCPP-HIGH cells, TCPP-IM cells and TCPP-LOW cells were further analyzed. TCPP-HIGH cells show a broad light scattering profile on the FSC vs. SSC dot plot 0x / y axis.

[0111] The CD45+ compartment of TCPP-HIGH cells is enriched for alveolar macrophages (CD45+;CD206++ cells), whereas the CD45- compartment is enriched for EpCAM+;panCK+ double positive cells. TCPP-IM cells represent the majority of sputum cells, and thus the profile of this subpopulation resembles that of the whole sample. TCPP-LOW cells exhibit relatively low light scattering properties compared to TCPP-HIGH cells and are predominantly CD45- with no expression of the epithelial markers EpCAM or panCK.

[0112] Based on gating of different cell lineage markers by FCM, sputum cell populations with different TCPP fluorescence intensities were identified. A dot plot analysis displaying TCPP versus FITC / Alexa488 fluorescence (i.e., CD66b / CD3 / CD19 in the leukocyte tube and panCK in the non-leukocyte tube) was used to define the TCPP-HIGH cutoff. A dot plot of TCPP versus PE-CF594 fluorescence can also be used for this purpose, although cells with the highest FI for TCPP are easier to identify with the former.

[0113] The TCPP-HIGH cutoff is derived from a gate placed on a dot plot of sputum cells with TCPP on the y-axis and CD66b / CD3 / CD19 on the x-axis. The TCPP-LOW population is defined as the intersection of unstained sputum with the TCPP-stained sample. The population with intermediate TCPP staining, TCPP-IM, is defined as the population between the TCPP-HIGH and TCPP-LOW populations.

[0114] The inherent characteristics of the TCPP-HIGH population showed some significant differences between the high-risk and cancer groups. First, TCPP-HIGH cells from samples of the cancer group showed lower side scatter values ​​than those from the high-risk group. Second, the CD45- compartment of the TCPP-HIGH population contained a higher percentage of EpCAM+panCK+ cells. Furthermore, this double-positive population from samples of the cancer group expressed higher levels of EpCAM, but not panCK, compared to cells in the same quadrant belonging to samples of the high-risk group.

[0115] Differences in sputum cell characteristics between cancer and high-risk sputum samples were identified with the TCPP-HIGH population of cancer samples exhibiting smaller SSCs than the TCPP-HIGH population of high-risk samples (**p<0.01). In cancer samples, the percentage of EpCAM+panCK+ cells in the CD45- fraction of the TCPP-HIGH subpopulation is greater than the corresponding CD45- fraction of high-risk samples (**p<0.01). The mean fluorescence intensity (MFI) of EpCAM in TCPP-HIGH CD45-EpCAM+panCK+ cells is higher in cancer samples than the corresponding cell subset of high-risk samples (*p<0.05).

[0116] Further analysis of the cancer groups revealed significantly higher EpCAM A higher mean fluorescence intensity was observed in early cancer samples (stage I / II) compared to late cancer samples (stage III / IV) (p=0.047). No significant differences based on cancer type (squamous cell carcinoma vs. adenocarcinoma) or smoking history (current vs. former smokers) were identified. Interestingly, when we separated high-risk smokers based on smoking history, the profile of current high-risk smokers showed the presence of significantly more TCPP-HIGH;EpCAM+;panCK+ cells (p=0.0008) and macrophage populations compared to former smokers in both the alveoli (p=<0.0001) and interstitium (p=0.0141).

[0117] Current methodologies used for sputum analysis pose challenges that have limited their clinical use. Sputum cytology has low sensitivity due to the high skill required to identify subtle nuclear changes. The need to screen a large number of slides is time consuming and also hampers its clinical use. Although imaging and molecular techniques can assess genetic alterations in sputum-derived cells, screening methods based on nuclear ploidy or in situ hybridization to detect genetic abnormalities use only a few hundred cells per sputum sample, and microchip analysis of enriched epithelial cells derived from sputum analyzed for genetic abnormalities screens only 2000 cells per slide. Exclusion of the majority of sputum cells from the analysis may mask important disease parameters and make the analysis less sensitive than would be clinically useful. The limitations of these various techniques should not be confused with the highly useful nature of sputum as a bodily fluid, which can provide an important cellular snapshot of the lung environment.

[0118] Flow cytometry platforms are well suited to analyze exfoliated cells isolated from sputum to identify tumor-associated changes in leukocyte and non-leukocyte populations that would otherwise go undetected by traditional cytological methods. FCM's ability to detect and analyze cells based on physical properties (i.e., size and granularity) and cell surface molecules is powerful. Unlike microscopy or cytology, flow cytometry can analyze large numbers of cells in a short period of time. The variability in autofluorescence and nonspecific binding properties of cell populations within and between sputum samples prohibits the use of commercially available biological controls that are often used for immunophenotyping of highly characterized hematopoietic populations. For this reason, internal FMO controls have been used to establish the positivity threshold for the macrophage gate. The ability to identify alveolar macrophages as a distinct leukocyte subpopulation allowed us to include built-in flow cytometry quality control parameters to determine the pulmonary origin of each sputum sample. To ensure quality control according to one embodiment of the present invention, a cytology-based sample quality confirmation was required.

[0119] The lungs are constantly exposed to pathogens and harmful particles. Alveolar macrophages are the main primary innate defense to maintain a healthy lung environment. Alveolar macrophages are characterized as a distinct CD45+ population with high CD206 expression (CD206++) and moderate to high signals on the granulocyte / lymphocyte axis due to their autofluorescence. Furthermore, the results show that The light scattering profile of alveolar macrophages Previous observations that overlapped with those of contaminating SEC were confirmed, highlighting the need to isolate SEC from further analysis.

[0120] Although CD206 intermediate positive cells (CD206+) are also macrophages, they are smaller than alveolar CD206++ macrophages and show minimal FITC autofluorescence, indicating that this macrophage population likely represents interstitial macrophages. Although interstitial macrophages (in contrast to alveolar macrophages) are not normally in contact with the airway lumen, the proinflammatory environment caused by chronic smoking is ideal for the penetration of interstitial macrophages into the airways. Thus, their presence in sputum obtained from heavy smokers is not unexpected. This is further demonstrated by our findings that current high-risk smokers have significantly more macrophages in their sputum than former high-risk smokers.

[0121] In one embodiment of the present invention, the minimum number of sputum-derived cells in a sputum sample for automated FCM to give a suitable profile so that the presence of macrophages can be determined was approximately 1.5 million cells. A cutoff of 5 macrophages per 10,000 cells (0.05%) to determine sample adequacy was well within the detection range of the flow cytometer. Interstitial macrophages were included in the 0.05% macrophage cutoff for sample adequacy because the presence of both alveolar and interstitial macrophages are lung tissue-specific cell populations. Samples without alveolar macrophages were considered insufficient because the presence of interstitial macrophages (a rare occurrence) without the presence of alveolar macrophages is difficult to interpret biologically.

[0122] Comparative multiparameter analysis of sputum samples from individuals with confirmed lung cancer versus those at high risk of developing the disease revealed significant differences between the two groups: cancer samples contained significantly more CD45+ cells, specifically more granulocytes / lymphocytes and interstitial macrophages, than did high-risk samples.

[0123] By adding the porphyrin TCPP to our staining protocol, we were able to identify several significant differences in the most brightly stained subset (TCPP-HIGH) between the cancer and high-risk groups. TCPP-HIGH cells from the cancer group, regardless of their CD45 lineage, showed lower side scatter characteristics than TCPP-HIGH cells from the high-risk group, suggesting the reduced cytoplasmic content, organelle degranulation, and vacuolization documented for malignant tumors.

[0124] Analysis of the non-leukocyte (CD45-) subpopulation of TCPP-HIGH cells revealed that the cancer group contained a greater percentage of cells stained for the epithelial markers panCK and EpCAM. This difference with the high-risk group is mainly caused by the fact that former high-risk smokers had significantly fewer of these cells in their sputum compared to current smokers. The epithelial cell subpopulation from the cancer group expressed equal levels of panCK but also higher levels of EpCAM. This was most pronounced in the stage I / II subgroup.

[0125] Detection of epithelial-derived cancers and circulating tumor cells has historically relied on detection of both EpCAM and cytokeratin expression. Our flow cytometry-based analysis identifying increased expression of EpCAM in samples with confirmed stage I / II cancers as well as in samples from high-risk participants who continue to smoke suggests that EpCAM expression may be particularly important in early lung cancer detection.

[0126] According to one embodiment of the Lung assay, analysis of light scattering and fluorescence signals from viable single cells identified by automated FCM is determined. Logistic regression models the relationship between predictor variables and a categorical (in our case, binary cancer / non-cancer) response variable. Stepwise regression is a supervised machine learning process in which potential predictor variables are added and removed and the resulting model is examined for goodness of fit. Clinical factors for which complete data were available (Table 1) were included as potential predictors. Age was the clinical parameter that was repeatedly assessed as significant during forward and backward stepwise regression.

[0127] The performance of the Lung assay was evaluated for 122 high-risk and 28 cancer samples listed in Table 1, as well as an additional 32 samples (Table 2) processed on a different FCM instrument (Navios EX). These 32 samples comprise a different set of patients than that used for assay development using the LSRII cytometer. The same model with the same coefficients was used for both instruments, but the cutoff for Navios samples was 0.5 instead of 0.28. The results shown in Table 3 demonstrate that the Lung assay performed very well with sensitivity, specificity and accuracy all >80% for the LSRII samples, and with a very similar number for the smaller set of Navios EX samples. Very robust negative predictive values ​​(NPV) > 95% were obtained for both platforms. Table 2. Patient characteristics of Navios EX validation samples [Table 2]

[0128] Table 3. Lung assay performance [Table 3]

[0129] The Lung assay also performed remarkably well for cases in which nodules ≥20 mm in diameter were detected by LDCT (Table 3, "All nodules <20 mm"), with a sensitivity of 92% and a specificity of 87%, with an area under the ROC curve of 94%. Furthermore, the Lung assay performed well in all tumor types represented and in all disease stages, including I and II (Table 4, Table 5).

[0130] Table 4. Performance of the Lung assay by tumor type and stage (LSRII) [Table 4]

[0131] Table 5. Performance of the Lung assay by tumor type and stage (Navios EX) [Table 5]

[0132] Each of the retained predictors contributed significantly to the model (Wald test p-value < 0.05), and their individual removal negatively affected the ability to accurately classify cancer and high-risk samples (Table 6). Age is a well-established clinical correlate of lung cancer, as it is in our model. 31 Nevertheless, the correlation between age and model values ​​was not overwhelming in either the LSRII or Navios EX samples (Figure 8), with some younger patients being called "cancer" and many older patients being called "non-cancer." Indeed, CD206 低 CD3 / CD19CD66b 中 Signal elimination is achieved by elimination of age and FVS510-A / log 10 Its interaction with FSC-A R2 resulted in the same number of misclassified samples (Table 6).

[0133] Table 6. Impact of model predictors on classification [Table 6]

[0134] Table 7. Reagents used for staining and flow cytometry analysis of sputum according to one embodiment of the present invention [Table 7] JPEG2025505917000014.jpg27161

[0135] One aspect of the present invention provides an automated flow cytometry system and method for machine learning analysis to predict the presence of lung cancer from sputum samples. One hypothesis, without being limited thereto, is that sputum as diagnostic material provides a snapshot of the tumor itself, its microenvironment (ME) and its field of carcinogenesis (FoC). Although specialized cytological analysis of sputum can detect cancerous and precancerous cells, this is a very laborious approach that is not well suited for non-automated large-scale screening, prone to observer bias, and does not have the ability to scrutinize a large number of cells from a sample in a few seconds, as cytological samples are limited to the size of the slide that limits the number of cells analyzed. Automated image processing has been used with some success to capture malignancy-associated changes in cells, but is still hampered by technical complexity and the low number of cells analyzed.

[0136] Another aspect of the present invention provides a system and method for analyzing biological samples such as sputum for the presence of cancer cells by a high-throughput automated flow cytometry-based approach combined with machine learning to provide the following advantages: a) the assay can be subjected to routine laboratory use without the need for expert evaluation of the sample or being subject to operator bias, b) the entire sputum sample can be rapidly analyzed, c) the numerical analysis can capture the complex interactions between lung cancer, ME and FoC cells that would be difficult for an individual to detect reliably. During the development of the Lung assay, for example, it was unexpectedly found that the predictive value of viability staining density suggests an association with apoptosis. Furthermore, it was unexpectedly observed that specific markers of immune function would be beneficial.

[0137] One aspect of the present invention provides an automated flow cytometry-based test that queries three aspects of tumorigenesis: TCPP staining, programmed cell death and immune response. Other researchers have shown that the performance of sputum-based tests for early lung cancer detection can be significantly increased when combining different types of measurements, e.g., cytology and methylation biomarkers with gene mutations or microRNAs. Although we used the same technology platform to measure different cancer-related processes, these additional parameters are likely to contribute to the performance improvement from slide-based to flow cytometry-based assays (Figure 7). Furthermore, flow cytometry-based assays read the entire sample, which was also predicted to increase test performance.

[0138] All study participants except one met the criteria for lung cancer screening most recently issued by the US Preventive Services Task Force. Although our study group can be considered a sample from those eligible for lung cancer screening (one of the target populations of Lung), sampling was small and minorities were underrepresented as were women in the cancer group. Furthermore, the prevalence of cancer in our study was less than 19% for both data sets, which is considerably higher than the lung cancer screening population or the group of patients with lung nodules between 7 and 19 mm (the other target populations of the Lung assay).

[0139] The American Thoracic Society, in its 2017 official policy statement, states that molecular biomarkers should influence clinical management decisions to improve clinical outcomes in order to be considered clinically useful. The authors discuss a use case in which screening is expanded to include participants currently ineligible for LDCT screening, lowering the cancer prevalence from the NLST level of 1 / 120 to a hypothetical 1 / 500. They assume a reasonable harm threshold of 0.83%, based on the NLST data, which would result in a minimum positive diagnostic likelihood ratio (PDLR) of 4.18, a level met by the larger LSRII group (Table 3). Using a hypothetical prevalence of 1 / 400 rather than 1 / 500 with the same harm threshold would result in a PDLR of 3.35, met by both the LSRII and Navios groups. The Lung assay system and method disclosed herein may be useful in expanding early lung cancer screening to relatively underserved populations, such as young female and male African American smokers.

[0140] Lung assays, perhaps in conjunction with risk calculators (e.g., Brock University's Lung Cancer Risk Calculator), can also aid clinical decision-making in LDCT-positive patients with intermediate-sized nodules. Although the Pan-Canadian study found that the largest nodules were nonmalignant in 20% of participants, those smaller than 7 mm only 2% of NLST patients underwent invasive follow-up, and those larger than 20 mm may err on the side of caution and prompt immediate follow-up. However, intermediate-sized nodules are notoriously challenging. If we estimate the risk threshold (R) above which invasive follow-up is worthwhile to be the frequency of cancer in the NLST population with nodules 7–19 mm in diameter (4.8%) and assume a cancer prevalence of 3.8% in the LDCT-positive population, then the sensitivity / (1–specificity) should be ≥ [(1–prevalence) / prevalence] × R / (1–R) = [(1–0.038) / 0.038] × 0.048 / (1–0.48) = 1.28, a threshold that is comfortably met by our assay. 44 (Table 3, PDLR).

[0141] One aspect of the Lung assay according to one embodiment of the present invention is a non-invasive sputum-based test for the detection of early stage lung cancer. It uses a flow cytometry platform to analyze the cellular content of sputum, and the analysis is fully automated and therefore unbiased. Test performance in cases with small nodules (<20mm) showed a sensitivity of 92% and a specificity of 87%.

[0142] Where a range of values ​​is provided, it is understood that each intervening value, to the tenth of the unit of the lower limit, between the upper and lower limits of that range, unless the context clearly dictates otherwise, and any other stated or intervening value within that stated range, is encompassed within the invention. The upper and lower limits of these smaller ranges may independently be included in the smaller ranges and are encompassed within the invention, subject to any specifically excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the invention.

[0143] Certain ranges are presented herein with numerical values ​​preceded by the term "about". The term "about" is used herein to provide literal support for the exact number it precedes, as well as a number that is close or approximately close to the number it precedes. In determining whether a number is close or approximately a specifically recited number, the near or approximate unrecited number may be a number that provides a substantial equivalent to the specifically recited number in the context in which it is presented.

[0144] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar or equivalent to those described herein can also be used in the practice or testing of the present invention, representative exemplary methods and materials are described herein.

[0145] It should be noted that, as used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. It should be further noted that the claims may be drafted to exclude any optional element. Thus, this statement is intended to serve as a predicate for use of exclusive terminology, such as "solely," "only," and the like, in connection with the recitation of claim elements or the use of a "negative" limitation.

[0146] As will be apparent to those skilled in the art upon reading this disclosure, each of the individual embodiments described and illustrated herein has individual components and features which may be readily separated or combined with the features of any of the other several embodiments without departing from the scope or spirit of the invention. Any recited method can be carried out in the order of events recited or in any other order which is logically possible.

[0147] Although the systems and methods have been described for grammatical fluidity with functional descriptions, it is to be clearly understood that the claims should not be construed as necessarily limited by "means" or "step" limitation constructions unless expressly formulated under 35 U.S.C. § 112, but rather should be given the full scope of the meaning and equivalents of the definitions provided by the claims under the legal theory of equivalents, and that if the claims are expressly formulated under 35 U.S.C. § 112, the full legal equivalents under 35 U.S.C. § 112 should be given. How to Classify Flow Cytometer Data

[0148] As discussed above, aspects of the present invention include methods for classifying flow cytometer data. "Flow cytometer data" refers to information about the characteristics of sample particles (e.g., beads or cells or debris) collected by any number of detectors in a particle analyzer. As discussed herein, a "particle analyzer" is an analytical tool (e.g., a flow cytometer) that allows for the characterization of particles based on certain (e.g., optical) parameters. "Particle" refers to individual components of a biological sample, such as molecules, analyte-bound beads, individual cells, etc.

[0149] The method of interest involves classifying one or more population clusters based on determined parameters (e.g., fluorescence) of events (e.g., particles) in a sample. As used herein, a "population" or "subpopulation" of events, such as cells or other particles, generally refers to a group of events that have characteristics (e.g., optical properties, impedance properties, or time properties) related to one or more measured parameters, such that the measured parameter data form a cluster in the data space. Data obtained from analysis of cells (or other particles) by flow cytometry is often multidimensional, with each cell corresponding to a point in the multidimensional space defined by the measured parameters. In embodiments, the data is composed of signals from multiple different parameters, including, for example, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, etc., and more. Thus, populations are recognized as clusters in the data. Conversely, each data cluster is generally interpreted as corresponding to a population of a particular type of cell or particle, although clusters corresponding to noise or background are also commonly observed. Clusters may be defined within a subset of dimensions, for example with respect to a subset of measurement parameters (e.g., fluorescent dyes) that correspond to populations that differ only in a subset of the measurement parameters or features extracted from the sample measurements.

[0150] Aspects of the subject method include receiving a first gate having a defined boundary. As discussed herein, a "gate" generally refers to a classifier boundary that identifies a subset of data of interest (the data representing features or characteristics of particles / cells in a sample). In cytometry, a gate can bind a group (i.e., a population) of events of particular interest. In other words, a gate defines a boundary for classifying a population of flow cytometer data. In embodiments, a gate identifies flow cytometer data exhibiting the same or similar set of parameters. Furthermore, "gating" generally refers to the process of classifying data using a gate defined for a given data set. For example, a cell or particle is an event detected by a sensor when the cell or particle passes between the sensor and an interrogating light source of the flow cytometer, and the optical features or characteristics of the event detected by the detector / sensor can be analyzed for each event.

[0151] In some embodiments, the first gate is a gate drawn by a trained algorithm. In such embodiments, the trained algorithm may define the boundaries of a region (e.g., in two-dimensional space) within which a particular classification can be assigned to the flow cytometer data. For example, drawing the first gate may include overlaying a polygon on a two-dimensional plot representing the flow cytometer data. For example, the first gate may be received from a database of gates used in previous attempts to classify the flow cytometer data.

[0152] In an embodiment, the method includes receiving flow cytometer data, calculating parameters for each population, and gating the populations for further analysis based on the target population of interest. For example, an experiment may include particles / cells labeled with several fluorophores or fluorescently labeled antibodies, and groups of particles may be defined by populations corresponding to one or more fluorescence measurements. In this example, a first group may be defined by a certain range of light scattering for a first fluorophore, a second group may be defined by a certain range of light scattering for a selected population from the first group, and a third group may be defined by a third fluorophore based on one or more selected populations of the first group, the second group, or a combination thereof.

[0153] The flow cytometer data may be received from any suitable source. In some embodiments, the flow cytometer data is received from a memory of a storage device. In such embodiments, the flow cytometer data may have been previously generated and stored in the memory of a storage device for subsequent recall and analysis. In other embodiments, the flow cytometer data is received in real-time. Stated differently, the flow cytometer data generated during operation of the flow cytometer may subsequently (e.g., immediately) populate a data space (e.g., a two-dimensional plot) having a first gate. In some cases, the flow cytometer may be operated to generate data until a recording criterion is met. A "recording criterion" as discussed herein is a condition that terminates operation of the flow cytometer and data collection. Any suitable recording criterion may be used. In one particular case, the recording criterion is a time limit. When the recording criterion is a time limit, the flow cytometer data collection stops after a defined amount of time (e.g., ranging from a few seconds to three hours). In additional cases, the recording criterion is a total number of events. In such cases, the flow cytometer data collection stops after a certain number of particles (e.g., defined by the user) have been analyzed. In yet additional cases, the recording criterion is the number of events in a population. Flow cytometer data collection may, in such cases, be stopped after a certain number of particles (e.g., defined by a user) in a particular population (e.g., exhibiting a particular phenotype) have been analyzed.

[0154] In certain embodiments, particles are detected and uniquely identified by exposing them to excitation light and measuring the fluorescence of each particle in one or more detection channels as desired.The fluorescence emitted in the detection channels used to identify particles and their associated binding complexes can be measured after excitation by a single light source, or can be measured separately after excitation by separate light sources.When separate excitation light sources are used to excite particle labels, the labels can be selected so that all labels are excitable by each of the excitation light sources used.

[0155] In an embodiment, the flow cytometer data is received from a forward scatter detector. The intended forward scatter detector provides information regarding the overall size of the particle. In an embodiment, the flow cytometer data is received from a side scatter detector. The intended side scatter detector detects refracted and reflected light from the surface and internal structure of the particle, which tends to increase as the particle's complexity of structure increases (e.g., particle size). In an embodiment, the flow cytometer data is received from a fluorescence detector. The intended fluorescence detector is configured to detect fluorescent emission from fluorescent molecules associated with the particles in the flow cell, such as labeled specific binding members (such as labeled antibodies that specifically bind to markers of interest). In certain embodiments, the method includes detecting fluorescence from the sample with one or more fluorescence detectors, including 2 or more, 3 or more, 4 or more, 5 or more, 6 or more, 7 or more, 8 or more, 9 or more, 10 or more, 15 or more, etc., and 25 or more fluorescence detectors.

[0156] The method in certain embodiments also includes data acquisition, analysis, and recording, such as by a computer, where multiple data channels record data from each detector for light scattering and fluorescence emitted by each particle as it passes through the sample interrogation area of ​​the flow cytometer. In these embodiments, the analysis includes classifying and counting the particles so that each particle is present as a set of digitized parameter values. The target system can be set to trigger on selected parameters to distinguish particles of interest from background and noise or cell populations that are not of interest. "Trigger" refers to a pre-set threshold for detecting a parameter and can be used as a means for detecting the passage of a particle through a light source. Detection of an event exceeding the threshold of the selected parameter triggers the acquisition of light scattering and fluorescence data of the particle. No data is acquired for particles or other components in the medium being assayed that cause a response below the threshold. The trigger parameter can be detection of forward scattered light caused by a particle passing through a light beam. The flow cytometer then detects and collects the light scattering and fluorescence data of the particle. The data recorded for each particle is analyzed in real time or stored in a data storage and analysis means, such as a computer, as necessary.

[0157] In at least one embodiment, as will be readily understood by those skilled in the art, the apparatus according to the present invention comprises a general purpose or special purpose computer or distributed system programmed with computer software to perform the above-mentioned steps, which may be any suitable computer language, including R, Python, C++, C#, Perl, Java, PHP, HTML, MySQL, distributed programming languages, etc. The apparatus may also comprise a plurality of such computers / distributed systems in various hardware implementations (e.g., connected via the Internet and / or one or more intranets). For example, data processing may be performed by a suitably programmed microprocessor, computing cloud, application specific integrated circuit (ASIC), field programmable gate array (FPGA), etc., in conjunction with suitable memory, network, and bus elements. Alternatively, a container may be used.

[0158] The embodiments of the present invention provide a technology-based solution that overcomes the existing problems of the current state of the art in a technological way to solve existing problems for people who may have early stage cancer, healthcare providers, insurance companies and diagnostic laboratories. One embodiment of the present invention is necessarily based on computer technology such as computer learning. An embodiment of the present invention achieves significant advantages over the current state of the art such as increased flexibility, faster results, non-invasive procedures, and automated screening of samples. For example, using flow cytometry and automated analysis of data in a time frame of minutes to hours, thousands of cells from a biological specimen can be analyzed and characterized at a speed and accuracy that is not possible with analysis by a human observer to obtain the same data in the same amount of time.

[0159] The foregoing examples can be repeated with similar success by substituting the generally or specifically described reactants and / or operating conditions of the present invention for those used in the foregoing examples.

[0160] In the specification and claims, "about" or "approximately" means within twenty percent (20%) of the cited numerical value. All computer software disclosed herein may be embodied on any computer readable medium (including combinations of media), including, but not limited to, CD-ROM, DVD-ROM, hard drive (local or network storage device), USB key, other removable drive, ROM, virtual machine, software container (e.g., Docker), and firmware.

[0161] Although the invention has been described in detail with particular reference to these embodiments, other embodiments can achieve the same results. Variations and modifications of the present invention will be obvious to those skilled in the art, and it is intended to cover in the appended claims all such modifications and equivalents.

[0162] All publications and patents cited herein are incorporated by reference as if each individual publication or patent was specifically and individually indicated to be incorporated by reference, and are incorporated by reference herein to disclose and describe the methods and / or materials for which the publications are cited. The citation of any publication is for its disclosure prior to the filing date and should not be construed as an admission that the present invention is not entitled to antedate such publication by prior invention. Further, the publication dates provided may be different from the actual publication dates which may need to be independently confirmed.

Claims

1. 1. A flow cytometer method for automatically analyzing a sputum sample from a subject suspected of having lung cancer using a computer, comprising: obtaining a plurality of cells from a sputum sample from the subject suspected of having lung cancer; marking said plurality of cells with i) a plurality of lineage-specific marker compositions, ii) a cell viability composition, and iii) a tetra(4-carboxyphenyl)porphyrin (TCPP) composition; analyzing the plurality of cells marked in i-iii with the flow cytometer to obtain a cell size selected subpopulation from the plurality of cells based on an automatically selected bead size exclusion gate; automatically selecting, by the computer, a viable singlet population of cells from the cell size selected subpopulations using an automatic non-debris gate and an automatic singlet gate; automatically obtaining, by the computer, from the viable singlet population of cells, a flow cytometer value based on the plurality of lineage-specific marker compositions, the cell viability composition, and the TCPP marker; automatically applying, by the computer, a trained classifier to metadata and obtained flow cytometry values ​​from the subject, wherein the subject metadata includes age; and automatically generating, by the computer, a classification of the sputum sample based on application of the trained classifier, wherein the classification is selected from a plurality of classification options including cancer and non-cancer, the cell viability composition is FVS510, the plurality of lineage specific marker compositions are selected from fluorescent anti-CD206, fluorescent anti-CD3, fluorescent anti-CD19, and fluorescent anti-CD66b, and the trained classifier is [0010] (wherein coefficient b ０ ~b ５ is determined by applying the trained classifier to multiple sputum samples used to construct the classifier.

2. The method of claim 1 , wherein the sputum sample is a single cell suspension.

3. 2. The method of claim 1, wherein the multiple cell lineage specific marker composition is selected from CD206, CD3, CD19, CD66b, CD45, EpCAM, PanCK, and any combination thereof.

4. The method of claim 1 , wherein the cell viability composition preferentially labels dead cells over live cells.

5. 2. The method of claim 1, wherein the automatically analyzing step comprises obtaining flow cytometry values ​​for side scatter, forward scatter, fluorescence from TCPP, fluorescence from the cell viability composition, and fluorescence from the plurality of lineage-specific marker compositions from the plurality of cells.

6. 2. The method of claim 1, wherein the bead size exclusion gate is set between 5 μm and about 30 μm, and events less than about 5 μm and greater than about 30 μm are not further analyzed.

7. The method of claim 1 , wherein the automated non-debris gate excludes most dead cells from the non-debris population.

8. 2. The method of claim 1, wherein the automated singlet gate is applied to a population of cells selected in the automated non-debris gate.

9. 2. The method of claim 1, wherein the sputum sample comprises a minimum number of CD206 expressing cells for the sputum sample to be accepted for a determination of pulmonary health.

10. 1. A system for automated analysis of flow cytometry data, comprising: a computer processor in communication with a memory storing flow cytometry data from a plurality of markers in a plurality of cells from a sputum sample of a subject, the plurality of markers including: i) a plurality of lineage specific marker compositions; ii) a cell viability composition; and iii) a tetra(4-carboxyphenyl)porphyrin (TCPP) composition; A computer program product embodied in a non-transitory computer-readable medium, the computer program product comprising: receiving said flow cytometry data obtained from a plurality of cells from a sputum sample; selecting from the plurality of cells in the sputum sample a subpopulation of cells that is automatically selected based on application of an automated gate selected from a bead size exclusion gate, a viability gate, and a singlet gate; determining flow cytometry values ​​of interest for the plurality of lineage-specific marker compositions, the cell viability composition, and the TCPP marker from the subpopulation; automatically applying a classifier to said flow cytometry values ​​of interest and said subject metadata; and a computer program product comprising instructions for automatically generating an output on a display device with an identification of one or more classifications of said sputum sample, said classification comprising cancer or non-cancer; the cell viability composition is FVS510, the plurality of lineage specific marker compositions are selected from fluorescent anti-CD206, fluorescent anti-CD3, fluorescent anti-CD19, and fluorescent anti-CD66b, and the trained classifier is [0025] (wherein coefficient b ０ ~b ５ is determined by applying the trained classifier to a plurality of sputum samples used to build the classifier.

11. The system of claim 10 , wherein the cell viability composition preferentially labels dead cells over live cells.

12. 11. The system of claim 10, wherein flow cytometry values ​​of the plurality of cells are obtained for side scatter, forward scatter, fluorescence from TCPP, fluorescence from the cell viability composition, and fluorescence from the plurality of lineage-specific marker compositions.

13. 11. The system of claim 10, wherein the bead size exclusion gate is set to exclude events having a size less than about 5 μm and greater than about 30 μm.

14. The system of claim 10 , wherein the automated non-debris gate excludes most dead cells from the non-debris population.

15. A non-transitory computer-readable medium that, when executed, causes a processing circuit to: obtaining flow cytometer values ​​from a viable singlet population of the subject's sputum sample based on side scatter, forward scatter, fluorescence from the TCPP, fluorescence from the cell viability composition, and fluorescence from the plurality of lineage-specific marker compositions; applying the trained classifier to metadata from a subject and the resulting flow cytometry values, where the subject metadata is age; and and program code for causing the generation of a classification of the sputum sample based on application of the trained classifier, wherein the classification is selected from a plurality of classification options including cancer and non-cancer, the cell viability composition is FVS510, the plurality of lineage specific marker compositions are selected from fluorescent anti-CD206, fluorescent anti-CD3, fluorescent anti-CD19, and fluorescent anti-CD66b, and the trained classifier is [0030] (wherein coefficient b ０ ~b ５ is determined by applying the trained classifier to a plurality of sputum samples used to construct the classifier.

16. 16. The non-transitory computer readable medium of claim 15, wherein the cell viability composition preferentially labels dead cells over live cells.

17. The non-transitory computer readable medium of claim 15, wherein a minimum number of CD206 positive cells are present in the sputum sample that is analyzed.

18. 16. The non-transitory computer readable medium of claim 15, wherein the bead size exclusion gate is set at 5 μm to 30 μm.

19. 16. The non-transitory computer readable medium of claim 15, wherein the automated non-debris gate excludes a majority of dead cells from the non-debris population.

Images