Systems for analyzing multimodal flow cytometry data sets and methods of use thereof

Abstract

The present disclosure includes a system for classifying sample cells in a flow stream, for example, according to a biological parameter determined from a feature vector. The system includes a light source configured to illuminate sample cells in a flow stream, a light detection system having a photodetector to detect light from illuminated cells, and a processor having a memory operably coupled to the processor, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to receive two or more data sets, wherein each of the two or more data sets is associated with one of a plurality of data patterns, apply an algorithm to convert each of the two or more data sets to a feature vector based on the associated data pattern, receive a classification task with respect to the two or more data sets, and apply one or more of a plurality of classification models to the feature vector to determine a class based on the received classification task.

Description

Technical Field

[0001] This application relates to the technical field of systems for analyzing multimodal flow cytometry datasets and methods of using them. Background Technology

[0002] Flow cytometry particle sorting systems, such as sorting flow cytometers, are used to sort particles in a fluid sample based on at least one measurement characteristic of the particles. In a flow cytometry particle sorting system, particles (such as analyte-bound beads or single cells in a fluid suspension) flow through a detection zone, where sensors detect particles of the type to be sorted contained in the flow. When the sensor detects particles of the type to be sorted, a sorting mechanism is triggered, selectively separating the particles of interest.

[0003] Particle sensing is typically performed by passing a flow through a detection zone where particles are exposed to illumination from one or more lasers, and the fluorescence from the particles is measured. Particles or their components can be tagged with fluorescent dyes for easy detection, and multiple different particles or components can be detected simultaneously by using fluorescent dyes with different spectra to tag different particles or components. Detection is performed using one or more photoelectric sensors to independently measure the fluorescence of each different fluorescent dye.

[0004] Using data generated from detected light, the distribution of components can be recorded, and the desired material can be sorted. To sort particles in a sample, a droplet charging mechanism charges droplets containing the types of particles to be sorted in the flow at the breakpoint of the flow. The droplets pass through an electrostatic field and are deflected into one or more collection containers based on the polarity and amplitude of the charge on the droplets. Uncharged droplets are not deflected by the electrostatic field.

[0005] Extracting biological insights from flow cytometry data is challenging due to the heterogeneity of data types, the large volume of data, and the lack of universally accessible algorithms for biologists. Summary of the Invention

[0006] The inventors recognized the need for high-throughput phenotypic cell analysis using high-parameter, large-volume heterogeneous data provided by label-free and fluorescence imaging. They also discovered that scalable architectures on neural networks can be used for image-enabled cell phenotypic analysis, combining imaging with spectral analysis.

[0007] This disclosure includes a system for classifying cells (e.g., single cells) in a sample in a flowing stream. A system according to some embodiments includes: a light source configured to illuminate sample cells in the flowing stream; a light detection system having a photodetector to detect light from the illuminated cells; and a processor having a memory operatively coupled to the processor, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to receive two or more datasets, each of the two or more datasets being associated with one of a plurality of data patterns; apply an algorithm to convert each of the two or more datasets into a feature vector based on the associated data patterns; receive a classification task regarding the two or more datasets; and, based on the received classification task, apply one or more of the plurality of classification patterns to the feature vectors to determine a category.

[0008] In some embodiments, the multiple data patterns include image data patterns. In some instances, the algorithm is a rule-based image processing algorithm. In some instances, the algorithm includes a neural network. In some instances, the memory includes instructions for inverting the optical loss channels of a dataset associated with the image data pattern, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to zero mean, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to unit standard deviation, or a combination thereof. In some instances, the multiple data patterns include waveform data patterns. In some instances, the algorithm includes a sequence learning machine model. In some instances, the multiple data patterns include spectral data patterns. In some instances, the algorithm includes a compensation model, a spectral unmixing model, or a combination thereof. In some instances, the multiple data patterns include scattering data patterns. In some instances, the algorithm includes a calibration model.

[0009] In some embodiments, the classification task includes single-cell classification, survival classification, whole blood classification, T-cell activation classification, or any combination thereof. In some instances, the classification task is a single-cell classification task, and the categories include single-cell categories, doublets categories, separated doublets categories, triplets categories, or fragment categories. In some instances, the classification task includes a survival classification task, and the categories include a survival category, a death category, or an apoptosis category. In some instances, the classification task includes a whole blood classification task, and the whole blood classification includes a granulocyte category, a monocyte category, or a lymphocyte category. In some instances, the classification task includes a T-cell activation classification task, and the categories include an activated category or a non-activated category.

[0010] In some embodiments, the multiple classification models include end-to-end models. In some instances, the end-to-end models include uniform manifold approximation and projection (UMAP) models, T-distributed random nearest neighbor embedding (TSNE) models, principal component analysis (PCA) models, or any combination thereof. In some instances, the multiple classification models are trained using supervised methods. In some instances, the multiple classification models are trained using a weighted cross-entropy loss function.

[0011] In some embodiments, the light source is configured to illuminate the sample with a frequency-modulated light beam. In some instances, the light source includes one or more lasers. In some instances, the light detection system includes multiple photodetectors. In some embodiments, one or more photodetectors are photomultiplier tubes. In some embodiments, one or more photodetectors are photodiodes (e.g., avalanche photodiodes, APDs). In some embodiments, the light detection system includes a photodetector array, such as a photodetector array having multiple photodiodes or charge-coupled devices (CCDs).

[0012] In some instances, the system includes a sorting mechanism for distributing sample cells into multiple sample containers. In some instances, the sorting mechanism is configured to sort cells based on their presence, classification, or both. In some instances, the memory includes instructions for sorting cells based on generated cell images and spectral data from the cells. In some instances, the memory includes instructions for sorting cells based on computed image parameters.

[0013] This disclosure also includes methods for classifying sample cells (e.g., single cells) in a flowing stream, such as classification based on biological parameters of the cells determined from feature vectors. A method according to some embodiments includes: receiving two or more datasets, each of the two or more datasets being associated with one of a plurality of data patterns; applying an algorithm to convert each of the two or more datasets into a feature vector based on the associated data patterns; receiving a classification task for the two or more datasets; and applying one or more of a plurality of classification models to the feature vectors based on the received classification task to determine a category. In some instances, the method further includes illuminating a sample containing cells in the flowing stream with a light source and measuring the light from the illuminated cells using a light detection system with a photodetector.

[0014] In some embodiments, the multiple data patterns include image data patterns. In some instances, the algorithm is a rule-based image processing algorithm. In some instances, the algorithm includes a neural network. In some instances, the memory includes instructions for reversing the optical loss channels of a dataset associated with the image data patterns, scaling the pixel values ​​of each channel of the dataset associated with the image data patterns to zero mean, and scaling the pixel values ​​of each channel of the dataset associated with the image data patterns to unit standard deviation or a combination thereof. In some instances, the multiple data patterns include waveform data patterns. In some instances, the algorithm includes a sequence learning machine model. In some instances, the multiple data patterns include spectral data patterns. In some instances, the algorithm includes a compensation model, a spectral unmixing model, or a combination thereof. In some instances, the multiple data patterns include scattering data patterns. In some instances, the algorithm includes a calibration model.

[0015] In some embodiments, the classification task includes a single-cell classification task, a survival classification task, a whole blood classification task, a T-cell activation classification task, or any combination thereof. In some instances, the classification task is a single-cell classification task, and the categories include single-cell categories, adherent two-cell categories, separated two-cell categories, three-cell categories, or fragment categories. In some instances, the classification task includes a survival classification task, and the categories include a survival category, a death category, or an apoptosis category. In some instances, the classification task includes a whole blood classification task, and the whole blood classification includes a granulocyte category, a monocyte category, or a lymphocyte category. In some instances, the classification task includes a T-cell activation classification task, and the categories include an activated category or a non-activated category.

[0016] In some embodiments, the multiple classification models include end-to-end models. In some instances, the end-to-end models include uniform manifold approximation and projection (UMAP) models, t-distributed random nearest neighbor embedding (TSNE) models, principal component analysis (PCA) models, or any combination thereof. In some instances, the multiple classification models are trained using supervised methods. In some instances, the multiple classification models are trained using a weighted cross-entropy loss function.

[0017] In some instances, the method includes identifying one or more sorting gates for the classified cells of a sample. In some instances, the method includes computing one or more sorting gates that capture clusters of target cells and exclude clusters of non-target cells. In some instances, the method includes computing a sorting gate that maximizes the inclusion rate of target cell clusters. In some instances, the method includes computing a sorting gate that maximizes the exclusion of clusters of non-target cells.

[0018] A non-transitory computer-readable storage medium with instructions for an algorithm is also provided. According to certain embodiments, the non-transitory computer-readable storage medium includes: an algorithm for receiving two or more datasets, wherein each of the two or more datasets is associated with one of a plurality of data patterns; an algorithm for applying the algorithm to convert each of the two or more datasets into a feature vector based on the associated data patterns; an algorithm for receiving a classification task about the two or more datasets; and an algorithm for applying one or more of a plurality of classification models to the feature vectors based on the received classification task to determine a category. In some instances, the non-transitory computer-readable storage medium includes an algorithm for illuminating a sample containing cells in a flowing stream with a light source, and an algorithm for measuring the light from the illuminating cells using a light detection system with a photodetector.

[0019] In some embodiments, the multiple data patterns include image data patterns. In some instances, the algorithm is a rule-based image processing algorithm. In some instances, the algorithm includes a neural network. In some instances, the memory includes instructions for inverting the optical loss channels of a dataset associated with the image data pattern, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to zero mean, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to unit standard deviation, or a combination thereof. In some instances, the multiple data patterns include waveform data patterns. In some instances, the algorithm includes a sequence learning machine model. In some instances, the multiple data patterns include spectral data patterns. In some instances, the algorithm includes a compensation model, a spectral unmixing model, or a combination thereof. In some instances, the multiple data patterns include scattering data patterns. In some instances, the algorithm includes a calibration model.

[0020] In some embodiments, the classification task includes a single-cell classification task, a survival classification task, a whole blood classification task, a T-cell activation classification task, or any combination thereof. In some instances, the classification task is a single-cell classification task, and the categories include single-cell categories, adherent two-cell categories, separated two-cell categories, three-cell categories, or fragment categories. In some instances, the classification task includes a survival classification task, and the categories include a survival category, a death category, or an apoptosis category. In some instances, the classification task includes a whole blood classification task, and the whole blood classification includes a granulocyte category, a monocyte category, or a lymphocyte category. In some instances, the classification task includes a T-cell activation classification task, and the categories include an activated category or a non-activated category.

[0021] In some embodiments, the multiple classification models include end-to-end models. In some instances, the end-to-end models include uniform manifold approximation and projection (UMAP) models, T-distributed random nearest neighbor embedding (TSNE) models, principal component analysis (PCA) models, or any combination thereof. In some instances, the multiple classification models are trained using supervised methods. In some instances, the multiple classification models are trained using a weighted cross-entropy loss function.

[0022] In some instances, the non-transitory computer-readable storage medium includes algorithms for determining one or more sorting gates for classifying cells in a sample. In some instances, the non-transitory computer-readable storage medium includes algorithms for computing one or more sorting gates that capture clusters of target cells and exclude clusters of non-target cells. In some instances, the non-transitory computer-readable storage medium includes algorithms for computing sorting gates that maximize the inclusion rate of clusters of target cells. In some instances, the non-transitory computer-readable storage medium includes algorithms for computing sorting gates that maximize the exclusion of clusters of non-target cells.

[0023] In some instances, the non-transitory computer-readable storage medium includes algorithms for sorting cells in a sample into multiple sample containers. In some instances, the non-transitory computer-readable storage medium includes algorithms for sorting cells based on the presence, classification, or both. In some instances, the non-transitory computer-readable storage medium includes algorithms for sorting cells based on generated cell images. In some instances, the non-transitory computer-readable storage medium includes algorithms. Attached Figure Description

[0024] The invention can be best understood by reading the following detailed description in conjunction with the accompanying drawings. The drawings include the following figures:

[0025] Figure 1A A flowchart illustrating the classification of sample cells based on feature vectors according to certain embodiments is depicted. Figure 1B An analytical workflow that integrates image, spectral, and scattering information from measured single cells, according to certain embodiments, is described. Figure 1C An example of a labelless image encoder according to certain embodiments is shown. Figure 1D Images of light loss for different cell classifications according to certain embodiments are depicted. Figure 1E Example performance of a classifier in whole blood classification according to certain embodiments is described. Figure 1F Example performance of the classifier in T cell activation determination according to certain embodiments is described. Figure 1G The receiver operating characteristic curves of a T-cell activation dataset according to certain embodiments are shown.

[0026] Figure 2 A flow cytometer system according to certain embodiments is shown.

[0027] Figure 3A An image-enabled particle sorter according to certain embodiments is depicted. Figure 3B Image-enabled particle sorting data processing according to certain embodiments is described.

[0028] Figure 4A A functional block diagram of a particle analysis system according to certain embodiments is depicted. Figure 4B A flow cytometer according to certain embodiments is described.

[0029] Figure 5 A functional block diagram of an example control system according to certain embodiments is depicted.

[0030] Figure 6A A schematic diagram of a particle sorting system according to certain embodiments is depicted. Figure 6B A schematic diagram of a particle sorting system according to certain embodiments is depicted.

[0031] Figure 7 A block diagram of a computing system according to certain embodiments is depicted. Detailed Implementation

[0032] This disclosure includes systems for classifying cells (e.g., single cells) in a flowing stream of samples, such as classifying based on biological parameters determined by feature vectors. A system according to some embodiments includes: a light source configured to illuminate sample cells in a flowing stream; a light detection system having a photodetector for detecting light from the illuminated cells; and a processor having memory operatively coupled to the processor, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to receive two or more datasets, each of the two or more datasets being associated with one of a plurality of data patterns; apply an algorithm to convert each of the two or more datasets into a feature vector based on the associated data patterns; receive a classification task regarding the two or more datasets; and apply one or more of a plurality of classification models to the feature vectors based on the received classification task to determine a category. Methods using the system of this subject matter are also described. A non-transitory computer-readable storage medium is also provided.

[0033] Before describing the invention in more detail, it should be understood that the invention is not limited to the specific embodiments described, as these embodiments can certainly be varied. It should also be understood that, since the scope of the invention will be limited only by the appended claims, the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0034] Where numerical ranges are provided, it should be understood that, unless the context explicitly specifies otherwise, every value between the upper and lower limits of that range (accurate to one-tenth of the lower limit unit), as well as any other specified value or intermediate value within that range, is included in this invention. The upper and lower limits of these smaller ranges may be independently included within those smaller ranges and also within this invention, subject to any specific exclusions within the range. Where the range includes one or both of the included limits, the range excluding one or both of those included limits is also included in this invention.

[0035] This article provides certain ranges whose numerical values ​​are preceded by the term "approximately". The term "approximately" is used in this article to provide textual support for the exact number preceding it and for numbers that are close to or approximate to the number preceding the term. In determining whether a number is close to or approximate to a specifically listed number, an unlisted number that is close to or approximates can be a number that provides a basic equivalent to the specifically listed number in the context in which it appears.

[0036] 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 pertains. Although any methods and materials similar to or equivalent to those described herein may be used in the practice or testing of this invention, representative illustrative methods and materials are described hereafter.

[0037] All publications and patents referenced in this specification are incorporated herein by reference as if each individual publication or patent were expressly and individually indicated to be incorporated herein by reference, and by reference to disclose and describe the methods and / or materials relating to the referenced publication. References to any publication refer to its publication prior to the filing date and should not be construed as an admission that the present invention is not entitled to priority over that publication by virtue of a prior invention. Furthermore, the publication dates provided may differ from the actual publication dates, which may require independent verification.

[0038] It is worth noting that, unless the context clearly specifies otherwise, the singular forms “a,” “an,” “the,” etc., used herein and in the appended claims include plural references. It should also be noted that claims can be drafted to exclude any optional elements. Therefore, this statement is intended to serve as a preliminary basis for the use of exclusive terms such as “unique,” ​​“only,” etc., when referencing claim elements or using the “negative” limitation.

[0039] As will be apparent to those skilled in the art upon reading this disclosure, each individual embodiment described and illustrated herein has discrete components and features that can be readily separated from or combined with features of any of the other several embodiments without departing from the scope or spirit of the invention. Any enumerated method may be performed in the order of the enumerated events or in any other logically feasible order.

[0040] Although the apparatus and method have been or will be described for grammatical fluency and functional interpretation, it should be clearly understood that, unless expressly provided in 35 U.SC §112, the claims should never be construed as necessarily limited to “means” or “steps,” but rather should be given the full scope of meaning and equivalence provided by the definition of the claims in accordance with the doctrine of judicial equivalence, and if the claims are expressly provided in 35 U.SC §112, they should be given the full scope of their legal equivalence in accordance with 35 U.SC §112. A system for classifying single cells in a sample in a flowing stream.

[0041] As described above, aspects of this disclosure include systems for classifying cells (e.g., single cells) in samples from a flowing stream, such as based on biological parameters determined from feature vectors. In some embodiments, the system provides a scalable neural network-based architecture for automating image-based cell phenotypic analysis and integrating images with spectral analysis. The ability of the methods and systems described herein to receive image data, waveform data, spectral data, and scattering data enables the inclusion of data in various forms from different sources. In some instances, the system and methods provide a workflow for flow cytometry analysis that integrates multimodal datasets, such as image data, waveform data, spectral data, and scattering data measured from single cells, and is used to implement an image encoder. The application of neural networks described herein provides biological insights into single cells in a flowing stream from large, heterogeneous datasets, including imaging datasets, allowing general algorithms to generate insightful biological data. In some instances, the biological data is sufficient to determine whether further downstream biological sampling or assays are needed.

[0042] In some instances, the system is configured to generate feature vectors that can be used to classify cells based on physically and human-interpretable properties. In some embodiments, this subject matter system provides a high-throughput and robust scheme for generating single-cell phenotypic analysis from image data, sequence data (e.g., raw waveform data), and tabular data (e.g., raw or unmixed spectral and scattered light data) to assess cell type and morphological characteristics in a sample. In some instances, this subject matter system is configured to classify (and sort, as described below) rare cells in a sample. In some instances, the system provides identification of cells in a sample.

[0043] In some embodiments, the system described herein is capable of improving the accuracy of cell classification in a sample by 5% or more, such as 10% or more, such as 15% or more, such as 25% or more, such as 50% or more, such as 75% or more, and including 99% or more. In some instances, it is capable of sorting or preparing classified cells in a sample such that 50% or more of the cells in the sample are suitable for downstream bioassays, such as 60% or more, such as 70% or more, such as 80% or more, such as 90% or more, such as 95% or more, such as 97% or more, and including 99% or more.

[0044] In this embodiment, the sample irradiated in the flow stream can be a biological sample. A "biological sample" can refer to a whole organism, plant, fungus, or a subset of animal tissue, cells, or components, and in some embodiments, can be found in blood, mucus, lymph, synovial fluid, cerebrospinal fluid, saliva, bronchoalveolar lavage fluid, amniotic fluid, amniotic cord blood, urine, vaginal secretions, and semen. Therefore, a "biological sample" refers both to a natural organism or a subset of its tissues and to homogenates, lysates, or extracts prepared from a subset of an organism or its tissues, including but not limited to, plasma, serum, cerebrospinal fluid, lymph, skin sections, respiratory tract, gastrointestinal tract, cardiovascular and genitourinary tract, tears, saliva, breast milk, blood cells, tumors, and organs. A biological sample can be any type of body tissue, including healthy tissue and diseased tissue (e.g., cancerous tissue, malignant tissue, necrotic tissue, etc.). In some embodiments, the biological sample is a liquid sample, such as blood or its derivatives, such as plasma, tears, urine, semen, etc. In some instances, the sample is a blood sample, including whole blood, such as blood obtained from venipuncture or finger prick (where the blood may or may not be bound to any reagents such as preservatives, anticoagulants, etc., before testing). In some instances, the sample is a whole blood sample.

[0045] In some embodiments, the source of the sample is "mammal" or "milk animal," terms that are widely used to describe organisms within the class Mammalia, including Carnivora (e.g., dogs and cats), Rodentia (e.g., mice, guinea pigs, and rats), and Primates (e.g., humans, chimpanzees, and monkeys). In some instances, the test subject is a human. The method can be applied to samples obtained from human test subjects of both sexes and at any developmental stage (i.e., newborns, infants, adolescents, teenagers, and adults), wherein, in some embodiments, the human test subject is an adolescent, teenager, or adult. While the invention can be applied to samples from human test subjects, it should be understood that the method can also be performed on samples from other animal test subjects (i.e., "non-human test subjects") (e.g., but not limited to birds, mice, rats, dogs, cats, livestock, and horses).

[0046] In some embodiments, the systems and methods of this subject matter are configured for phenotypic analysis and determination of biological parameters of single cells in a sample. The term "single cell" is used herein in its conventional sense, referring to the characterization of a sample cell at the level of a single cell. Thus, in some instances, the data generated herein comes from each individual cell in the sample, rather than the average or statistical derivative of a large population of cells. In some instances, image data is generated for each individual cell in the sample. In some instances, spectral data of light at a predetermined wavelength is generated from each individual cell in the sample. In some instances, scattering data is generated for each individual cell in the sample. In some instances, raw detector waveforms are generated for each individual cell in the sample.

[0047] In an embodiment, the method includes illuminating sample cells in a flowing stream with a light source. The light source can be any suitable broadband or narrowband light source. Depending on the composition of the sample, the light source can be configured to emit light of different wavelengths, ranging from 200 nm to 1500 nm, for example from 250 nm to 1250 nm, for example from 300 nm to 1000 nm, for example from 350 nm to 900 nm, and including 400 nm to 800 nm. For example, the light source can include a broadband light source that emits light with wavelengths from 200 nm to 900 nm. In other instances, the light source includes a narrowband light source that emits light with wavelengths ranging from 200 nm to 900 nm. For example, the light source can be a narrowband LED (1 nm–25 nm) that emits light with wavelengths ranging from 200 nm to 900 nm. In some embodiments, the light source is a laser. In some instances, the subject system includes gas lasers, such as helium-neon lasers, argon lasers, krypton lasers, xenon lasers, nitrogen lasers, CO2 lasers, CO lasers, argon-fluorine (ArF) excimer lasers, krypton-fluorine (KrF) excimer lasers, xenon-chlorine (XeCl) excimer lasers, or xenon-fluorine (XeF) excimer lasers, or combinations thereof. In other instances, the subject system includes dye lasers, such as stilbene, coumarin, or rhodamine lasers. In still other instances, lasers of interest include metal vapor lasers, such as helium-cadmium (HeCd) lasers, helium-mercury (HeHg) lasers, helium-selenium (HeSe) lasers, helium-silver (HeAg) lasers, strontium lasers, neon-copper (NeCu) lasers, copper lasers, or gold lasers, or combinations thereof. In other instances, the subject system includes solid-state lasers, such as ruby ​​lasers, Nd:YAG lasers, NdCrYAG lasers, Er:YAG lasers, Nd:YLF lasers, Nd:YVO4 lasers, Nd:yCa4O(BO3)3 lasers, Nd:YCOB lasers, Ti:sapphire lasers, thulium YAG lasers, ytterbium YAG lasers, ytterbium₂O₃ lasers, or cerium-doped lasers and combinations thereof.

[0048] In other embodiments, the light source is a non-laser light source, such as a lamp, including but not limited to halogen lamps, deuterium arc lamps, xenon arc lamps, light-emitting diodes (e.g., broadband LEDs with a continuous spectrum), superluminescent light-emitting diodes, semiconductor light-emitting diodes, broadband LED white light sources, and multi-LED integration. In some instances, the non-laser light source is a stable fiber-coupled broadband light source, a white light source, and other light sources or any combination thereof.

[0049] The light source can be positioned at any suitable distance from the sample (e.g., the flow stream in a flow cytometer), such as 0.001 mm or greater, for example 0.005 mm or greater, 0.01 mm or greater, 0.05 mm or greater, 0.1 mm or greater, 0.5 mm or greater, 1 mm or greater, 5 mm or greater, 10 mm or greater, 25 mm or greater, and including distances of 100 mm or greater. Furthermore, the light source illuminates the sample at any suitable angle (e.g., relative to the vertical axis of the flow stream), such as angles from 10° to 90°, for example 15° to 85°, 20° to 80°, 25° to 75°, and including angles from 30° to 60°, for example 90°.

[0050] The light source can be configured to illuminate the sample continuously or at discrete intervals. In some instances, the system includes a light source configured to continuously illuminate the sample, such as using a continuous-wave laser to continuously illuminate the flow stream at the interrogation point of a flow cytometer. In other cases, the system of interest includes a light source configured to illuminate the sample at discrete intervals, such as every 0.001 milliseconds, every 0.01 milliseconds, every 0.1 milliseconds, every 1 millisecond, every 10 milliseconds, every 100 milliseconds, and including every 1000 milliseconds or other intervals. When the light source is configured to illuminate the sample at discrete intervals, the system may include one or more additional components to provide intermittent illumination of the sample with the light source. For example, the subject system in these embodiments may include one or more laser beam choppers, manually or computer-controlled beam blockers, for blocking the sample and exposing it to the light source.

[0051] In some embodiments, the light source is a laser. Lasers of interest may include pulsed lasers or continuous-wave lasers. For example, the laser may be a gas laser, such as a helium-neon laser, an argon laser, a krypton laser, a xenon laser, a nitrogen laser, a CO2 laser, a CO laser, an argon-fluorine (ArF) excimer laser, a krypton-fluorine (KrF) excimer laser, a xenon-chlorine (XeCl) excimer laser, or a combination thereof; a dye laser, such as a stilbene, coumarin, or rhodamine laser; or a metal vapor laser, such as a helium-cadmium (HeCd) laser, a helium-mercury (HeHg) laser, a helium-selenium (HeSe) laser, a helium-silver (HeAg) laser, a strontium laser, or a neon-copper (Neon) laser. NeCu lasers, copper lasers, or gold lasers and combinations thereof; solid-state lasers, such as ruby ​​lasers, Nd:YAG lasers, NdCrYAG lasers, Er:YAG lasers, Nd:YLF lasers, Nd:YVO4 lasers, Nd:yCa4O(BO3)3 lasers, Nd:YCOB lasers, Ti:sapphire lasers, thulium YAG lasers, ytterbium YAG lasers, ytterbium₂O₃ lasers, or cerium-doped lasers and combinations thereof; semiconductor diode lasers, optically pumped semiconductor lasers (OPSL), or frequency doubling or third harmonics of any of the above lasers.

[0052] In some embodiments, the light source is a beam generator configured to generate two or more frequency-shifted beams. In some instances, the beam generator includes a laser, a radio frequency (RF) generator configured to apply an RF drive signal to the acousto-optic device to generate two or more angle-deflected laser beams. In these embodiments, the laser can be a pulsed laser or a continuous-wave laser. For example, the laser in the beam generator of interest can be a gas laser, such as a helium-neon laser, an argon laser, a krypton laser, a xenon laser, a nitrogen laser, a CO2 laser, a CO laser, an argon-fluorine (ArF) excimer laser, a krypton-fluorine (KrF) excimer laser, a xenon-chlorine (XeCl) excimer laser, or a xenon-fluorine (XeF) excimer laser or a combination thereof; a dye laser, such as a stilbene, coumarin, or rhodamine laser; or a metal vapor laser, such as a helium-cadmium (HeCd) laser, a helium-mercury (HeHg) laser, or a helium-selenium (H) laser. eSe) lasers, helium-silver (HeAg) lasers, strontium lasers, neon-copper (NeCu) lasers, copper lasers, or gold lasers and combinations thereof; solid-state lasers, such as ruby ​​lasers, Nd:YAG lasers, NdCrYAG lasers, Er:YAG lasers, Nd:YLF lasers, Nd:YVO4 lasers, Nd:yCa4O(BO3)3 lasers, Nd:YCOB lasers, titania-sapphire lasers, thulium YAG lasers, ytterbium YAG lasers, ytterbium₂O₃ lasers, or cerium-doped lasers and combinations thereof.

[0053] The acousto-optic device can be any suitable acousto-optic scheme configured to frequency-shift a laser using applied acoustic waves. In some embodiments, the acousto-optic device is an acousto-optic deflector. The acousto-optic device in this subject matter system is configured to generate an angle-deflected laser beam from light from a laser and an applied radio frequency (RF) drive signal. Any suitable RF drive signal source can be used to apply the RF drive signal to the acousto-optic device, such as a direct digital synthesizer (DDS), an arbitrary waveform generator (AWG), or an electrical pulse generator.

[0054] In an embodiment, the controller is configured to apply radio frequency drive signals to the acousto-optic device to generate a laser beam with a desired number of angular deflections in the output laser beam. For example, it is configured to apply 3 or more radio frequency drive signals, such as 4 or more radio frequency drive signals, such as 5 or more radio frequency drive signals, such as 6 or more radio frequency drive signals, such as 7 or more radio frequency drive signals, such as 8 or more radio frequency drive signals, such as 9 or more radio frequency drive signals, such as 10 or more radio frequency drive signals, such as 15 or more radio frequency drive signals, such as 25 or more radio frequency drive signals, such as 50 or more radio frequency drive signals, and includes being configured to apply 100 or more radio frequency drive signals.

[0055] In some instances, in order to generate an intensity distribution of an angle-deflected laser beam in the output laser beam, the controller is configured to apply an amplitude-varying radio frequency drive signal, for example from about 0.001V to about 500V, for example from about 0.005V to about 400V, for example from about 0.01V to about 300V, for example from about 0.05V to about 200V, for example from about 0.1V to about 100V, for example from about 0.5V to about 75V, for example from about 1V to 50V, for example from about 2V to 40V, for example from 3V to about 30V, and including from about 5V to about 25V. In some embodiments, each applied radio frequency drive signal has a frequency of about 0.001 MHz to about 500 MHz, for example about 0.005 MHz to about 400 MHz, for example about 0.01 MHz to about 300 MHz, for example about 0.05 MHz to about 200 MHz, for example about 0.1 MHz to about 100 MHz, for example about 0.5 MHz to about 90 MHz, for example about 1 MHz to about 75 MHz, for example about 2 MHz to about 70 MHz, for example about 3 MHz to about 65 MHz, for example about 4 MHz to about 60 MHz, and includes about 5 MHz to about 50 MHz.

[0056] In some embodiments, the controller has a processor having a memory operatively coupled to the processor, such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to produce an output laser beam with angular deflections of laser beams having a desired intensity distribution. For example, the memory may include instructions to produce two or more laser beams with the same intensity at angular deflections, such as three or more, four or more, five or more, ten or more, 25 or more, or 50 or more, and the memory may include instructions to produce 100 or more laser beams with the same intensity at angular deflections. In other embodiments, the memory may include instructions to produce two or more laser beams with different intensities at angular deflections, such as three or more, four or more, five or more, ten or more, 25 or more, or 50 or more, and the memory may include instructions to produce 100 or more laser beams with different intensity at angular deflections.

[0057] In some embodiments, the controller has a processor having a memory operatively coupled to the processor, such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam whose intensity increases along a horizontal axis from the edge to the center. In these examples, the intensity of the laser beam angularly deflected at the center of the output beam can be from 0.1% to about 99% of the intensity of the laser beam angularly deflected at the edge of the output laser beam along the horizontal axis, for example, from 0.5% to about 95%, from 1% to about 90%, from about 2% to about 85%, from about 3% to about 80%, from about 4% to about 75%, from about 5% to about 70%, from about 6% to about 65%, from about 7% to about 60%, from about 8% to about 55%, and includes from about 10% to about 50% of the intensity of the laser beam angularly deflected at the edge of the output laser beam along the horizontal axis. In other embodiments, the controller has a processor having a memory operatively coupled to the processor, such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam whose intensity increases along a horizontal axis from the edge to the center. In these examples, the intensity of the laser beam angularly deflected at the edge of the output beam can be from 0.1% to about 99%, for example from 0.5% to about 95%, for example from 1% to about 90%, for example from about 2% to about 85%, for example from about 3% to about 80%, for example from about 4% to about 75%, for example from about 5% to about 70%, for example from about 6% to about 65%, for example from about 7% to about 60%, for example from about 8% to about 55%, and includes from about 10% to about 50% of the intensity of the laser beam angularly deflected along the horizontal axis at the center of the output laser beam. In other embodiments, the controller has a processor with a memory operatively coupled to the processor, such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam with an intensity distribution having a Gaussian distribution along a horizontal axis. In other embodiments, the controller has a processor with a memory operatively coupled to the processor, such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam with a top-hat intensity distribution along a horizontal axis.

[0058] In embodiments, the beam generator of interest can be configured to generate spatially separated angle-deflected laser beams within the output laser beam. Depending on the applied radio frequency drive signal and the desired illumination profile of the output laser beam, the angle-deflected laser beams can be separated by 0.001 μm or more, for example, 0.005 μm or more, for example, 0.01 μm or more, for example, 0.05 μm or more, for example, 0.1 μm or more, for example, 0.5 μm or more, for example, 1 μm or more, for example, 5 μm or more, for example, 10 μm or more, for example, 100 μm or more, for example, 500 μm or more, for example, 1000 μm or more, and including 5000 μm or more. In some embodiments, the system is configured to generate mutually overlapping angle-deflected laser beams within the output laser beam, for example, overlapping with adjacent angle-deflected laser beams along the horizontal axis of the output laser beam. The overlap between laser beams with adjacent angle deflections (e.g., beam spot overlap) can be 0.001 μm or more, such as 0.005 μm or more, such as 0.01 μm or more, such as 0.05 μm or more, such as 0.1 μm or more, such as 0.5 μm or more, such as 1 μm or more, such as 5 μm or more, such as 10 μm or more, and includes overlaps of 100 μm or more.

[0059] In some instances, beam generators configured to produce two or more frequency-shifted beams include laser excitation modules, as described by Diebold et al. in Nature Photonics Vol. 7(10); 806-810(2013), and as described in U.S. Patents 9,423,353, 9,784,661, 9,983,132, 10,006,852, 10,078,045, 10,036,699, 10,222,316, 10,288,546, and 10,324,010. The disclosures described in U.S. Patent Publications Nos. 9, 10,408,758, 10,451,538, 10,620,111, and 2017 / 0133857, 2017 / 0328826, 2017 / 0350803, 2018 / 0275042, 2019 / 0376895, and 2019 / 0376894 are incorporated herein by reference.

[0060] In embodiments, the system includes a light detection system with multiple photodetectors for measuring light from cells in a sample. In some instances, one or more photodetectors of the light detection system are light loss photodetectors. In some instances, one or more photodetectors of the light detection system are configured to measure scattered light. In some instances, one or more photodetectors of the light detection system are configured to measure side-scattered light. In some instances, one or more photodetectors of the light detection system are configured to measure forward-scattered light. In some instances, one or more photodetectors of the light detection system are configured to measure back-scattered light. The photodetectors of interest may include, but are not limited to, optical sensors such as active pixel sensors (APS), avalanche photodiodes (APDs), image sensors, charge-coupled devices (CCDs), enhancement-mode charge-coupled devices (ICCDs), light-emitting diodes, photon counters, calorimeters, pyroelectric detectors, photoresistors, photovoltaic cells, photodiodes, photomultiplier tubes, phototransistors, quantum dot photoconductors, or combinations thereof, as well as other photodetectors. In some embodiments, a charge-coupled device (CCD), a semiconductor charge-coupled device (CCD), an active pixel sensor (APS), a complementary metal-oxide-semiconductor (CMOS) image sensor, or an N-type metal-oxide-semiconductor (NMOS) image sensor is used to measure the light from the sample.

[0061] In some embodiments, the light detection system of interest includes a plurality of photodetectors. In some instances, the light detection system includes a plurality of solid-state detectors, such as photodiodes. In some instances, the light detection system includes a photodetector array, such as a photodiode array. In these embodiments, the photodetector array may include four or more photodetectors, such as ten or more photodetectors, such as 25 or more photodetectors, such as 50 or more photodetectors, such as 100 or more photodetectors, such as 250 or more photodetectors, such as 500 or more photodetectors, such as 750 or more photodetectors, and may include 1000 or more photodetectors. For example, the detector may be a photodiode array having four or more photodiodes, such as ten or more photodiodes, such as 25 or more photodiodes, such as 50 or more photodiodes, such as 100 or more photodiodes, such as 250 or more photodiodes, such as 500 or more photodiodes, such as 750 or more photodiodes, and may include 1000 or more photodiodes.

[0062] The photodetectors can be arranged in any geometric configuration as needed, including, but not limited to, square, rectangular, trapezoidal, triangular, hexagonal, heptagonal, octagonal, nonagonal, decagonal, dodecagonal, circular, elliptical, and irregular pattern configurations. The photodetectors in the photodetector array can be oriented at an angle of 10° to 180° relative to another (e.g., a reference XZ plane), such as 15° to 170°, 20° to 160°, 25° to 150°, 30° to 120°, and including 45° to 90°. The photodetector array can be any suitable shape and can be linear, such as square, rectangular, trapezoidal, triangular, hexagonal, etc., or curved, such as circular, elliptical, and irregular, such as a parabolic base coupled to the top of a plane. In some embodiments, the photodetector array has a rectangular active surface.

[0063] Each photodetector (e.g., a photodiode) in the array may have an active surface with a width ranging from 5 μm to 250 μm, such as 10 μm to 225 μm, 15 μm to 200 μm, 20 μm to 175 μm, 25 μm to 150 μm, 30 μm to 125 μm, and including 50 μm to 100 μm, and a length ranging from 5 μm to 250 μm, such as 10 μm to 225 μm, 15 μm to 200 μm, 20 μm to 175 μm, 25 μm to 150 μm, 30 μm to 125 μm, and including 50 μm to 100 μm, wherein the surface area of ​​each photodetector (e.g., a photodiode) in the array is 25 μm. 2 Up to 10000μm 2 For example, 50μm 2 Up to 9000μm 2 For example, 75μm 2 up to 8000μm 2 For example, 100μm 2 up to 7000μm 2 For example, 150μm 2 up to 6000μm 2 And including 200μm 2 up to 5000μm 2 .

[0064] The size of the photodetector array can vary depending on the amount and intensity of light, the number of photodetectors, and the desired sensitivity. The length can range from 0.01 mm to 100 mm, for example, 0.05 mm to 90 mm, 0.1 mm to 80 mm, 0.5 mm to 70 mm, 1 mm to 60 mm, 2 mm to 50 mm, 3 mm to 40 mm, 4 mm to 30 mm, and includes 5 mm to 25 mm. The width of the photodetector array can also vary from 0.01 mm to 100 mm, for example, 0.05 mm to 90 mm, 0.1 mm to 80 mm, 0.5 mm to 70 mm, 1 mm to 60 mm, 2 mm to 50 mm, 3 mm to 40 mm, 4 mm to 30 mm, and includes 5 mm to 25 mm. Thus, the active surface area of ​​the photodetector array can range from 0.1 mm. 2 Up to 10000mm 2 For example, 0.5mm 2 Up to 5000mm 2 For example, 1mm 2 Up to 1000mm 2 For example, 5mm 2 Up to 500mm 2 And including 10mm 2 Up to 100mm 2 .

[0065] The photodetector of interest is configured to measure collected light at one or more wavelengths, such as at two or more wavelengths, such as at five or more different wavelengths, such as at ten or more different wavelengths, such as at 25 or more different wavelengths, such as at 50 or more different wavelengths, such as at 100 or more different wavelengths, such as at 200 or more different wavelengths, such as at 300 or more different wavelengths, and includes measuring light emitted by a sample in the flow at 400 or more different wavelengths.

[0066] In some embodiments, the photodetector is configured to measure light collected within a wavelength range (e.g., 200 nm to 1000 nm). In some embodiments, the photodetector of interest is configured to collect a spectrum within a wavelength range. For example, the system may include one or more detectors configured to collect a spectrum within one or more wavelength ranges from 200 nm to 1000 nm. In other embodiments, the photodetector of interest is configured to measure light from a sample in a flowing stream at one or more specific wavelengths. For example, the system may include one or more detectors configured to measure light at one or more of the following wavelengths: 450 nm, 518 nm, 519 nm, 561 nm, 578 nm, 605 nm, 607 nm, 625 nm, 650 nm, 660 nm, 667 nm, 670 nm, 668 nm, 695 nm, 710 nm, 723 nm, 780 nm, 785 nm, 647 nm, 617 nm, and any combination thereof.

[0067] The light detection system is configured to measure light continuously or at discrete intervals. In some instances, the photodetector of interest is configured to continuously measure the collected light. In other instances, the light detection system is configured to perform measurements at discrete intervals, such as every 0.001 milliseconds, every 0.01 milliseconds, every 0.1 milliseconds, every 1 millisecond, every 10 milliseconds, every 100 milliseconds, and including measuring light at 1000 milliseconds or other intervals.

[0068] In some embodiments, the system includes a processor having a memory operatively coupled to the processor, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate data in response to measured light. Data signals from the light detection system can be generated in multiple different photodetector channels, such as two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, twelve or more, sixteen or more, twenty-four or more, thirty-two or more, sixty-four or more, and including 128 photodetector channels. In some instances, data is generated in one or more fluorescence photodetector channels. In some instances, data is generated in light loss photodetector channels. In some instances, data is generated in scattered light photodetector channels (e.g., forward-scattering photodetector channels, side-scattering photodetector channels).

[0069] In some instances, the generated data is image data. In some instances, the image data includes dark-field images of cells. In some instances, the image data includes light-field images of cells. In some instances, the image data includes fluorescently tagged images of cells. In some instances, the image data includes untagged images of cells. In some instances, the image data includes autofluorescence image data. The image data may include one or more images of cells, such as two or more, three or more, four or more, five or more, ten or more, fifteen or more, twenty-five or more, and images of 50 or more cells. In some instances, the image data includes two or more images of different types (e.g., fluorescently tagged images, autofluorescence images, dark-field images, etc.), such as three or more images of different types, such as four or more images of different types, and five or more images of different types. In some embodiments, the image data includes images that have been converted from one type to another, for example, an image generated in a light-loss photodetector channel (e.g., a bright-field image) is inverted to create a dark-field image.

[0070] In some embodiments, the system includes a memory storing instructions for generating an image based on detected light absorption, detected light scattering, detected light emission, or any combination thereof. In some instances, the memory includes instructions for generating an image based on light absorption detected from a sample (e.g., from a bright-field detector). In some instances, the memory includes instructions for generating an image based on light scattering detected from a sample, such as from a side-scattering detector, a forward-scattering detector, or a combination of both. In some instances, the memory includes instructions for generating an image based on light emitted from the sample. In other instances, the memory includes instructions for generating an image based on a combination of detected light absorption and detected light scattering.

[0071] In embodiments, the memory includes instructions for receiving two or more datasets associated with multiple data patterns. In some instances, the data patterns are image data patterns. In some instances, the data patterns are sequence data patterns (e.g., waveforms). In some instances, the data patterns are tabular data patterns. The methods and systems of this invention are capable of receiving image data, waveform data, spectral data, and scattering data, enabling the inclusion of data of various forms from different sources. The capabilities of the methods and systems of this invention provide a general workflow for flow cytometry data analysis that integrates multimodal datasets such as image, waveform, and tabular data.

[0072] In some embodiments, the multiple data modes include image data modes. In some instances, the memory includes instructions for generating a single image for each cell in the sample based on each form of detection light. In other embodiments, the memory includes instructions for generating multiple images for each cell, such as two or more, three or more, five or more, ten or more, and including generating 25 or more images for each cell. For example, the memory includes instructions for generating a first image of a cell based on fluorescence detected from a tagged cell; instructions for generating a second image of the cell based on detected light absorption; and instructions for generating a third image based on detected light scattering. In other embodiments, the memory includes instructions for generating two or more images from each form of detection light, such as three or more, four or more, five or more, and including ten or more images or combinations thereof.

[0073] In some instances, the algorithm is a rule-based image processing algorithm. In some instances, the algorithm includes a neural network. In some instances, the memory includes instructions for reversing the light loss channels of a dataset associated with an image data pattern, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to zero mean, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to unit standard deviation, or a combination thereof.

[0074] In some instances, the memory includes instructions for generating image data, such as digital waveforms generated in one or more photodetector channels. In some instances, the memory includes instructions for calculating image parameters directly from the waveform. In some instances, the memory includes instructions for calculating image parameters from image data, the generated image, or a combination thereof. In some instances, the memory includes instructions for determining cell parameters solely from the calculated image parameters (e.g., radial moment, eccentricity, etc.). In some instances, the memory includes instructions for determining cell parameters solely from an image of the particle, rather than from another data source (e.g., a data signal waveform). In some instances, the memory includes instructions for determining cell parameters using a combination of the generated particle image and a data signal waveform generated in response to measurement light from the irradiated particle.

[0075] In some embodiments, the system includes a memory with instructions for generating frequency-coded data (e.g., frequency-coded spatial data) from measurement light of sample cells in a flowing stream. In some instances, the memory includes instructions for generating one or more images from the frequency-coded data. The frequency-coded data may be generated in one or more detection channels, such as two or more, three or more, four or more, five or more, six or more, and including eight or more detection channels. In some embodiments, the frequency-coded data includes data components acquired (or derived) from light from different detectors, such as detected light absorption or detected light scattering. In some instances, the memory includes instructions for performing phase correction on the frequency-coded data. In some instances, the memory includes instructions for generating a phase-corrected image of the cells by performing a transform on the frequency-coded data. In one example, the memory includes instructions for performing phase correction on the frequency-coded data by performing a Fourier transform (FT). In another example, the memory includes instructions for performing phase correction on the frequency-coded data by performing a discrete Fourier transform (DFT). In yet another example, the memory includes instructions to perform phase correction on the frequency-coded data by performing a short-time Fourier transform (STFT) on the frequency-coded data. In some embodiments, the memory includes instructions to perform a transform on the frequency-coded data without performing any mathematical imaginary calculations (i.e., only performing mathematical real-valued calculations on the transform) to generate an image from the frequency-coded data.

[0076] In some embodiments, the multiple data modes include a sequence data mode. In some instances, the sequence data includes waveforms generated in one or more different photodetector channels. In some instances, the sequence data includes raw waveforms. In some instances, the sequence data includes raw image data in the form of raw waveforms generated in imaging photodetector channels. In some instances, the sequence data includes raw waveforms, and the image data is generated in real time from the raw waveforms.

[0077] In some embodiments, multiple data modes include a tabular data mode. In some instances, tabular data is applied as a stream cytometry data file of tabular data to a neural network. In some instances, the tabular data is spectral tabular data generated in one or more fluorescence photodetector channels. In some instances, the spectral tabular data is generated from fluorescence measured under light in one or more spectral wavelength ranges, such as 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, twelve or more, sixteen or more, and includes light in 24 or more different spectral wavelength ranges. In some instances, the spectral tabular data is uncompensated or unprocessed spectral data. In some instances, the spectral tabular data is compensated or unmixed spectral data. Spectral tabular data can be unmixed by spectral analysis of the light from each fluorophore in the sample (e.g., using a weighted least squares algorithm or a generalized least squares algorithm). In some embodiments, overlap between each distinct fluorophore is determined, and the contribution of each fluorophore to the overlapping fluorescence is calculated. In some embodiments, spectral analysis of the spectral tabulation data is performed by calculating the spectral unmixing matrix of the fluorescence spectra of each of the plurality of fluorophores with overlapping fluorescence in the sample detected by the photodetector system. For example, spectral unmixing of the spectral tabulation data may include the Moore-Penrose inverse or pseudo-inverse of the spectral matrix. In some instances, the algorithm used for spectral unmixing is characterized by the Cholesky decomposition of the unmixing matrix. In some embodiments, the unmixed spectral tabulation data is calibrated using a calibration scaling factor and fluorophore abundance.

[0078] In some instances, spectral table data from each fluorophore (e.g., calculating the spectral unmixing matrix for each fluorophore) can be used to estimate the abundance of each fluorophore in the sample. In some embodiments, the abundance of each fluorophore associated with the target particle can be determined. In some embodiments, the spectral table data is spectrally unmixed using algorithms such as those described in U.S. Patent No. 11,009,400, U.S. Patent Publication No. 2024 / 0192122, filed December 12, 2023, and U.S. Patent Application No. 18 / 986,295, filed December 18, 2024, the disclosures of which are incorporated herein by reference.

[0079] In some instances, the cells in the sample are labeled with one or more fluorescent markers. In some embodiments, the fluorescent dyes of interest may include, but are not limited to, bodipy dyes, coumarin dyes, rhodamine dyes, acridine dyes, anthraquinone dyes, arylmethane dyes, diarylmethane dyes, chlorophyll-containing dyes, triarylmethane dyes, azo dyes, diazo dyes, nitro dyes, nitroso dyes, phthalocyanine dyes, anthocyanin dyes, asymmetric anthocyanin dyes, quinone imine dyes, azazine dyes, ohhodin dyes, saffron dyes, indamine, indophenol dyes, fluorine dyes, oxazine dyes, oxazone dyes, thiazine dyes, thiazolium dyes, xanthan dyes, fluorene dyes, pyranine dyes, fluorine dyes, rhodamine dyes, phenanthridine dyes, squaric acid cyanine, BODIPY, squaric acid roxane, naphthalenes, coumarin, oxadiazole, anthracene, pyrene, acridine, arylimine, or tetrapyrrole and combinations thereof. In some embodiments, the conjugate may include two or more dyes, such as two or more dyes selected from the following: bodie dyes, coumarin dyes, rhodamine dyes, acridine dyes, anthraquinone dyes, arylmethane dyes, diarylmethane dyes, chlorophyll-containing dyes, triarylmethane dyes, azo dyes, diazo dyes, nitro dyes, nitroso dyes, phthalocyanine dyes, cyanine dyes, asymmetric cyanine dyes, quinone imine dyes, azazine dyes, ohhodin dyes, saffron dyes, indamine, indophenol dyes, fluorine dyes, oxazine dyes, oxazone dyes, thiazine dyes, thiazole dyes, xanthan dyes, fluorene dyes, pyranine dyes, fluorine dyes, rhodamine dyes, phenanthridine dyes, squaric acid cyanine, BODIPY, squaric acid roxane, naphthalene, coumarin, oxadiazole, anthracene, pyrene, acridine, arylimine, or tetrapyrrole and combinations thereof.

[0080] In some embodiments, the fluorescent dyes of interest may include, but are not limited to: fluorescein isothiocyanate (FITC), phycoerythrin (PE) dyes, polydinophytic chlorophyll-cyanin dyes (e.g., PerCP-Cy5.5), phycoerythrin-cyanin (PE-Cy) dyes (PE-Cy7), allophycocyanin (APC) dyes (e.g., APC-R700), allophycocyanin-cyanin dyes (e.g., APC-Cy7), and coumarin dyes (e.g., V450 or V500). In some instances, fluorescent dyes may include one or more selected from the following: 1,4-bis(o-methylstyryl)-benzene (bis-MSB 1,4-bis[2-(2-methylphenyl)vinyl]-benzene), C510 dye, C6 dye, Nile Red dye, T614 dye (e.g., N-[7-(methanesulfonamido)-4-oxo-6-phenoxychromene-3-yl]formamide), LDS 821 dye ((2-(6-(p-dimethylaminophenyl)-2,4-neopentenyl-1,3,5-hextrienyl)-3-ethylbenzothiazole perchlorate), mFluor dye (e.g., mFlur red dye, such as mFluor780NS)).

[0081] Fluorescent dyes of interest may include, but are not limited to: fluorescein, hydroxycoumarin, aminocoumarin, methoxycoumarin, cascade blue, Pacific blue, Pacific orange, fluorescent yellow, NBD, R-phycoerythrin (PE), PE-Cy5 conjugate, PE-Cy7 conjugate, red 613, PerCP, TruRed, FluorX, BODIPY-FL, TRITC, X-rhodamine, serinerhodamine B, Texas red, allophycocyanin (APC), APC-Cy7 conjugate, Cy2, Cy3, Cy3B, Cy3.5, Cy5, Cy5.5, Cy7, Hoechst 33342, DAPI, Hoechst 33258, SYTOX blue, chromomycin A3, mitomycin, YOYO-1, ethidium bromide, acridine orange, SYTOX green, TOTO-1, TO-PRO-1, thiazole orange, propidium iodide (PI), LDS. 751, 7-AAD, SYTOX Orange, TOTO-3, TO-PRO-3, DRAQ5, Indo-1, Fluo-3, DCFH, DHR, SNARF, Y66H, Y66F, EBFP, EBFP2, Azurite, GFPuv, T-Sapphire, TagBFP, Sky Blue, mCFP, ECFP, CyPet, Y66W, dKeima-Red, mKeima-Red, TagCFP, AmCyan1, mTFP1 (Cyan), S65A, Midoriishi-Cyan, Wild-type GFP, S65C, TurboGFP, TagGFP, TagGFP2, AcGFP1, S65L, Emerald, S65T, EGFP, Azami-Green, ZsGreen1, Dronpa-Green, TagYFP, EYFP, Topaz, Venus, mCi trine, YPet, TurboYFP, PhiYFP, PhiYFP-m, ZsYellow1, mBanana, Kusabira-Orange, mOrange, mOrange2, mKO, TurboRFP, tdTomato, DsRed-Express2, TagRFP, DsRed Monomer, DsRed2 (“RFP”), mStrawberry, TurboFP602, AsRed2, mRFP1, J-Red, mCherry, HcRed1, mKate2, Katushka (TurboFP635), mKate (TagFP635), TurboFP635, mPlum, mRaspberry, mNeptune, E2-Crimson, Manganese Monochloro, Calceflavin, Alexa Fluor 350, Alexa Fluor 405, Alexa Fluor 430, Alexa Fluor488, AlexaFluor500, Alexa Fluor 514, Alexa Fluor 532, Alexa Fluor 546, Alexa Fluor 555, Alexa Fluor 568, Alexa Fluor 594, Alexa Fluor 610, Alexa Fluor 633, Alexa Fluor 647, Alexa Fluor 660, Alexa Fluor 680, Alexa Fluor 700, Alexa Fluor 750, Alexa Fluor 790, and Hyper, etc. In some embodiments, the fluorescent dye is selected from: 7-AAD, Alexa Fluor 488, Alexa Fluor 647, Alexa Fluor... 700, AmCyan, APC, APC-Cy7, APC-H7, APC-R700, BB660-P2, BB790-P, BUV395, BUV615, BUV661, BV570, BV605, BV650, BV711, BV750, BV786, BYG584-P, Calcein AM, Calcein Blue AM, CFSE, DAPI, DRAQ5, DRAQ7, FITC, Fluo-4AM, FVS440UV, FVS450, FVS510, FVS520, FVS570, FVS575V, FVS620, FVS660, FVS700, FVS780, Indo-1Hi, Indo-1Lo, JC-1, MitoStatus Red, MitoStatus TMRE, Pacific Blue, PE, PE-CF594, PE-Cy5, PE-Cy7, PerCP, PerCP-Cy5.5, PI, R718, RB545, RB613, RB744, RB780, RY586, RY610, V450, V500, Via-Probe Green, Via-Probe Red, VPD450, Alexa Fluor 532, Alexa Fluor 561, Alexa Fluor 660, APC-eFluor 780, APC / Fire750, APC / Fire 810, BV785, eBFP, eCFP, eFluor 450, eFluor 506, eFluor 660, eGFP, eYFP, Hoechst33258, KIRAVIA blue 520, mCherry, NFB510, NFB530, NFB555, NFB585, NFB610-70S, NFB660-120S, NFR660, NFR685 , NFR700, NFR710, NFY570, NFY590, NFY610, NFY660, NFY690, NFY700, NFY730, Pacific Orange, PE-Cy5.5, PE-eFluor 610, PE / Dazzle594, PE / Fire 640, PE / Fire 700, PE / Fire 810, PerCP-eFluor 710, SB436, SB600, SB645, SB702, SB780, Spark Blue 550, Spark Blue 574, Spark NIR 685, Spark UV 387, Spark Purple 423, Spark Purple 538, Spark YG 581, Spark YG 593 and tdTomato.

[0082] In some instances, the fluorescent dye is a polymer dye (e.g., a fluorescent polymer dye). The fluorescent polymer dyes used in the methods and systems of this subject are diverse. In some instances of this method, the polymer dye includes conjugated polymers. Conjugated polymers (CPs) are characterized by a delocalized electronic structure, comprising a backbone of alternating unsaturated bonds (e.g., double and / or triple bonds) and saturated bonds (e.g., single bonds), where π-electrons can move from one bond to another. Therefore, the conjugated backbone can impart an extended linear structure to the polymer dye, with finite bond angles between the repeating units of the polymer. For example, proteins and nucleic acids, while also polymers, do not form extended rod-like structures in some cases, but rather fold into higher-order three-dimensional shapes. Furthermore, CPs can form “rigid rod” polymer backbones and undergo finite twist (e.g., torsion) angles between the monomeric repeating units along the polymer backbone. In some instances, the polymer dye includes CPs with a rigid rod structure. The structural features of the polymer dye can influence the fluorescence properties of the molecule.

[0083] Polymer dyes of interest include, but are not limited to: U.S. Patents Nos. 7,270,956, 7,629,448, 8,158,444, 8,227,187, 8,455,613, 8,575,303, 8,802,450, 8,969,509, 9,139,869, and 9,371,559. The dyes described in Nos. 9,547,008, 10,094,838, 10,302,648, 10,458,989, 10,641,775 and 10,962,546, the disclosures of which are incorporated herein by reference in their entirety; and the dyes described in Gaylord et al., J. Am. Chem. Soc., 2001, 123(26), pp 6417–6418; Feng et al., Chem. Soc. Rev., 2010, 39, 2411–2419; and Traina et al., J. Am. Chem. Soc., 2011, 133(32), pp 12600–12607, the disclosures of which are incorporated herein by reference in their entirety. Specific polymer dyes that may be used include, but are not limited to, BD Horizon Brilliant. TM Dyes, such as BD HorizonBrilliant TM Purple dyes (e.g., BV421, BV510, BV605, BV650, BV711, BV786); BD HorizonBrilliant TM UV dyes (e.g., BUV395, BUV496, BUV737, BUV805); and BD HorizonBrilliant TM Blue dyes (e.g., BB515) (BD Biosciences, San Jose, CA). Any fluorescent dye known to those skilled in the art—including, but not limited to, those described above—or any fluorescent dye not yet discovered may be used.

[0084] In some embodiments, the table data includes scattering table data. In some instances, the scattering table data is generated from measured forward-scattered light from the sample (i.e., the photodetector signal generated in the forward-scattering photodetector channel FSC). In some instances, the scattering table data is generated from measured side-scattered light from the sample (i.e., the photodetector signal generated from the side-scattering photodetector channel SSC).

[0085] In embodiments, the memory includes instructions stored thereon that, when executed by the system's processor, cause the processor to apply algorithms to transform each of two or more datasets into feature vectors. In some instances, the feature vectors concisely summarize and normalize the parameters and information contained in the generated data. In some instances, the feature vectors described herein are human-interpretable features of cells. In some instances, the feature vectors are physical features of cells. In some instances, the feature vectors are units of measurement of cells (e.g., dimensions in nanometers). In some instances, the feature vectors are abundance features of cells, such as the abundance of cellular biomarkers, fluorophores, autofluorescence, or binding molecules. In some cases, converting datasets into feature vectors allows the systems and methods described herein to form a flexible analytical framework to integrate prior knowledge of biological systems by providing only relevant features to a specific classifier. In some cases, converting datasets into feature vectors allows the systems and methods described herein to provide a general pre-trained base model that allows users to train a classification model for their specific task or fine-tune a body model with a smaller dataset.

[0086] Any suitable machine learning algorithm can be implemented to convert various data patterns (e.g., image data patterns, sequence data patterns, and tabular data patterns) into feature vectors. The machine learning methods of interest may include, but are not limited to, linear regression, logistic regression, Naive Bayes, k-nearest neighbor (kNN), random forest, decision tree, support vector machine, gradient boosting, and clustering algorithms. In some embodiments, the system is configured to apply neural networks, such as artificial neural networks, convolutional neural networks, or recurrent neural networks. In some instances, the system is configured to implement a Python script. In some instances, the system is configured to apply artificial neural networks (e.g., convolutional neural networks), Bayesian statistics, learning automata, hidden Markov modeling, linear classifiers, quadratic classifiers, association rule learning, etc.

[0087] In some embodiments, the neural network includes a network of nodes. Nodes can be organized into layers, where the first layer is the input layer from which data flows in. The neural network may also include an output layer from which transformed data flows out. Each individual node may have multiple inputs and a single output (e.g., input layer nodes have only a single input). The output of a node represents a linear combination of the inputs. In other words, the inputs can be multiplied by a relevant constant. The product can accumulate along a node path with a constant offset. The constant offset, or “bias,” can represent another degree of freedom that can be adjusted during training. For example, the constant offset could be a threshold, as this can reduce node values ​​below zero, causing the activation function to output zero.

[0088] An activation function is used to evaluate the resulting value, and this value is used as the output of a node. Nodes in a given layer of a neural network are connected to every node in adjacent layers. A neural network can be trained by minimizing its error using gradient descent and an error function to compare the network's expected output with its actual output. The weights of one or more nodes can be adjusted to model the desired outcome produced by the network. In some instances, neural networks apply the sigmoid activation function, the Adam optimizer, adaptive learning rates, dynamic thresholds, binary classification (e.g., positive vs. negative), binary cross-entropy loss, or any combination thereof.

[0089] In some embodiments, the neural network includes a backpropagation neural network. In some instances, the neural network includes two or more stages, such as three or more stages, such as four or more stages, such as five or more stages, such as six or more stages, such as seven or more stages, such as eight or more stages, such as nine or more stages, such as ten or more stages, such as twelve or more stages, such as sixteen or more stages, such as twenty or more stages, and includes twenty-four or more stages. In some instances, the neural network includes 2 to 32 stages, such as 2 to 24 stages, such as 2 to 16 stages, and includes 2 to 10 stages. In some instances, the neural network includes at least one dropout stage. In some instances, the dropout stage is configured to prevent overfitting during training.

[0090] In some embodiments, each stage includes one or more hidden layers, such as two or more hidden layers, three or more hidden layers, four or more hidden layers, five or more hidden layers, six or more hidden layers, seven or more hidden layers, eight or more hidden layers, nine or more hidden layers, ten or more hidden layers, twelve or more hidden layers, sixteen or more hidden layers, twenty-four or more hidden layers, thirty-two or more hidden layers, and includes sixty-four or more hidden layers. In some embodiments, each stage includes at least two hidden layers, at least four hidden layers, at least eight hidden layers, at least twelve hidden layers, at least sixteen hidden layers, at least twenty-four hidden layers, and at least thirty-two hidden layers. In some instances, each stage includes eight to sixty-four hidden layers.

[0091] In some instances, the memory includes instructions for applying a neural network to convert the generated data into feature vectors. In some instances, the neural network includes feature engineering layers for generating feature vectors from two or more datasets (e.g., image data, sequence data, and tabular data). In some instances, the feature engineering layers convert image data into image parameters of cells. In some instances, the feature vectors are quantitative image parameters. In some instances, the feature vectors are the radial moments of the cells. In some instances, the feature vectors are the size of the cells. In some instances, the feature vectors are the diffusion rate of the cells. In some instances, the feature vectors are the eccentricity of the cells. In some instances, the feature vectors are the pointillism of the cells. In some instances, the feature vectors are the shape of the cells. In some instances, the feature vectors are one or more morphological features of the cells.

[0092] In some instances, the feature engineering layer incorporates algorithms for extracting sequential data features from photodetector waveforms. In some instances, the algorithm includes a sequence learning machine learning model. In some instances, the feature engineering layer applies a machine learning model to extract features from waveforms.

[0093] In some instances, the feature-engineered layer includes algorithms for spectral unmixing of spectral tabular data. In some instances, the algorithm includes a compensation model, a spectral unmixing model, or a combination thereof. In some instances, the algorithm includes a calibration model. In some instances, the feature-engineered layer includes algorithms for calculating the abundance of biomarker molecules (e.g., surface biomarker fractions). In some instances, the feature-engineered layer includes algorithms for calculating fluorophore abundance. In some instances, the feature-engineered layer includes algorithms for calculating cellular physical measurements based on scattered light tabular data. For example, the feature-engineered layer may apply a machine learning model that converts the raw scattered channel tabular data into cellular measurements, such as the size of a cell or cellular component in nanometers.

[0094] In one embodiment, the memory includes instructions to apply machine learning algorithms (e.g., neural networks) to determine one or more biological parameters of a cell based on a feature vector. Depending on the application, different analytical models may be applied to extract biological insights into the cell from the feature vector.

[0095] In some embodiments, the analytical model is an end-to-end model. In some instances, the end-to-end model is designed to identify predefined cell phenotypes. In some instances, the end-to-end model is a gated hierarchy or classification model, such as a single-classifier or multi-classifier model (e.g., a single-layer or multi-layer perceptron), a random forest classifier, or a support vector machine. In some embodiments, the analytical model is an exploratory model, such as a model capable of discovering novel cell populations. In some instances, the applied exploratory model is a dimensionality reduction algorithm, such as Uniform Manifold Approximation and Projection (UMAP), t-Distributed Random Nearest Neighbor Embedding (TSNE), and Principal Component Analysis (PCA).

[0096] In some embodiments, the analytical model can select all or a subset of the feature vectors based on prior knowledge of which features are associated with the biological parameters of interest. For example, as described in more detail below, a survival classification task can be developed to identify live cells based on a combination of bright-field and dark-field image data from cells. In these instances, the analytical model can utilize feature vectors from light loss, side scattering, and forward scattering image data. In other instances, the TBNK classification task can be designed to classify cells into T cells, B cells, and natural killer (NK) cells using image data. In some instances, the analytical model applies unlabeled image data and autofluorescence image data to determine biological parameters from the feature vectors.

[0097] In some embodiments, the neural network includes a machine learning encoder layer. In some instances, the encoder layer is a rule-based image processing pipeline. In some instances, the encoder layer includes an image encoder. In some instances, the image encoder has an image recognition model architecture. In some instances, the image encoder has an encoder model architecture such as RestNet34, ShuffleNet V2, EfficientNet V2-S, Inception V3, and combinations thereof. In some instances, the memory includes instructions for training the encoder layer using a supervised method with a weighted cross-entropy loss function. In some instances, the memory includes instructions for reconstructing images in real time based on data waveforms. In some instances, the trained encoder layer is validated based on the evaluated feature quality. In some instances, metrics such as precision, recall, and F1 score are used to determine the evaluated feature quality.

[0098] In some embodiments, the memory includes instructions for applying a neural network to the data to classify cells based on one or more defined feature vectors. In some instances, the memory includes instructions for applying a dimensionality reduction algorithm to the generated data. In some instances, the memory includes instructions for applying a dimensionality reduction algorithm to high-dimensional feature vectors. In some instances, the dimensionality reduction algorithm is selected from Uniform Manifold Approximation and Projection (UMAP), t-Distributed Random Nearest Neighbor Embedding (t-SNE), Principal Component Analysis (PCA), and combinations thereof. In some instances, the memory includes instructions for classifying cells based on a gating hierarchy. In some instances, the memory includes instructions for classifying cells using a classification model, such as a single-layer perceptron, a multilayer perceptron, a random forest classifier, a support vector machine, and combinations thereof.

[0099] In some instances, the memory includes instructions for applying a classification task to two or more datasets and applying one or more classification models to a feature vector based on the received classification task to determine the category. In some instances, receiving and applying the classification task enables the systems and methods of this paper to extract biological insights from heterogeneous datasets and provides generalizable algorithms to generate insightful biological data. The classification task enables the systems and methods to classify cell types and identify specific cells in a sample and their states (e.g., live, activated live cells, dead cells, or inactive live cells). The classification task enables the systems and methods to provide human-interpretable data about cells and, in some instances, to provide sorting decision features.

[0100] In some instances, the classification task is a survival classification task. In some instances, the classification task includes a survival classification task, with categories including a survival category, a death category, or an apoptosis category. In some instances, the memory includes instructions for classifying data using a cell type classification task. In some instances, the cell type classification task classifies cells into T cells, B cells, and natural killer (NK) cells. In some instances, the memory includes instructions for applying a classification task, which includes single-cell classification, and categories including single-cell category, adherent two-cell category, separated two-cell category, three-cell category, or fragment category. In some instances, the classification task includes a whole blood classification task, and whole blood classification includes granulocyte category, monocyte category, or lymphocyte category. In some instances, the classification task includes a T cell activation classification task, and categories including an activated category or an inactivated category.

[0101] Figure 1AA flowchart illustrating the classification of sample cells based on feature vectors according to certain embodiments is described. In step 101, sample cells in a flowing stream are illuminated using a light source. In step 102, light (e.g., fluorescence and light scattering) from the illuminated cells is measured using a light detection system with a photodetector. In step 103, multi-modal data is generated in response to the measured light, including image data modes, sequence data modes (raw waveforms), and tabular data modes (spectral tabular data, scattered light tabular data). Two or more datasets are received by an associated processor (step 104), and in step 105, an algorithm is applied to convert each of the two or more datasets into a feature vector based on the data mode. In step 106, a classification task is received regarding the two or more datasets, and in step 107, based on the received classification task, one or more classification models are applied to the feature vectors to determine the category.

[0102] In some embodiments, the neural network includes a learning algorithm configured to train and optimize cell classification. In some instances, different datasets are combined and shuffled to ensure that each training batch contains samples from different datasets. In some instances, the dataset is divided into three distinct sets: 1) for training; 2) for validation; and 3) for testing. In some instances, the training dataset comprises 40% to 80% of the data, e.g., 50% to 75%, and includes 70% of the training data. In some instances, the validation dataset comprises 10% to 30% of the data, e.g., 15% to 25%, and includes 30% of the data used for validation. In some instances, the test dataset comprises 5% to 20% of the data, e.g., 7.5% to 15%, and includes 10% of the data used for testing.

[0103] In some instances, one or more of image data patterns, sequence data patterns, and tabular data patterns are processed in real time during the application of the training algorithm. In some instances, images are reconstructed in real time from raw data waveforms during the application of the training algorithm. In some instances, the memory includes instructions for image transformations prior to application to the neural network. In some instances, data from the light loss channel is inverted to create a dark-field image. In some instances, all photodetector channels are zero-padded. In some instances, the pixel values ​​of each photodetector channel are scaled to have zero mean and unit standard deviation.

[0104] In some embodiments, the training algorithm includes a supervised method using a weighted cross-entropy loss function. In some instances, the weights are calculated as the product of class weights and task weights, where the class weights correspond to the class abundance in the task and the task weights correspond to the task abundance across all datasets. In some instances, the Adam optimizer with a predetermined learning rate and no weight decay is used. In some instances, the predetermined learning rate ranges from 1e... -1 up to 1e -5 For example, from 1e -2 up to 1e -4 And including 1e -3 The learning rate. In some instances, the learning rate does not decay with weights. In some embodiments, the learning rate decreases by a predetermined factor if the validation loss remains constant, for example, if the validation loss remains constant over one or more time steps, such as two or more time steps, such as three or more time steps, such as five or more time steps, such as ten or more time steps, such as fifteen or more time steps, such as twenty or more time steps, such as twenty or five or more time steps, and including fifty or more time steps. In these embodiments, if the validation loss remains constant, the learning rate will decrease by a factor of 2 or more, such as a factor of 3 or more, such as a factor of 4 or more, such as a factor of 5 or more, such as a factor of 10 or more, and including twenty or more. In some embodiments, if the validation loss remains constant over 10 time steps, the learning rate will decrease by a factor of 10. The batch size for training can vary; in some instances, the batch size is 64 or more, such as 128 or more, such as 256 or more, and including 512 or more. In some instances, the batch size is 128. In other instances, the batch size is 512.

[0105] In some embodiments, the performance of the algorithm is evaluated. In some instances, the performance of the trained model is evaluated using validation or test datasets. In some instances, dimensionality reduction is applied to high-dimensional feature vectors to evaluate feature quality. In some instances, applying dimensionality reduction to high-dimensional feature vectors can examine how known groups are separated in a low-dimensional space. In some instances, the performance of the algorithm is measured using one or more of the precision, recall, and F1 score for each class. In some instances, receiver operating feature curves are generated for each class. In some instances, the area under the curve (AUC) is evaluated to assess the performance of the feature vector algorithm.

[0106] Figure 1BAn analytical workflow according to certain embodiments is described, which integrates image, spectral, and scattering information from the single cell being measured. The workflow includes a data acquisition phase in which the single cell is illuminated with a light source, and the light from the illuminated sample is measured in different photodetector channels. Figure 1B The data acquisition for raw data in the light loss detector channel (LL), forward scattering channel (FSC), side scattering channel (SSC), and fluorescence (UV1, R4) and imaging photodetector channels (ImgB3) is illustrated. The feature-engineered layers of the neural network include image or waveform encoders for analyzing raw sequence waveforms across different imaging modalities (light loss, side scattering, forward scattering, and different fluorescence channels). The spectral module provides analysis of the fluorescence photodetector channels as spectral tabular data, and the scattering module provides analysis of the scattering photodetector channels as scattering tabular data.

[0107] Feature vectors generated using the feature engineering layer are applied to different classifiers to generate biological insights (biological parameters of cells) through insight distillation. In some instances, the classifiers include a viability classifier to determine whether a cell is alive or dead. In other instances, the classifiers include a single-cell classifier to provide single-cell identification, distinguishing cells from fragments, two-celled cells, and three-celled cells. Further classifiers can be generated by integrating spectral tabular and scatter plot data, for example, using exploratory models (e.g., dimensionality reduction algorithms).

[0108] Table 1 summarizes the example datasets used for training. Figure 1B The image encoder of the feature engineering layer in the workflow shown.

[0109] Table 1

[0110]

[0111] Figure 1C An example of an unlabeled image encoder according to certain embodiments is shown. Feature vectors are concatenated with multiple classification heads, each dedicated to a different cell phenotype task. The image encoder, with its feature vector engineering layer, converts a multi-channel unlabeled image into feature vectors. The image encoder converts light loss images (LL), forward scatter images (FSC), and side scatter images (SSC) into feature vectors for constructing a classifier. The generated classification heads include single-cell, survival rate, whole blood, and T-cell activation. Figure 1D Examples of light loss images for different cell phenotypes used in the classification head are shown.

[0112] Figure 1E Example performance of a classifier in whole blood classification according to certain embodiments is described. Figure 1FThe performance of the classifier according to certain embodiments in T-cell activation determination is shown. The top rows show the results of various dimensionality reduction techniques applied to image feature vectors. Uniform manifold approximation and projection (UMAP) and T-distributed random nearest neighbor embedding (TSNE) plots distinguish different phenotypes, while principal component analysis (PCA) plots show overlapping phenotypes in T-cell activation determination, as PCA has poor performance in preserving local structure in high-dimensional spaces. The lower left plot shows the confusion matrix of the classification results, from which metrics such as precision, recall, and F1 score can be derived. These results are summarized in the lower right table, showing that all scores exceed 0.90, highlighting the effectiveness of the multi-head classifier.

[0113] Figure 1G The receiver operating characteristic (ROC) curves for the T-cell activation dataset according to certain embodiments are shown. The area under the curve (AUC) of the classifier was evaluated as 0.977, which is significantly higher than the value of random guessing and close to the theoretical optimum of 1. Table 2 summarizes the performance examples of different classifications according to certain embodiments.

[0114] Table 2

[0115]

[0116] In some embodiments, the memory includes instructions for determining one or more sorting gates for sample cells, for example by assigning each cell to a particle swarm cluster. In some instances, the memory includes instructions for generating one or more sorting gates that capture cells of a target particle swarm cluster and exclude cells of non-target particle swarm clusters. In some instances, the memory includes instructions for determining sorting gates configured to maximize the inclusion rate of cells of the target particle swarm cluster (e.g., cells of a specific classification), for example, sorting gates configured to achieve an inclusion rate of 50% or higher, such as 55% or higher, such as 60% or higher, such as 65% or higher, such as 70% or higher, such as 75% or higher, such as 80% or higher, such as 85% or higher, such as 90% or higher, such as 95% or higher, such as 97% or higher, such as 99% or higher, and includes determining a sorting gate configured to achieve an inclusion rate of 99.9% or higher for cells of the target particle swarm cluster. In some instances, the memory includes instructions for determining sorting gates configured to maximize the purity yield of cells in a target particle swarm. For example, the sorting gates are configured to achieve a purity yield of 50% or higher, such as 55% or higher, 60% or higher, 65% or higher, 70% or higher, 75% or higher, 80% or higher, 85% or higher, 90% or higher, 95% or higher, 97% or higher, 99% or higher, and includes determining a sorting gate configured to achieve a purity yield of 99.9% or higher for the cells in the target particle swarm.

[0117] In some instances, the memory includes instructions for determining sorting gates configured to exclude cells from non-target particle swarms to the maximum extent possible. In some instances, the memory includes instructions for determining sorting gates configured to exclude 50% or more of cells from non-target particle swarms, such as 55% or more, 60% or more, 65% or more, 70% or more, 75% or more, 80% or more, 85% or more, 90% or more, 95% or more, 97% or more, 99% or more, and includes determining a sorting gate configured to exclude 99.9% or more of cells from non-target particle swarms.

[0118] In some instances, the memory includes instructions for evaluating sorting gates of a gating strategy and adjusting one or more generated sorting gates. This adjustment can be based on a metric that indicates the accuracy of the sorting gates according to the desired sorting strategy. In some instances, this metric can be an accuracy (e.g., purity) metric, where purity is compared to a predetermined threshold. This metric can be generated based on the confidence level of a classifier included in the sorting strategy. In other instances, the metric can be a yield metric, where the yield of target cells in a particle swarm is compared to a predetermined threshold.

[0119] In some embodiments, the memory includes instructions for generating sorting decisions based on sorting gates determined for the cells in the sample. In some instances, the memory includes instructions for generating particle sorting decisions using a gating strategy with eight or fewer sorting gates (e.g., seven or fewer, six or fewer, five or fewer, four or fewer, three or fewer, two or fewer). In some embodiments, the memory includes instructions for generating sorting gates using a graphical display that shows one or more analysis algorithms for applying classification parameters to image parameters determined for the cells.

[0120] In some embodiments, the memory includes instructions for generating sorting gates using a graphical display that shows one or more analysis algorithms for applying determined classification parameters to image parameters determined for cells. In some embodiments, the system includes a display for visualizing the gating strategy on a graphical user interface.

[0121] In some instances, the graphical user interface applies analytical algorithms to generate gating policies. For example, the analytical algorithm may be one or more of a spectral compensation matrix, a clustering algorithm, and a t-distributed random nearest neighbor embedding (t-SNE) algorithm. In some instances, the analytical algorithm is applied to a particle swarm cluster by dragging its icon onto the cluster. In other instances, the particle swarm cluster is selected from a drop-down menu, and the analytical algorithm is applied. In some instances, the analytical algorithm is a spectral unmixing algorithm, such as the algorithm described in U.S. Patent No. 11,009,400 and International Patent Application No. PCT / US2021 / 46741, filed August 19, 2021, the disclosure of which is incorporated herein by reference.

[0122] In some instances, the system includes memory having instructions for determining a gating strategy using a computational sorting algorithm, such as that described in U.S. Patent No. 11,513,054, the disclosure of which is incorporated herein by reference. In some instances, the system includes memory having computer software for determining the sorting gate, such as HyperFinder (e.g., as described by Bonavia et al. in “Frontiers in Immunology” 2022; 13:1007016) and computational sorting using HyperFinder, FlowJo software, and BD FACSDiva software (Becton Dickinson, 2021), the disclosure of which is incorporated herein by reference. In some embodiments, the gating strategy is developed on a separate computational system (e.g., a different computer system or network) and transmitted to a flow cytometer to implement the gating strategy, such as a particle sorter using flow cytometry (e.g., having a sorting decision module).

[0123] In some embodiments, the system for generating the gating strategy is part of, or operatively coupled to, a particle analyzer system (e.g., a flow cytometer) for generating the flow cytometry data described herein.

[0124] In some instances, the system includes an integrated circuit device programmed to implement one or more of the methods described herein. In some embodiments, the integrated circuit device of interest includes a field-programmable gate array (FPGA). In other embodiments, the integrated circuit device includes an application-specific integrated circuit (ASIC). In still other embodiments, the integrated circuit device includes a complex programmable logic device (CPLD).

[0125] In some embodiments, the system includes a particle sorter component. In some embodiments, the sorting mechanism is configured to sort cells into containers based on the presence, classification, or both of cells in the sample. The term “sorting” is used herein in its conventional sense to mean separating components of a sample (e.g., target cells with mitochondrial morphology of interest, non-target cells with mitochondrial morphology of no interest, non-cellular particles such as biomacromolecules), and in some instances, delivering the separated components to one or more sample collection containers. For example, the system of this subject matter may be configured to sort samples having two or more components, such as three or more components, such as four or more components, such as five or more components, such as ten or more components, such as fifteen or more components, and includes sorting samples having 25 or more components. One or more sample components may be separated from the sample and transported to a sample collection container, such as two or more sample components, three or more sample components, four or more sample components, five or more sample components, ten or more sample components, and including 15 or more components that may be separated from the sample and transported to a sample collection container.

[0126] In some embodiments, the particle sorting system of interest is configured to sort particles using enclosed particle sorting modules, such as those described in U.S. Patent Publication No. 2017 / 0299493, filed March 28, 2017, the disclosure of which is incorporated herein by reference. In some embodiments, a sorting decision module having multiple sorting decision units is used to sort particles (e.g., cells) in a sample, such as those described in U.S. Patent Publication No. 2020 / 0256781, the disclosure of which is incorporated herein by reference. In some embodiments, the system of interest includes a particle sorting module with deflection plates, such as those described in U.S. Patent Publication No. 2017 / 0299493, filed March 28, 2017, the disclosure of which is incorporated herein by reference.

[0127] The system according to some embodiments may include a display and an operator input device. For example, the operator input device may be a keyboard, mouse, etc. The processing module includes a processor that can access memory on which instructions for performing steps of the methods of this subject are stored. The processing module may include an operating system, a graphical user interface (GUI) controller, system memory, memory storage devices and input / output controllers, a cache, a data backup unit, and many other devices. The processor may be a commercially available processor or one of other processors that are already available or not. The processor executes the operating system, which interacts with firmware and hardware in well-known ways and assists the processor in coordinating and executing the functions of various computer programs, which may be written in various programming languages, such as Java, Perl, C++, other high-level or low-level languages ​​and combinations thereof, as well as are well known in the art. The operating system typically works with the processor to coordinate and execute the functions of other computer components. The operating system also provides scheduling, input / output control, file and data management, memory management, and communication control and related services according to known techniques. The processor may be any suitable analog or digital system. In some embodiments, the processor includes analog electronics that provide feedback control (e.g., negative feedback control).

[0128] System memory can be any of a variety of known or future memory storage devices. Examples include any common random access memory (RAM), magnetic media (such as fixed hard disks or magnetic tapes), optical media (such as optical discs), flash memory devices, or other memory storage devices. Memory storage devices can be any of a variety of known or future devices, including optical disc drives, magnetic tape drives, removable hard disk drives, or floppy disk drives. This type of memory storage device typically reads from and / or writes to program storage media (not shown), such as optical discs, magnetic tapes, removable hard disks, or floppy disks. Any of these program storage media, or other media currently in use or that may be developed in the future, can be considered a computer program product. As will be understood, these program storage media typically store computer software programs and / or data. Computer software programs, also known as computer control logic, are typically stored in system memory and / or in program storage devices used in conjunction with memory storage devices.

[0129] In some embodiments, a computer program product is described, comprising a computer-usable medium having control logic (computer software program, including program code) stored thereon. When the control logic is executed by a computer's processor, the processor performs the functions described herein. In other embodiments, some functions are implemented primarily in hardware, for example using a hardware state machine. Implementing a hardware state machine to perform the functions described herein will be apparent to those skilled in the art.

[0130] The memory can be any suitable device in which the processor can store and retrieve data, such as magnetic, optical, or solid-state storage devices (including disks, optical discs, magnetic tapes, RAM, or any other suitable fixed or portable devices). The processor can include a general-purpose digital microprocessor, which is appropriately programmed by a computer-readable medium carrying the necessary program code. The programming can be provided to the processor remotely via a communication channel, or the programming can be pre-stored in a computer program product (e.g., memory or a portion of other portable or fixed computer-readable storage media) using any device associated with the memory. For example, a disk or optical disc can carry the programming and can be read by a disk writer / reader. The system of the present invention also includes programming, for example, in the form of a computer program product, an algorithm for practicing the methods described above. The programming according to the present invention can be recorded on a computer-readable medium, for example, any medium that can be directly read and accessed by a computer. These media include, but are not limited to: magnetic storage media, such as floppy disks, hard disk storage media, and magnetic tape; optical storage media, such as CD-ROMs; electrical storage media, such as RAM and ROM; portable flash drives; and mixtures of these classes, such as magnetic / optical storage media.

[0131] The processor can also access communication channels to communicate with users in remote locations. A remote location refers to a user who does not directly interact with the system but relays input information from an external device to the input manager. This external device could be a computer connected to a wide area network (“WAN”), telephone network, satellite network, or any other suitable communication channel, including a mobile phone (i.e., a smartphone).

[0132] In some embodiments, the system according to this disclosure may be configured to include a communication interface. In some embodiments, the communication interface includes a receiver and / or transmitter for communicating with a network and / or another device. The communication interface can be configured for wired or wireless communication, including but not limited to radio frequency (RF) communication (e.g., RFID, Zigbee communication protocol, WiFi, infrared, wireless universal serial bus (USB), ultra-wideband (UWB)). Communication protocols and cellular communications, such as Code Division Multiple Access (CDMA) or Global System for Mobile Communications (GSM).

[0133] In one embodiment, the communication interface is configured to include one or more communication ports, such as physical ports or interfaces, such as USB ports, RS-232 ports, or any other suitable electrical connection ports, to allow data communication between the system and other external devices (e.g., computer terminals in a doctor's office or hospital environment) configured for similar complementary data communication.

[0134] In one embodiment, the communication interface is configured for infrared communication. Communication or any other suitable wireless communication protocol to enable the system of this subject to communicate with other devices (such as computer terminals and / or networks, communication-enabled mobile phones, personal digital assistants, or any other communication devices that the user can use in conjunction with them).

[0135] In one embodiment, the communication interface is configured to provide data transmission connectivity via Internet Protocol (IP), through mobile networks, short message service (SMS), wireless connection to a personal computer (PC) on a local area network (LAN) connected to the Internet, or WiFi connection to the Internet at a WiFi hotspot.

[0136] In one embodiment, the system is configured to communicate wirelessly with a server device via a communication interface, such as using 802.11 or... A common standard for RF or IrDA infrared protocols. The server device can be another portable device, such as a smartphone, personal digital assistant (PDA), or laptop; or a larger device, such as a desktop computer, appliance, etc. In some embodiments, the server device has a display, such as a liquid crystal display (LCD), and input devices, such as buttons, a keyboard, a mouse, or a touchscreen.

[0137] In some embodiments, the communication interface is configured to automatically or semi-automatically transmit data stored in the subject system, such as data stored in an optional data storage unit, to a network or server device using one or more of the communication protocols and / or mechanisms described above.

[0138] The output controller may include any of a variety of known display devices for presenting information to a user (whether human or machine, local or remote). If one of the display devices provides visual information, that information may typically be logically and / or physically organized as an array of image elements. The graphical user interface (GUI) controller may include any of a variety of known or future software programs for providing a graphical input and output interface between the system and the user, and for processing user input. Functional elements of the computer may communicate with each other via a system bus. In alternative embodiments, some of these communications may be accomplished using a network or other types of remote communication. The output manager may also provide information generated by the processing module to a user at a remote location, according to known technologies, such as via the Internet, telephone, or satellite networks. The presentation of data by the output manager may be implemented according to a variety of known technologies. As some examples, the data may include SQL, HTML, or XML documents, emails or other files, or other forms of data. The data may include Internet URLs that allow the user to retrieve other SQL, HTML, XML, or other documents or data from a remote source. One or more platforms present in the subject system may be any type of known computer platform or type to be developed in the future, although they typically belong to the class of computers commonly referred to as servers. However, they can also be mainframes, workstations, or other computer types. They can be connected via any known or future type of cable or other communication system (including wireless systems, whether networked or otherwise). They can be located in the same place or physically separated. Various operating systems can be used on any computer platform, depending on the type and / or architecture of the chosen platform. Suitable operating systems include Windows 10, Windows... Windows XP, Windows 7, Windows 8, iOS, Sun Solaris, Linux, OS / 400, Compaq Tru64Unix, SGI IRIX, Siemens Reliant Unix, Ubuntu, Zorin OS, etc.

[0139] In some embodiments, the subject system includes one or more optical adjustment components for adjusting light, such as light incident on a sample (e.g., from a laser) or light collected from a sample (e.g., scattered light, fluorescence). For example, optical adjustment can be increasing the size of the light, focusing the light, or collimating the light. In some instances, optical adjustment is an amplification scheme to increase the size of the light (e.g., beam spot), such as increasing the size by 5% or more, such as increasing by 10% or more, such as increasing by 25% or more, such as increasing by 50% or more, and including increasing the size by 75% or more. In other embodiments, optical adjustment includes focusing the light to reduce the size of the light, such as reducing by 5% or more, such as reducing by 10% or more, such as reducing by 25% or more, such as reducing by 50% or more, and including reducing the beam spot size by 75% or more. In some embodiments, optical adjustment includes collimating the light. The term "collimating" is used in its conventional sense to refer to optically adjusting the collinearity of light propagation or reducing the divergence of light from a common propagation axis. In some instances, collimating includes narrowing the spatial cross-section of the beam (e.g., reducing the beam profile of a laser).

[0140] In some embodiments, the optical adjustment component is a focusing lens with a magnification of 0.1 to 0.95, such as 0.2 to 0.9, 0.3 to 0.85, 0.35 to 0.8, 0.5 to 0.75, and including 0.55 to 0.7, such as 0.6. For example, in some instances, the focusing lens is a biachromatic reduction lens with a magnification of approximately 0.6. The focal length of the focusing lens can vary, ranging from 5 mm to 20 mm, such as 6 mm to 19 mm, 7 mm to 18 mm, 8 mm to 17 mm, 9 mm to 16 mm, and including focal lengths ranging from 10 mm to 15 mm. In some embodiments, the focal length of the focusing lens is approximately 13 mm.

[0141] In other embodiments, the optical adjustment component is a collimator. The collimator can be any suitable collimation scheme, such as one or more mirrors or curved lenses, or combinations thereof. For example, in some instances, the collimator is a single collimating lens. In other instances, the collimator is a collimating mirror. In other instances, the collimator includes two lenses. In other instances, the collimator includes a mirror and a lens. In examples where the collimator includes one or more lenses, the focal length of the collimating lens can vary from 5 mm to 40 mm, for example, 6 mm to 37.5 mm, for example, 7 mm to 35 mm, for example, 8 mm to 32.5 mm, for example, 9 mm to 30 mm, for example, 10 mm to 27.5 mm, for example, 12.5 mm to 25 mm, and includes focal lengths ranging from 15 mm to 20 mm.

[0142] In some embodiments, the subject matter system includes a flow cell nozzle having a nozzle orifice configured to allow a flow stream to pass through the flow cell nozzle. The subject matter flow cell nozzle has an orifice that propagates a fluid sample to a sample interrogation region, wherein, in some embodiments, the flow cell nozzle includes a proximal cylindrical portion defining a longitudinal axis and a distal truncated conical portion terminating on a flat surface and having a nozzle orifice transverse to the longitudinal axis on the flat surface. The length of the proximal cylindrical portion (measured along the longitudinal axis) can vary from 1 mm to 15 mm, for example from 1.5 mm to 12.5 mm, for example from 2 mm to 10 mm, for example from 3 mm to 9 mm, including from 4 mm to 8 mm. The length of the distal truncated conical portion (measured along the longitudinal axis) can also vary from 1 mm to 10 mm, for example from 2 mm to 9 mm, for example from 3 mm to 8 mm, including from 4 mm to 7 mm. In some embodiments, the diameter of the flow pool nozzle chamber can vary, ranging from 1 mm to 10 mm, for example from 2 mm to 9 mm, for example from 3 mm to 8 mm, and including from 4 mm to 7 mm.

[0143] In some instances, the nozzle chamber does not include a cylindrical portion, and the entire flow cell nozzle chamber is a truncated cone. In these embodiments, the length of the truncated cone nozzle chamber (measured along the longitudinal axis transverse to the nozzle orifice) can range from 1 mm to 15 mm, for example 1.5 mm to 12.5 mm, for example 2 mm to 10 mm, for example 3 mm to 9 mm, including 4 mm to 8 mm. The diameter of the proximal portion of the truncated cone nozzle chamber can range from 1 mm to 10 mm, for example 2 mm to 9 mm, for example 3 mm to 8 mm, and including 4 mm to 7 mm.

[0144] In an embodiment, the sample flow exits from an orifice at the distal end of the flow cell nozzle. Depending on the desired characteristics of the flow, the flow cell nozzle orifice can be of any suitable shape, wherein the cross-sectional shape of interest includes, but is not limited to: linear cross-sectional shapes (e.g., square, rectangle, trapezoid, triangle, hexagon, etc.), curved cross-sectional shapes (e.g., circular, elliptical), and irregular shapes (e.g., a parabolic bottom portion coupled to a flat top portion). In some embodiments, the flow cell nozzle of interest has a circular orifice. The nozzle orifice size can vary, and in some embodiments, ranges from 1 μm to 20000 μm, for example, 2 μm to 17500 μm, for example, 5 μm to 15000 μm, for example, 10 μm to 12500 μm, for example, 15 μm to 10000 μm, for example, 25 μm to 7500 μm, for example, 50 μm to 5000 μm, for example, 75 μm to 1000 μm, for example, 100 μm to 750 μm, and includes 150 μm to 500 μm. In some embodiments, the nozzle orifice is 100 μm.

[0145] In some embodiments, the flow cell nozzle includes a sample injection port configured to provide a sample to the flow cell nozzle. In embodiments, the sample injection system is configured to provide a suitable sample flow to the flow cell nozzle chamber. Depending on the desired characteristics of the flow flow, the sample rate delivered to the flow cell nozzle chamber through the sample injection port can be 1 μL / sec or higher, for example, 2 μL / sec or higher, for example, 3 μL / sec or higher, for example, 5 μL / sec or higher, for example, 10 μL / sec or higher, for example, 15 μL / sec or higher, for example, 25 μL / sec or higher, for example, 50 μL / sec or higher, for example, 100 μL / sec or higher, for example, 150 μL / sec or higher, for example, 200 μL / sec or higher, for example, 250 μL / sec or higher, for example, 300 μL / sec or higher, for example, 350 μL / sec or higher, for example, 400 μL / sec or higher, for example, 450 μL / sec or higher, and includes 500 μL / sec or higher. For example, the sample flow rate can be from 1 μL / sec to about 500 μL / sec, for example from 2 μL / sec to about 450 μL / sec, for example from 3 μL / sec to about 400 μL / sec, for example from 4 μL / sec to about 350 μL / sec, for example from 5 μL / sec to about 300 μL / sec, for example from 6 μL / sec to about 250 μL / sec, for example from 7 μL / sec to about 200 μL / sec, for example from 8 μL / sec to about 150 μL / sec, for example from 9 μL / sec to about 125 μL / sec, and includes 10 μL / sec to about 100 μL / sec.

[0146] The sample injection port can be an orifice located in the wall of the nozzle chamber or a conduit located near the proximal end of the nozzle chamber. When the sample injection port is an orifice located in the wall of the nozzle chamber, the orifice can be of any suitable shape, wherein the cross-sectional shapes of interest include, but are not limited to: linear cross-sectional shapes (e.g., square, rectangle, trapezoid, triangle, hexagon, etc.), curved cross-sectional shapes (e.g., circle, ellipse, etc.), and irregular shapes (e.g., parabolic bottom portion coupled to a flat top portion). In some embodiments, the sample injection port has a circular orifice. The size of the sample injection port orifice can vary depending on the shape, and in some instances, its opening ranges from 0.1 mm to 5.0 mm, for example 0.2 mm to 3.0 mm, for example 0.5 mm to 2.5 mm, for example 0.75 mm to 2.25 mm, for example 1 mm to 2 mm, including 1.25 mm to 1.75 mm, for example 1.5 mm.

[0147] In some instances, the sample injection port is a conduit positioned proximal to the flow cell nozzle chamber. For example, the sample injection port can be a conduit positioned such that its orifice aligns with the flow cell nozzle orifice. When the sample injection port is a conduit positioned to align with the flow cell nozzle orifice, the cross-sectional shape of the sample injection tube can be any suitable shape, wherein the cross-sectional shapes of interest include, but are not limited to: straight cross-sectional shapes (e.g., square, rectangle, trapezoid, triangle, hexagon, etc.), curved cross-sectional shapes (e.g., circular, elliptical), and irregular shapes (e.g., parabolic bottom portion coupled to a flat top portion). The orifice of the conduit can vary in shape, and in some instances, its opening ranges from 0.1 mm to 5.0 mm, for example 0.2 mm to 3.0 mm, for example 0.5 mm to 2.5 mm, for example 0.75 mm to 2.25 mm, for example 1 mm to 2 mm, and includes 1.25 mm to 1.75 mm, for example 1.5 mm. The shape of the tip of the sample injection port can be the same as or different from the cross-sectional shape of the sample injection tube. For example, the orifice of the sample injection port may include a beveled tip with an angle ranging from 1° to 10°, such as 2° to 9°, such as 3° to 8°, such as 4° to 7°, and including a 5° angle.

[0148] In some embodiments, the flow cell nozzle further includes a sheath fluid injection port configured to supply sheath fluid to the flow cell nozzle. In embodiments, the sheath fluid injection system is configured to supply a sheath fluid flow to the flow cell nozzle chamber, for example, in conjunction with a sample to generate a sheath fluid laminar flow around a sample flow. Depending on the desired characteristics of the flow flow, the rate of sheath fluid delivered to the flow cell nozzle chamber can be 25 μL / sec or higher, for example 50 μL / sec or higher, for example 75 μL / sec or higher, for example 100 μL / sec or higher, for example 250 μL / sec or higher, for example 500 μL / sec or higher, for example 750 μL / sec or higher, for example 1000 μL / sec or higher, and includes 2500 μL / sec or higher. For example, the range of sheath fluid flow rate can be from 1 μL / sec to about 500 μL / sec, such as 2 μL / sec to about 450 μL / sec, such as 3 μL / sec to about 400 μL / sec, such as 4 μL / sec to about 350 μL / sec, such as 5 μL / sec to about 300 μL / sec, such as 6 μL / sec to about 250 μL / sec, such as 7 μL / sec to about 200 μL / sec, such as 8 μL / sec to about 150 μL / sec, such as 9 μL / sec to about 125 μL / sec, and includes 10 μL / sec to about 100 μL / sec.

[0149] In some embodiments, the sheath fluid injection port is an orifice located on the wall of the nozzle chamber. The sheath fluid injection port orifice can be of any suitable shape, wherein the cross-sectional shapes of interest include, but are not limited to: straight cross-sectional shapes (e.g., square, rectangle, trapezoid, triangle, hexagon, etc.), curved cross-sectional shapes (e.g., circular, elliptical), and irregular shapes (e.g., parabolic bottom portion coupled to a flat top portion). The size of the orifice of the sample injection port can vary depending on its shape, and in some instances, it has an opening ranging from 0.1 mm to 5.0 mm, for example 0.2 mm to 3.0 mm, for example 0.5 mm to 2.5 mm, for example 0.75 mm to 2.25 mm, for example 1 mm to 2 mm, and including 1.25 mm to 1.75 mm, for example 1.5 mm.

[0150] In some instances, the subject matter system includes a sample interrogation region in fluid communication with the flow cell nozzle orifice. In these instances, the sample flow exits from the orifice at the distal end of the flow cell nozzle, and particles in the flow can be illuminated by a light source at the sample interrogation region. The size of the interrogation region can vary depending on the characteristics of the flow nozzle (e.g., the size of the nozzle orifice and the size of the sample injection port). In embodiments, the width of the interrogation region can be 0.01 mm or greater, for example 0.05 mm or greater, for example 0.1 mm or greater, for example 0.5 mm or greater, for example 1 mm or greater, for example 2 mm or greater, for example 3 mm or greater, for example 5 mm or greater, and includes 10 mm or greater. The length of the query area can also vary. In some instances, the length can be 0.01 mm or longer, such as 0.1 mm or longer, such as 0.5 mm or longer, such as 1 mm or longer, such as 1.5 mm or longer, such as 2 mm or longer, such as 3 mm or longer, such as 5 mm or longer, such as 10 mm or longer, such as 15 mm or longer, such as 20 mm or longer, such as 25 mm or longer, and includes 50 mm or longer.

[0151] The interrogation region can be configured to facilitate illumination of a planar cross-section of the outflowing flow, or it can be configured to facilitate illumination of a diffuse field of a predetermined length (e.g., in a diffuse laser or lamp). In some embodiments, the interrogation region includes a transparent window that facilitates illumination of a predetermined length of the outflowing flow, such as 1 mm or longer, 2 mm or longer, 3 mm or longer, 4 mm or longer, 5 mm or longer, and including 10 mm or longer. Depending on the light source used to illuminate the outflowing flow (described below), the interrogation region can be configured to allow light of 100 nm to 1500 nm, such as 150 nm to 1400 nm, 200 nm to 1300 nm, 250 nm to 1200 nm, 300 nm to 1100 nm, 350 nm to 1000 nm, 400 nm to 900 nm, and including allowing light of 500 nm to 800 nm to pass through. Therefore, the query area can be formed of any transparent material that passes through the desired wavelength range, including but not limited to optical glass, borosilicate glass, Pyrex glass, ultraviolet quartz, infrared quartz, sapphire, and plastics such as polycarbonate, polyvinyl chloride (PVC), polyurethane, polyether, polyamide, polyimide, or copolymers of these thermoplastics, such as PETG (glycol-modified polyethylene terephthalate), and other polymeric plastic materials, including polyesters, wherein the polyester of interest may include, but is not limited to: poly(alkylene terephthalates), such as polyethylene terephthalate (PET), bottle-grade PET (made from monoethylene glycol, terephthalic acid, and substances such as isophthalic acid and cyclohexene), etc. Copolymers made from other comonomers such as dimethanol, polybutylene terephthalate (PBT) and polyhexamethylene terephthalate; polyalkylene adipate, such as polyethylene adipate, polybutanediol adipate and polyhexamethylene adipate; polyalkylene suberates, such as polyethylene suberate; polyalkylene sebacate, such as polyethylene sebacate; poly-ε-caprolactone and poly-β-propiolactone; polyalkylene isophthalate, such as polyethylene isophthalate; polyalkylene 2,6-naphthalene dicarboxylate, such as polyethylene 2,6-naphthalene dicarboxylate;Poly(alkylene sulfonyl-4,4'-dibenzoates), such as polyethylene sulfonyl-4,4'-dibenzoate; poly(p-phenylene alkylene dicarboxylates), such as polyethylene ethylene dicarboxylates; poly(trans-1,4-cyclohexanediyl alkylene dicarboxylates), such as poly(trans-1,4-cyclohexanediyl alkylene dicarboxylates. ethylenedicarboxylate); poly(1,4-cyclohexane-dimethylene alkylene dicarboxylates), such as poly(1,4-cyclohexane-dimethylene ethylene dicarboxylate); poly([2.2.2]-bicyclooctane-1,4-dimethylenealkylene dicarboxylates), such as poly([2.2.2]-bicyclooctane-1,4-dimethylene ethylene dicarboxylate). dicarboxylate; lactic acid polymers and copolymers, such as (S)-polylactide, (R,S)-polylactide, poly(tetramethylglycolide) and poly(lactide-co-glycolide); and bisphenol A type polycarbonate, 3,3'-dimethylbisphenol A polycarbonate, 3,3',5,5'-tetrachlorobisphenol A polycarbonate (3,3′-dimethylbisphenol A), 3,3',5,5'-tetramethylbisphenol A polycarbonate (3,3′-thylbisphenethylbisphenol A);Polyamides, such as poly(p-phenylene terephthalamide); polyesters, such as polyethylene terephthalate, such as Mylar; TM Polyethylene terephthalate, etc. In some embodiments, the system includes a cuvette located in the sample interrogation area. In embodiments, the cuvette allows light of 100 nm to 1500 nm, such as 150 nm to 1400 nm, such as 200 nm to 1300 nm, such as 250 nm to 1200 nm, such as 300 nm to 1100 nm, such as 350 nm to 1000 nm, such as 400 nm to 900 nm, to pass through, and includes allowing light of 500 nm to 800 nm to pass through.

[0152] In some embodiments, the subject system is a flow cytometry system that includes a photodiode and an amplifier assembly as part of a light detection system for detecting light emitted from a sample in a flowing stream. Suitable flow cytometry systems may include, but are not limited to, those described in the following: “Ormerod (ed.), FlowCytometry: A Practical Approach, Oxford Univ. Press (1997)”; “Jaroszeski et al. (ed.), Flow Cytometry Protocols, Methods in Molecular Biology No. 91, Humana Press (1997)”; “Practical Flow Cytometry, 3rd Edition, Wiley-Liss (1995)”; “Virgo et al. (2012) Ann Clin Biochem. Jan; 49(pt1):17-28”; “Linden et al., Semin Throm Hemost. 2004 Oct; 30(5):502-11”; “Alison et al. J Pathol, 2010 Dec; 222(4):335-344”; and “Herbig et al. (2007) Crit Rev Ther Drug Carrier”. Syst. 24(3):203-255”; its publication is incorporated herein by reference. In some instances, flow cytometry systems of interest include BD Biosciences FACSCanto TM Flow cytometer, BD Biosciences FACSCanto TM II flow cytometer, BD Accuri TM Flow cytometer, BD Accuri TM C6Plus flow cytometer, BD Biosciences FACSCelestaTM Flow cytometer, BD Biosciences FACSLyric TM Flow cytometer, BD Biosciences FACSVerse TM Flow cytometer, BD Biosciences FACSymphony TM Flow cytometer, BD Biosciences LSRFortessa TM Flow cytometer, BD Biosciences LSL Fortessa TM X-20 flow cytometer, BD Biosciences FACSPresto TM Flow cytometry, BDBiosciences FACSVia TM Flow cytometer and BD Biosciences FACSCalibur TM Cell sorting instrument, BDBiosciences FACSCount TM Cell sorter, BD Biosciences FACSLyric TM Cell sorting instrument, BDBiosciences Via TM Cell sorter, BD Biosciences Influx TM Cell sorter, BD Biosciences Jazz TM Cell sorter, BD Biosciences Aria TM Cell sorting instrument, BD Biosciences FACSAria TM II Cell Sorter, BD Biosciences FACSAria TM III Cell Sorter, BD Biosciences FACSAria TM Fusion Cell Sorter and BD Biosciences FACSMelody TM Cell sorter, BD Biosciences FACSymphony TM S6 cell sorter, etc.

[0153] In some embodiments, the subject matter system is a flow cytometry system, such as U.S. Patent Nos. 10,663,476, 10,620,111, 10,613,017, 10,605,713, 10,585,031, 10,578,542, 10,578,469, and 10,481,074. Numbers: 10,302,545, 10,145,793, 10,113,967, 10,006,852, 9,952,076, 9,933,341, 9,726,527, 9,453,789, 9,200,334, 9,097,640, 9,09 No. 5,494, No. 9,092,034, No. 8,975,595, No. 8,753,573, No. 8,233,146, No. 8,140,300, No. 7,544,326, No. 7,201,875, No. 7,129,505, No. 6,821,740, No. 6,813,017, No. 6, The entire disclosure of the flow cytometry systems described in Nos. 809,804, 6,372,506, 5,700,692, 5,643,796, 5,627,040, 5,620,842, 5,602,039, 4,987,086, and 4,498,766 is incorporated herein by reference.

[0154] In some embodiments, the flow cytometer is configured as an imaging flow cytometer. For example, in some instances, the system of this subject is a flow cytometry system configured to image particles in a flowing stream using fluorescence imaging with radio frequency labeled emission (FIRE), as described, for example, in: "Diebold et al., Nature Photonics..." Vol. 7(10); 806-810(2013)”, and U.S. Patent Nos. 9,423,353, 9,784,661, 9,983,132, 10,006,852, 10,036,699, 10,078,045, 10,222,316, 10,288,546, 10,324,019, 10,408,758, 10,451,538, 10,620,111, and 10,684 The public information of Nos. 211, 10,845,295, 10,935,482, 10,935,485, 11,105,728, 11,280,718, 11,327,016, 11,366,052, 11,371,937, 11,692,926, 11,630,053, 11,774,343, 11,940,369, and 11,946,851 is incorporated herein by reference.

[0155] Figure 2 A flow cytometry system 200 according to an illustrative embodiment of the present disclosure is shown. System 200 includes a laser 201 configured to irradiate particles 211 in a flow stream 214 at an interrogation point 215 within a flow cell 210. Although Figure 2 The example shows a single laser, but it should be understood that multiple lasers can also be used. The laser beam from laser 201 is directed to focusing lens 202, which focuses the beam onto the flow stream portion containing sample particles 211 within flow cell 210. Flow cell 210 is part of a fluid system that directs particles (typically one at a time) in the flow stream toward the focused laser beam for interrogation. Alternatively, in the case of a flow cytometer that is a flow air cytometer, a nozzle tip can be used.

[0156] like Figure 2As shown, flow cell 210 is fluidly connected to a sheath fluid reservoir 203 containing sheath fluid and a sample fluid reservoir 204 containing sample fluid. Sheath fluid from sheath fluid reservoir 203 is supplied to at least one sheath fluid injection port 208 via a conduit (i.e., sheath fluid line) 207. Furthermore, sample fluid containing particles 211 from sample fluid reservoir 204 is supplied to sample injection port 206 via a conduit (i.e., sample fluid line) 205. Sample injection port 206 is fluidly connected to a sample injector 213 (e.g., a sample injection needle) configured to introduce particles 211 into the interior of flow cell 210. Particles 211 are hydrodynamically focused by the sheath fluid entering from sheath fluid injection port 208, such that a flow stream 214 is formed downstream of the conical portion 212 of flow cell 210. Particles emitted at the distal end of flow cell 210 can be processed and / or collected by any suitable method. For example, depending on the type of flow cytometry being performed, particles can be collected at the distal end of the flow cytometer 210, for example, via a waste line. Alternatively, the particles can be sorted.

[0157] Light from one or more laser beams interacts with particles 211 in the sample through diffraction, refraction, reflection, scattering, and absorption, and is re-emitted at various wavelengths depending on the characteristics of the particles (e.g., size, internal structure, and the presence of one or more fluorescent molecules attached to or naturally present on the particles). Fluorescence emission, as well as diffracted, refracted, reflected, and scattered light, can be directed to one or more detectors. Specifically, forward-scattered light (FSC) is directed to a forward-scattering photodetector 223. The forward-scattering photodetector 223 is positioned slightly off-axis from the direct beam passing through the flow cell 210 and is configured to detect diffracted light, i.e., excitation light that propagates primarily forward through or around the particles. The intensity of the light detected by the forward-scattering photodetector 223 depends on the overall size of the particles. The forward-scattering detector can include, for example, a photodiode. Between the forward-scattering photodetectors 223 are a filter 221a and a scattering strip 222. Filter 221a can be configured to filter out non-FSC light of at least one wavelength, while scattering bar 222 can be configured to prevent the incident beam (i.e. non-scattered light) from laser 201 from being detected by forward scattering detector 223.

[0158] Furthermore, side-scattered light (SSC) is detected by side-scattered light detector 224. In other words, side-scattered light detector 224 is configured to detect refracted and reflected light from the surface and internal structure of particle 211, which tends to increase with the complexity of the particle structure. Figure 2In the example, the flow cytometer 200 includes a dichroic mirror 220a configured to reflect SSC light to a side-scatter light detector 224 while allowing non-SSC (e.g., fluorescence) light to pass through. A filter 221b is configured to prevent non-SSC light of at least one wavelength from being detected by the side-scatter light detector 224. Fluorescence detectors 225a to 225c are also shown, each configured to detect fluorescence of a different wavelength. For example, the dichroic mirror 220b may be configured to reflect fluorescence (FL) corresponding to a first wavelength (or wavelength range) to the fluorescence detector 225a while allowing light of other wavelengths to pass through. The filter 221c may be configured to prevent light of at least one wavelength not corresponding to the first wavelength (or wavelength range) from being detected by the fluorescence detector 225a. Similarly, the dichroic mirror 220c is configured to reflect FL light corresponding to a second wavelength (or wavelength range) to the fluorescence detector 225b while allowing light of a third wavelength (or wavelength range) to pass through for detection by the fluorescence detector 225c. Filter 221d is configured to prevent light of at least one wavelength that does not correspond to the second wavelength (or wavelength range) from being detected by fluorescence detector 225b. Furthermore, filter 221e is configured to prevent light of at least one wavelength that does not correspond to the third wavelength (or wavelength range) from being detected by fluorescence detector 225c.

[0159] Those skilled in the art will recognize that the flow cytometers according to embodiments of this disclosure are not limited to... Figure 2 The flow cytometer shown is not limited to any flow cytometer known in the art. For example, a flow cytometer can have any number of lasers, beam splitters, filters, and detectors with various wavelengths and different configurations. For example, although... Figure 2 The embodiment shown has three fluorescence detectors for illustrative purposes, but it should be understood that any suitable number of fluorescence detectors can be used.

[0160] During operation, the cytometer is controlled by controller / processor 290, and measurement data from the detector can be stored in memory 295 and processed by controller / processor 290. Although not explicitly shown, controller / processor 290 is coupled to the detector to receive its output signal and can also be coupled to the electrical and electromechanical components of the flow cytometer to control laser 201, fluid flow parameters, etc. Input / output (I / O) capability 297 may also be provided in the system. Memory 295, controller / processor 290, and I / O 297 may be provided entirely as integrated parts of the flow cytometer. In this embodiment, a display may also form part of I / O capability 297 for presenting experimental data to the user of cytometer 200. Alternatively, memory 295 and controller / processor 290, as well as some or all of the I / O capability, may be part of one or more external devices (e.g., a general-purpose computer). In some embodiments, memory 295 and controller / processor 290 may be capable of wireless or wired communication with cytometer 210. The controller / processor 290, in conjunction with the memory 295 and I / O 297, can be configured to perform various functions related to the preparation and analysis of flow cytometry experiments.

[0161] In some embodiments, the system is an image-enabled particle sorter that uses frequency-coded data, such as Figure 3AAs shown. The particle sorter 300 includes an illumination assembly 300a, which includes a light source 301 (e.g., a 488nm laser) that generates an output beam 301a, which is split into beams 302a and 302b by a beam splitter 302. Beam 302a is propagated through an acousto-optic device (e.g., an acousto-optic deflector, AOD) 303 to generate an output beam 303a with one or more angle deflections. In some instances, the output beam 303a generated by the acousto-optic device 303 includes a local oscillator beam and multiple radio frequency comb beams. Beam 302b is propagated through an acousto-optic device (e.g., an acousto-optic deflector, AOD) 304 to generate an output beam 304a with one or more angle deflections. In some instances, the output beam 304a generated by the acousto-optic device 304 includes a local oscillator beam and multiple radio frequency comb beams. Output beams 303a and 304a, generated by acousto-optic devices 303 and 304 respectively, are combined with beam splitter 305 to generate output beam 305a, which is transmitted through optical component 306 (e.g., objective lens) to illuminate particles in flow cell 307. In some embodiments, acousto-optic device 303 (AOD) splits a single laser beam into an array of sub-beams, each sub-beam having a different optical frequency and angle. A second AOD 304 adjusts the optical frequency of a reference beam and then overlaps it with the sub-beam array at beam combiner 305. In some embodiments, the light illumination system having a light source and acousto-optic devices may also include those described in Schraivogel et al., “High-speed fluorescence image-enabled cellsorting”, Science (2022), 375(6578):315-320, and U.S. Patent Publication No. 2021 / 0404943, the disclosure of which is incorporated herein by reference.

[0162] Output beam 305a irradiates sample particles 308 propagating through flow cell 307 (e.g., with sheath 309) at irradiation region 310. As shown in irradiation region 310, multiple beams (e.g., angularly deflected radio frequency offset beams depicted as dots on irradiation region 310) overlap with a reference local oscillator beam (depicted as shaded lines on irradiation region 310). Due to their different optical frequencies, the overlapping beams exhibit beat frequency behavior, causing each sub-beam to carry a different frequency f. 1-n Sine modulation.

[0163] Light from the illuminated sample is transmitted to a light detection system 300b comprising multiple photodetectors. The light detection system 300b includes a forward-scattering photodetector 311 for generating a forward-scattering image 311a and a side-scattering photodetector 312 for generating a side-scattering image 312a. The light detection system 300b also includes a bright-field photodetector 313 for generating a light loss image 313a. In some embodiments, the forward-scattering detector 311 and the side-scattering detector 312 are photodiodes (e.g., avalanche photodiodes, APDs). In some instances, the bright-field photodetector 313 is a photomultiplier tube (PMT). Fluorescence from the illuminated sample is also detected using fluorescence photodetectors 314 to 317. In some instances, photodetectors 314 to 317 are photomultiplier tubes. Light from the illuminated sample is directed by a beam splitter 320 to the side-scattering detection channel 312 and the fluorescence detection channels 314 to 317. The optical detection system 300b includes bandpass optical components 321, 322, 323, and 324 (e.g., dichroic mirrors) for propagating light of a predetermined wavelength to photodetectors 314 to 317. In some instances, optical component 321 has a 534 nm / 40 nm bandpass. In some instances, optical component 322 has a 586 nm / 42 nm bandpass. In some instances, optical component 323 has a 700 nm / 54 nm bandpass. In some instances, optical component 324 has a 783 nm / 56 nm bandpass. The first number indicates the center of the spectral band. The second number provides the range of the spectral band. Therefore, the 510 / 20 filter extends 10 nm on each side of the center of the spectral band, i.e., from 500 nm to 520 nm.

[0164] Data signals generated in response to light detected in scattered light detection channels 311 and 312, bright field light detection channel 313, and fluorescence detection channels 314 to 317 are processed by processors 350 and 351 through real-time digital processing. Based on the data signals generated in processors 350 and 351, images 311a to 317a can be generated in each light detection channel. Image-enabled sorting is performed in response to a sorting signal generated in sorting trigger 352. Sorting assembly 300c includes deflection plate 331 for deflecting particles into sample container 332 or waste stream 333. In some instances, sorting assembly 300c is configured to sort particles using enclosed particle sorting modules, such as those described in U.S. Patent Publication No. 2017 / 0299493, filed March 28, 2017, the disclosure of which is incorporated herein by reference. In some embodiments, the sorting component 300c includes a sorting decision module having multiple sorting decision units, such as those described in U.S. Patent Publication No. 2020 / 0256781, the disclosure of which is incorporated herein by reference.

[0165] Figure 3B Image-enabled particle sorting data processing according to certain embodiments is described. In some instances, image-enabled particle sorting data processing is a low-latency data processing pipeline. Each photodetector generates pulses with high-frequency modulation to encode an image (waveform). Fourier analysis is performed to reconstruct the image from the modulated pulses. The image processing pipeline generates a set of image features (image analysis), which are combined with features (event packets) derived from the pulse processing pipeline. Real-time sorting and classification electronics then classify the particles based on the image features, generating sorting decisions for the selective charging of droplets.

[0166] In some embodiments, the system is a particle analyzer, wherein the particle analysis system 401 ( Figure 4A It can be used to analyze and characterize particles, whether or not the particles are physically sorted into a collection container. Figure 4A A functional block diagram of a particle analysis system for computation-based sample analysis and particle characterization is shown. In some embodiments, particle analysis system 401 is a fluid system. Figure 4A The particle analysis system 401 shown can be configured to perform all or part of the methods described herein. The particle analysis system 401 includes a fluid system 402. The fluid system 402 may include or be coupled to a sample tube 405 and a moving fluid column within the sample tube, wherein particles 403 (e.g., cells) in the sample move along a common sample path 409.

[0167] The particle analysis system 401 includes a detection system 404 configured to collect signals from each particle as it passes through one or more detection stations along a common sample path. Detection station 408 typically refers to a monitoring area 407 on the common sample path. In some embodiments, detection can include detecting the light or one or more other characteristics of particle 403 as it passes through monitoring area 407. Figure 4A The image shows a detection station 408 with a monitoring area 407. Some embodiments of the particle analysis system 401 can include multiple detection stations. Furthermore, some detection stations can monitor more than one area.

[0168] A signal value is assigned to each signal to form a data point for each particle. As mentioned above, this data can be referred to as event data. The data point can be a multi-dimensional data point, including values ​​of various attributes measured for the particle. The detection system 404 is configured to collect a series of such data points within a first time interval.

[0169] The particle analysis system 401 may also include a control system 406. The control system 406 may include one or more processors, amplitude control circuitry, and / or frequency control circuitry. The illustrated control system is operatively associated with the fluid system 402. The control system may be configured to generate a calculated signal frequency for at least a portion of the first time interval based on the number of data points collected by the detection system 404 during the first time interval. The control system 406 may also be configured to generate an experimental signal frequency based on the number of data points in that portion of the first time interval. The control system 406 may also compare the experimental signal frequency with a calculated signal frequency or a predetermined signal frequency.

[0170] Figure 4B A flow cytometry system 400 according to an exemplary embodiment of the present invention is shown. System 400 includes a flow cytometer 410, a controller / processor 490, and a memory 495. The flow cytometer 410 includes one or more excitation lasers 415a to 415c, a focusing lens 420, a flow cell 425, a forward scattering detector 430, a side scattering detector 435, a fluorescence collecting lens 440, one or more beam splitters 445a to 445g, one or more bandpass filters 450a to 450e, one or more long-pass (“LP”) filters 455a to 455b, and one or more fluorescence detectors 460a to 460f.

[0171] Lasers 115a to 115c are excited to emit light in the form of laser beams. Figure 4B In the example system, the laser beams emitted by excitation lasers 415a to 415c have wavelengths of 488 nm, 633 nm, and 325 nm, respectively. The laser beams are first guided through one or more beam splitters 445a and 445b. Beam splitter 445a transmits 488 nm light and reflects 633 nm light. Beam splitter 445b transmits ultraviolet light (light with wavelengths from 10 to 400 nm) and reflects both 488 nm and 633 nm light.

[0172] The laser beam is then guided to a focusing lens 420, which focuses the beam onto the flow stream portion containing the sample particles within a flow cell 425. The flow cell is part of a fluid system that guides particles in the flow stream to (typically one at a time) the focused laser beam for interrogation. The flow cell can include a flow cell in a benchtop cytometer or a nozzle tip in an airflow cytometer.

[0173] Light from one or more laser beams interacts with particles in the sample through diffraction, refraction, reflection, scattering, and absorption, and is re-emitted at various wavelengths depending on the characteristics of the particles (e.g., their size, internal structure, and the presence of one or more fluorescent molecules attached to or naturally present on the particles). The fluorescence emission, along with the diffracted, refracted, reflected, and scattered light, can be guided by one or more of beam splitters 445a to 445g, bandpass filters 450a to 450e, longpass filters 455a to 455b, and fluorescence collecting lenses 440 to one or more of forward scattering detectors 430, side scattering detectors 435, and fluorescence detectors 460a to 460f.

[0174] A fluorescence collecting lens 440 collects light emitted from the interaction of particles with the laser beam and directs this light to one or more beam splitters and filters. Bandpass filters, such as bandpass filters 450a to 450e, allow a narrow range of wavelengths to pass through the filter. For example, bandpass filter 450a is a 510 / 20 filter. The first number represents the center of the spectral band. The second number provides the range of the spectral band. Therefore, the 510 / 20 filter extends 10 nm on each side of the center of the spectral band, i.e., from 500 nm to 520 nm. Short-pass filters transmit light equal to or shorter than a specified wavelength. Long-pass filters, such as long-pass filters 455a to 455b, transmit light equal to or longer than a specified wavelength. For example, long-pass filter 455a is a 670 nm long-pass filter that transmits light equal to or longer than 670 nm. Filters are typically selected to optimize the detector's specificity for a particular fluorescent dye. Filters can be configured such that the spectral band of the light transmitted to the detector is close to the emission peak of the fluorescent dye.

[0175] Beam splitters direct light of different wavelengths in different directions. Beam splitters can be characterized by, for example, short-pass and long-pass filter characteristics. For example, beam splitter 445g is a 620SP beam splitter, meaning that beam splitter 445g transmits light with wavelengths of 620 nm or shorter and reflects light with wavelengths longer than 620 nm in different directions. In one embodiment, beam splitters 445a to 445g can include optical mirrors, such as dichroic mirrors.

[0176] A forward scattering detector 430 is positioned slightly off-axis from the direct beam passing through the flow cell and is configured to detect diffracted light, i.e., excitation light that passes primarily forward or surrounds the particle. The intensity of the light detected by the forward scattering detector depends on the overall size of the particle. The forward scattering detector can include a photodiode. A side scattering detector 435 is configured to detect refracted and reflected light from the particle surface and internal structure, and tends to increase with increasing particle structural complexity. Fluorescence emitted by fluorescent molecules associated with the particle can be detected by one or more fluorescence detectors 460a to 460f. The side scattering detector 435 and the fluorescence detector can include photomultiplier tubes. The signals detected at the forward scattering detector 430, the side scattering detector 435, and the fluorescence detector can be converted into electrical signals (voltages) by the detectors. These data can provide information about the sample.

[0177] Those skilled in the art will recognize that the flow cytometer according to embodiments of the present invention is not limited to Figure 4B The flow cytometer shown may include any flow cytometer known in the art. For example, a flow cytometer may have any number of lasers, beam splitters, filters, and detectors with various wavelengths and various different configurations.

[0178] During operation, the cytometer is controlled by a controller / processor 490, and measurement data from the detector can be stored in memory 495 and processed by the controller / processor 490. Although not explicitly shown, the controller / processor 490 is coupled to the detector to receive its output signals and may also be coupled to the electrical and electromechanical components of the flow cytometer 400 to control the laser, fluid flow parameters, etc. Input / output (I / O) capabilities 497 may also be provided in the system. Memory 495, controller / processor 490, and I / O 497 may be provided entirely as an integrated part of the flow cytometer 410. In this embodiment, a display may also be formed as part of the I / O capability 497 for presenting experimental data to the user of the cytometer 400. Optionally, some or all of the memory 495 and controller / processor 490, as well as some of the I / O capabilities, may be part of one or more external devices (e.g., a general-purpose computer). In some embodiments, some or all of the memory 495 and controller / processor 490 may be able to communicate wirelessly or wiredly with the cytometer 410. The controller / processor 490, together with the memory 495 and I / O 497, can be configured to perform various functions related to the preparation and analysis of flow cytometry experiments.

[0179] Figure 4BThe system shown includes six different detectors that detect fluorescence in six different wavelength bands (which may be referred to here as “filter windows” for a given detector), defined by the configuration of filters and / or spectrometers in the beam path from flow cell 425 to each detector. Different fluorescent molecules used in flow cytometry experiments will emit light in their own characteristic wavelength bands. Specific fluorescent tags used in experiments and their associated fluorescence emission bands can be selected to roughly correspond to the filter windows of the detectors. However, as more detectors are provided and more tags are used, a perfect correspondence between filter windows and fluorescence emission spectra is not possible. Typically, although the peak of the emission spectrum of a particular fluorescent molecule may lie within the filter window of a particular detector, a portion of the emission spectrum of that tag will also overlap with the filter windows of one or more other detectors. This can be referred to as overflow. I / O 497 can be configured to receive data on a flow cytometry experiment having a set of fluorescent tags and multiple cell populations with multiple labels, each cell population having a subset of multiple labels. I / O 497 can also be configured to receive biological data assigning one or more markers to one or more cell populations, marker density data, emission spectral data, data assigning tags to one or more markers, and cytometer configuration data. Flow cytometry experimental data (e.g., tag spectral characteristics and flow cytometry configuration data) can also be stored in memory 495. Controller / processor 490 can be configured to evaluate the assignment of one or more tags to markers.

[0180] Figure 5 A functional block diagram of an example particle analyzer control system (e.g., analysis controller 500) for analyzing and displaying biological events is shown. The analysis controller 500 can be configured to implement various processes for controlling the graphical display of biological events.

[0181] The particle analyzer or sorting system 502 can be configured to acquire biological event data. For example, a flow cytometer can generate flow cytometry event data. The particle analyzer 502 can be configured to provide the biological event data to the analysis controller 500. A data communication channel may be included between the particle analyzer or sorting system 502 and the analysis controller 500. The biological event data can be provided to the analysis controller 500 via the data communication channel.

[0182] Analysis controller 500 can be configured to receive biological event data from particle analyzer or sorting system 502. The biological event data received from particle analyzer or sorting system 502 can include flow cytometry event data. Analysis controller 500 can be configured to provide a graphical display of a first graph including the biological event data to display device 506. Analysis controller 500 can also be configured to present regions of interest as gates around the biological event data clusters shown on display device 506, for example, overlaying them on the first graph. In some embodiments, the gate can be a logical combination of one or more graphical regions of interest plotted on a single-parameter histogram or bivariate graph. In some embodiments, the display can be used to display particle parameter or saturation detector data.

[0183] The analysis controller 500 can also be configured to display biological event data inside the door on the display device 506, which is different from other events in the biological event data outside the door. For example, the analysis controller 500 can be configured to display the biological event data contained inside the door in a different color than the biological events outside the door. The display device 506 can be implemented as a monitor, tablet computer, smartphone, or other electronic device configured to display a graphical interface.

[0184] The analysis controller 500 can be configured to receive a door selection signal for identifying a door from a first input device. For example, the first input device can be implemented as a mouse 510. The mouse 510 can send a door selection signal to the analysis controller 500 to identify a door to be displayed on or operated via the display device 506 (e.g., by clicking the door when the cursor is on or inside the desired door). In some embodiments, the first device can be implemented as a keyboard 508 or other means for providing input signals to the analysis controller 500, such as a touchscreen, stylus, optical detector, or voice recognition system. Some input devices can include multiple input functions. In this embodiment, each input function can be considered an input device. For example, such as... Figure 5 As shown, mouse 510 can include a right mouse button and a left mouse button, and each button can generate a trigger event.

[0185] Triggering events can enable the analysis controller 500 to change the way data is displayed, the portion of data actually displayed on the display device 506, and / or provide input for further processing, such as selecting a group of interest for particle sorting.

[0186] In some embodiments, the analysis controller 500 can be configured to detect when the mouse 510 initiates gate selection. The analysis controller 500 can also be configured to automatically modify the drawing visualization to facilitate the gating process. This modification can be based on a specific distribution of the biological event data received by the analysis controller 500.

[0187] The analysis controller 500 can be connected to the storage device 504. The storage device 504 can be configured to receive and store biological event data from the analysis controller 500. The storage device 504 can also be configured to receive and store flow cytometry event data from the analysis controller 500. The storage device 504 can also be configured to allow the analysis controller 500 to retrieve biological event data, such as flow cytometry event data.

[0188] Display device 506 can be configured to receive display data from analysis controller 500. The display data can include gating of portions of graphs and delineations of biological event data. Display device 506 can also be configured to change the presented information based on input received from analysis controller 500, combined with input from particle analyzer 502, storage device 504, keyboard 508, and / or mouse 510.

[0189] In some implementations, the analysis controller 500 can generate a user interface to receive sample events for sorting. For example, the user interface can include controls for receiving sample events or sample images. The sample events, images, or sample gates can be provided before collecting event data for the samples, or based on an initial set of events from a portion of the samples.

[0190] Figure 6A This is a schematic diagram of a particle sorting system 600 (e.g., a particle analyzer or sorting system 502) according to one embodiment presented herein. In some embodiments, the particle sorting system 600 is a cell sorting system. Figure 6A As shown, a droplet-forming transducer 602 (e.g., a piezoelectric oscillator) is coupled to a fluid conduit 601, which is capable of being coupled to, including, or is a nozzle 603. Within the fluid conduit 601, a sheath fluid 604 hydrodynamically focuses a sample fluid 606, including particles 609, into a moving fluid column 608 (e.g., a stream). Within the moving fluid column 608, particles 609 (e.g., cells) align in a single file, pass through a monitoring region 611 (e.g., where laser streams intersect), and are irradiated by an irradiation source 612 (e.g., a laser). Vibration of the droplet-forming transducer 602 causes the moving fluid column 608 to split into multiple droplets 610, some of which contain particles 609.

[0191] In operation, a detection station 614 (e.g., an event detector) identifies when a particle (or cell) of interest crosses a monitoring area 611. The detection station 614 is fed into a timing circuit 628, which in turn feeds into a flash charge circuit 630. At the droplet interruption point, a flash charge can be applied to the moving fluid column 608 based on the timing droplet delay (Δt), causing the droplet of interest to carry a charge. The droplet of interest can comprise one or more particles or cells to be sorted. The charged droplet is then deflected by activating a deflection plate (not shown), sorting it into a container (e.g., a collection tube or a porous or microporous sample plate) where a pore or micropore can be associated with a specific droplet of interest. Figure 6A As shown, the droplets can be collected in the discharge container 638.

[0192] A detection system 616 (e.g., a droplet boundary detector) is used to automatically determine the phase of the droplet drive signal as a particle of interest passes through the monitoring region 611. An exemplary droplet boundary detector is described in U.S. Patent No. 7,679,039, the entire contents of which are incorporated herein by reference. The detection system 616 allows the instrument to accurately calculate the position of each detected particle within the droplet. The detection system 616 can be fed into an amplitude signal 620 and / or a phase signal 618, which (via amplifier 622) is further fed into an amplitude control circuit 626 and / or a frequency control circuit 624. The amplitude control circuit 626 and / or the frequency control circuit 624 then control the droplet forming transducer 602. The amplitude control circuit 626 and / or the frequency control circuit 624 can be included in a control system.

[0193] In some embodiments, the sorting electronics (e.g., detection system 616, detection station 614, and processor 640) can be coupled to a memory configured to store detected events and sorting decisions based thereon. The sorting decisions can be included in the event data of the particles. In some embodiments, detection system 616 and detection station 614 can be implemented as a single detection unit or communication coupling, such that event measurements can be collected by one of detection system 616 or detection station 614 and provided to non-collecting elements.

[0194] Figure 6B This is a schematic diagram of a particle sorting system according to an embodiment of the present document. Figure 6B The particle sorting system 600 shown includes deflection plates 652 and 654. Charge can be applied via a stream-charging wire in the barbs. This generates a droplet stream 610 containing particles 610 for analysis. The particles can be illuminated with one or more light sources (e.g., lasers) to produce light scattering and fluorescence information. This can be achieved, for example, through sorting electronics or other detection systems. Figure 6B(Not shown in the image) to analyze particle information. Deflecting plates 652 and 654 can be independently controlled to attract or repel charged droplets, thereby guiding the droplets to a destination collection container (e.g., one of 672, 674, 676, or 678). Figure 6B As shown, deflector plates 652 and 654 can be controlled to guide particles along a first path 662 to container 674 or along a second path 668 to container 678. If the particles are not of interest (e.g., do not exhibit scattering or illumination information within a specified sorting range), the deflector plates can allow the particles to continue flowing along flow path 664. Such uncharged droplets can enter the waste container, for example, via a suction device 670.

[0195] It can include sorting electronics to initiate the collection of measurement results, receive the fluorescence signal of the particles, and determine how to adjust the deflection plate for particle sorting. Figure 6B The example implementation of the illustrated embodiment includes BD FACSAria, commercially available from Becton, Dickinson and Company (Franklin Lakes, NJ). TM A series of flow cytometers.

[0196] A method for classifying single cells in a sample from a flowing stream.

[0197] This disclosure also includes methods for classifying cells (e.g., single cells) in a sample from a flowing stream, such as classifying them based on biological parameters determined by feature vectors. A method according to some embodiments includes: receiving two or more datasets, each of which is associated with one of a plurality of data patterns; applying an algorithm to convert each of the two or more datasets into a feature vector based on the associated data pattern; receiving a classification task for the two or more datasets; and applying one or more of a plurality of classification models to the feature vectors based on the received classification task to determine a category. In some instances, the method further includes illuminating a sample containing cells in the flowing stream with a light source and measuring the light from the illuminated cells using a light detection system with a photodetector.

[0198] In implementation of the subject method according to certain embodiments, a light detection system with a photodetector is used to measure light from a sample containing cells in a flowing stream (e.g., illuminating the sample with a light source). In some embodiments, the light source is a broadband light source that emits light with a wide wavelength range, such as spanning 50 nm or greater, for example 100 nm or greater, for example 150 nm or greater, for example 200 nm or greater, for example 250 nm or greater, for example 300 nm or greater, for example 350 nm or greater, for example 400 nm or greater, and including spans of 500 nm or greater. For example, a suitable broadband light source emits light with a wavelength range of 200 nm to 1500 nm. Another example of a suitable broadband light source includes a light source that emits light with a wavelength range of 400 nm to 1000 nm. When the method includes illumination with a broadband light source, the broadband light source schemes of interest may include, but are not limited to, halogen lamps, deuterium arc lamps, xenon arc lamps, stable fiber-coupled broadband light sources, broadband LEDs with a continuous spectrum, superluminescent diodes, semiconductor light-emitting diodes, broadband LED white light sources, multi-LED integrated white light sources, and other broadband light sources or any combination thereof.

[0199] In other embodiments, the method includes illumination using a narrowband light source that emits light of a specific wavelength or a narrow wavelength range, such as a light source emitting light of a narrow wavelength range, for example, a light source emitting light of 50 nm or less, such as 40 nm or less, such as 30 nm or less, such as 25 nm or less, such as 20 nm or less, such as 15 nm or less, such as 10 nm or less, such as 5 nm or less, such as 2 nm or less, and including light sources emitting light of a specific wavelength (i.e., monochromatic light). When the method includes illumination using a narrowband light source, the narrowband light source scheme of interest may include, but is not limited to, narrow-wavelength LEDs, laser diodes, or broadband light sources coupled to one or more optical bandpass filters, diffraction gratings, monochromators, or any combination thereof.

[0200] In some embodiments, the method includes irradiating the sample with one or more lasers. As described above, the type and number of lasers will vary depending on the sample and the light to be collected, and may be gas lasers, such as helium-neon lasers, argon lasers, krypton lasers, xenon lasers, nitrogen lasers, CO2 lasers, CO lasers, argon-fluorine (ArF) excimer lasers, krypton-fluorine (KrF) excimer lasers, xenon-chlorine (XeCl) excimer lasers, or xenon-fluorine (XeF) excimer lasers, or combinations thereof. In other instances, the method includes irradiating the flow with a dye laser, such as a stilbene, coumarin, or rhodamine laser. In still other instances, the method includes irradiating the flow with a metal vapor laser, such as a helium-cadmium (HeCd) laser, a helium-mercury (HeHg) laser, a helium-selenium (HeSe) laser, a helium-silver (HeAg) laser, a strontium laser, a neon-copper (NeCu) laser, a copper laser, or a gold laser, or combinations thereof. In other instances, the method involves irradiating the flow with a solid-state laser, such as a ruby ​​laser, an Nd:YAG laser, an NdCrYAG laser, an Er:YAG laser, an Nd:YLF laser, an Nd:YVO4 laser, an Nd:yCa4O(BO3)3 laser, an Nd:YCOB laser, a titanite sapphire laser, a thulium YAG laser, a ytterbium 2O3 laser, or a cerium-doped laser, or combinations thereof.

[0201] One or more of the above-described light sources can be used to illuminate the sample, such as two or more light sources, three or more light sources, four or more light sources, five or more light sources, and including ten or more light sources. The light sources can include any combination of light source types. For example, in some embodiments, the method includes illuminating the sample in the flowing stream using a laser array, such as an array having one or more gas lasers, one or more dye lasers, and one or more solid-state lasers.

[0202] The sample can be illuminated using wavelengths from 200 nm to 1500 nm, such as 250 nm to 1250 nm, 300 nm to 1000 nm, 350 nm to 900 nm, and including 400 nm to 800 nm. For example, when the light source is a broadband light source, the sample can be illuminated using wavelengths from 200 nm to 900 nm. In other instances, when the light source comprises multiple narrowband light sources, the sample can be illuminated using specific wavelengths from 200 nm to 900 nm. For example, the light source can be multiple narrowband LEDs (1 nm to 25 nm), each LED independently emitting light with wavelengths from 200 nm to 900 nm. In other embodiments, the narrowband light source comprises one or more lasers (e.g., a laser array) that illuminate the sample using specific wavelengths from 200 nm to 700 nm, such as laser arrays having gas lasers, excimer lasers, dye lasers, metal vapor lasers, and solid-state lasers as described above.

[0203] When using more than one light source, the light sources can be used to illuminate the sample simultaneously, sequentially, or in combination. For example, each of the light sources can be used to illuminate the sample simultaneously. In other embodiments, each of the light sources is used to illuminate the flow sequentially. When illuminating the sample sequentially using more than one light source, the duration of illumination for each light source can be independently 0.001 microseconds or longer, such as 0.01 microseconds or longer, 0.1 microseconds or longer, 1 microsecond or longer, 5 microseconds or longer, 10 microseconds or longer, 30 microseconds or longer, and includes 60 microseconds or longer. For example, the method may include illuminating the sample with a light source (e.g., a laser) for a duration of 0.001 microseconds to 100 microseconds, such as 0.01 microseconds to 75 microseconds, such as 0.1 microseconds to 50 microseconds, such as 1 microsecond to 25 microseconds, and including 5 microseconds to 10 microseconds. In embodiments where two or more light sources are used to illuminate the sample sequentially, the duration of illumination for each light source can be the same or different.

[0204] The time interval between each light source illumination can also vary as needed, independently with a delay of 0.001 microseconds or longer, such as 0.01 microseconds or longer, 0.1 microseconds or longer, 1 microsecond or longer, 5 microseconds or longer, 10 microseconds or longer, 15 microseconds or longer, 30 microseconds or longer, and including 60 microseconds or longer. For example, the time interval between each light source illumination can be from 0.001 microseconds to 60 microseconds, such as 0.01 microseconds to 50 microseconds, such as 0.1 microseconds to 35 microseconds, such as 1 microsecond to 25 microseconds, and including 5 microseconds to 10 microseconds. In some embodiments, the time interval between each light source illumination is 10 microseconds. In embodiments where more than two (i.e., three or more) light sources are used to sequentially illuminate the sample, the delay between each light source illumination can be the same or different.

[0205] The sample can be illuminated continuously or at discrete intervals. In some instances, the method involves continuously illuminating the sample with a light source. In other instances, the sample is illuminated with a light source at discrete intervals, such as every 0.001 milliseconds, every 0.01 milliseconds, every 0.1 milliseconds, every 1 millisecond, every 10 milliseconds, every 100 milliseconds, and including every 1000 milliseconds or other intervals.

[0206] Depending on the light source, the sample can be illuminated at varying distances, such as 0.01 mm or more, 0.05 mm or more, 0.1 mm or more, 0.5 mm or more, 1 mm or more, 2.5 mm or more, 5 mm or more, 10 mm or more, 15 mm or more, 25 mm or more, and including 50 mm or more. Furthermore, the angle or illumination can also vary, ranging from 10° to 90°, such as 15° to 85°, 20° to 80°, 25° to 75°, and including 30° to 60°, such as 90°.

[0207] In embodiments, the method includes irradiating a cell sample with a frequency-modulated light beam. In some instances, the method includes irradiating the sample with two or more frequency-shifted light beams. As described above, a beam generator assembly may be employed, having a laser and an acousto-optic device for frequency-shifting the laser. In these embodiments, the method includes irradiating the acousto-optic device with a laser. The specific wavelength of the laser can vary from 200 nm to 1500 nm, for example from 250 nm to 1250 nm, for example from 300 nm to 1000 nm, for example from 350 nm to 900 nm, and includes 400 nm to 800 nm, depending on the desired wavelength of the light generated in the output laser beam (e.g., for irradiating a sample in a flowing stream). One or more lasers can be used to irradiate the acousto-optic device, for example, two or more lasers, for example, three or more lasers, for example, four or more lasers, for example, five or more lasers, and including ten or more lasers. The lasers can include any combination of laser types. For example, in some embodiments, the method includes irradiating an acousto-optic device, such as an array of one or more gas lasers, one or more dye lasers, and one or more solid-state lasers, using a laser array.

[0208] When using more than one laser, the lasers can be used to illuminate the acousto-optic device simultaneously, sequentially, or in combination. For example, each of the lasers can be used to illuminate the acousto-optic device simultaneously. In other embodiments, each of the lasers is used to illuminate the acousto-optic device sequentially. When using more than one laser to illuminate the acousto-optic device sequentially, the duration for which each laser illuminates the acousto-optic device can be independently 0.001 microseconds or longer, for example, 0.01 microseconds or longer, for example, 0.1 microseconds or longer, for example, 1 microsecond or longer, for example, 5 microseconds or longer, for example, 10 microseconds or longer, 30 microseconds or longer, and including 60 microseconds or longer. For example, the method may include illuminating the acousto-optic device with a laser for a duration of 0.001 microseconds to 100 microseconds, for example, 0.01 microseconds to 75 microseconds, for example, 0.1 microseconds to 50 microseconds, for example, 1 microsecond to 25 microseconds, and including 5 microseconds to 10 microseconds. In embodiments where two or more lasers are used to illuminate the acousto-optic device sequentially, the duration for which each laser illuminates the acousto-optic device can be the same or different.

[0209] The time interval between each laser irradiation can also vary as needed, independently with a delay of 0.001 microseconds or longer, such as 0.01 microseconds or longer, 0.1 microseconds or longer, 1 microsecond or longer, 5 microseconds or longer, 10 microseconds or longer, 15 microseconds or longer, 30 microseconds or longer, and including 60 microseconds or longer. For example, the time interval between each light source irradiation can be from 0.001 microseconds to 60 microseconds, such as 0.01 microseconds to 50 microseconds, such as 0.1 microseconds to 35 microseconds, such as 1 microsecond to 25 microseconds, and including 5 microseconds to 10 microseconds. In some embodiments, the time interval between each light source irradiation is 10 microseconds. In embodiments where more than two (i.e., three or more) lasers are used to irradiate the acousto-optic device sequentially, the delay between each laser irradiation can be the same or different.

[0210] The acousto-optic device can be illuminated continuously or at discrete intervals. In some instances, the method includes continuously illuminating the acousto-optic device using a laser. In other instances, the acousto-optic device is illuminated using a laser at discrete intervals, such as every 0.001 milliseconds, every 0.01 milliseconds, every 0.1 milliseconds, every 1 millisecond, every 10 milliseconds, every 100 milliseconds, and including every 1000 milliseconds or other intervals.

[0211] Depending on the laser, the acousto-optic device can be illuminated at varying distances, such as 0.01 mm or more, 0.05 mm or more, 0.1 mm or more, 0.5 mm or more, 1 mm or more, 2.5 mm or more, 5 mm or more, 10 mm or more, 15 mm or more, 25 mm or more, and including 50 mm or more. Furthermore, the angle or illumination can also vary, ranging from 10° to 90°, such as 15° to 85°, 20° to 80°, 25° to 75°, and including 30° to 60°, such as 90°.

[0212] In an embodiment, the method includes applying radio frequency (RF) drive signals to an acousto-optic device to generate an angle-deflected laser beam. Two or more RF drive signals may be applied to the acousto-optic device to generate an output laser beam with a desired number of angle deflections, such as three or more RF drive signals, four or more RF drive signals, five or more RF drive signals, six or more RF drive signals, seven or more RF drive signals, eight or more RF drive signals, nine or more RF drive signals, ten or more RF drive signals, fifteen or more RF drive signals, twenty-five or more RF drive signals, fifty or more RF drive signals, and including one hundred or more RF drive signals.

[0213] Each angle-deflected laser beam generated by a radio frequency (RF) drive signal has an intensity based on the amplitude of the applied RF drive signal. In some embodiments, the method includes applying an RF drive signal with an amplitude sufficient to generate an angle-deflected laser beam with a desired intensity. In some instances, each applied RF drive signal independently has an amplitude of about 0.001V to about 500V, for example about 0.005V to about 400V, for example about 0.01V to about 300V, for example about 0.05V to about 200V, for example about 0.1V to about 100V, for example about 0.5V to about 75V, for example about 1V to 50V, for example about 2V to 40V, for example 3V to about 30V, and includes about 5V to about 25V. In some embodiments, each applied radio frequency drive signal has a frequency of about 0.001 MHz to about 500 MHz, for example about 0.005 MHz to about 400 MHz, for example about 0.01 MHz to about 300 MHz, for example about 0.05 MHz to about 200 MHz, for example about 0.1 MHz to about 100 MHz, for example about 0.5 MHz to about 90 MHz, for example about 1 MHz to about 75 MHz, for example about 2 MHz to about 70 MHz, for example about 3 MHz to about 65 MHz, for example about 4 MHz to about 60 MHz, and includes about 5 MHz to about 50 MHz.

[0214] In these embodiments, the angle-deflected laser beams in the output laser beam are spatially separated. Depending on the applied RF drive signal and the desired illumination profile of the output laser beam, the angle-deflected laser beams may be separated by 0.001 μm or more, for example 0.005 μm or more, for example 0.01 μm or more, for example 0.05 μm or more, for example 0.1 μm or more, for example 0.5 μm or more, for example 1 μm or more, for example 5 μm or more, for example 10 μm or more, for example 100 μm or more, for example 500 μm or more, for example 1000 μm or more, and including 5000 μm or more. In some embodiments, the angle-deflected laser beams may overlap, for example, along the horizontal axis of the output laser beam with adjacent angle-deflected laser beams. The overlap between adjacent angle-deflected laser beams (e.g., beam spot overlap) can be 0.001 μm or more, such as 0.005 μm or more, such as 0.01 μm or more, such as 0.05 μm or more, such as 0.1 μm or more, such as 0.5 μm or more, such as 1 μm or more, such as 5 μm or more, such as 10 μm or more, and includes overlap of 100 μm or more.

[0215] In some instances, a sample in a flowing stream is illuminated with multiple frequency-shifted beams, and mitochondrial cells in the flowing stream are imaged, as described by Diebold et al. in Nature Photonics Vol. 7(10); 806-810 (2013), and in U.S. Patent Nos. 9,423,353, 9,784,661, 9,983,132, 10,006,852, 10,078,045, 10,036,699, 10,222,316, 10,288,546, 10,324,019, and 10 The disclosures described in U.S. Patent Publications Nos. 408,758, 10,451,538, 10,620,111, 2017 / 0133857, 2017 / 0328826, 2017 / 0350803, 2018 / 0275042, 2019 / 0376895, and 2019 / 0376894 are incorporated herein by reference.

[0216] Light emitted by irradiated cells in the sample is transmitted to a light detection system as described above and measured by multiple photodetectors. In some embodiments, the method includes measuring light collected within a certain wavelength range (e.g., 200 nm–1000 nm). For example, the method may include the spectrum of light collected in one or more wavelength ranges from 200 nm to 1000 nm. However, in other embodiments, the method includes measuring light collected at one or more specific wavelengths. For example, light collected at one or more of the following wavelengths, and any combination thereof, may be measured: 450 nm, 518 nm, 519 nm, 561 nm, 578 nm, 605 nm, 607 nm, 625 nm, 650 nm, 660 nm, 667 nm, 670 nm, 668 nm, 695 nm, 710 nm, 723 nm, 780 nm, 785 nm, 647 nm, and 617 nm.

[0217] The collected light can be measured continuously or at discrete intervals. In some instances, the method involves measuring the light continuously. In other instances, the light is measured at discrete intervals, such as every 0.001 milliseconds, every 0.01 milliseconds, every 0.1 milliseconds, every 1 millisecond, every 10 milliseconds, every 100 milliseconds, and including every 1000 milliseconds or other intervals.

[0218] During the method described in this subject matter, the collected light may be measured one or more times, such as two or more times, three or more times, five or more times, and including ten or more times. In some embodiments, the light emitted by the sample is measured two or more times, and in some instances, the data is averaged.

[0219] It can measure the light emitted by cells in a sample at one or more wavelengths, such as at 5 or more different wavelengths, such as at 10 or more different wavelengths, such as at 25 or more different wavelengths, such as at 50 or more different wavelengths, such as at 100 or more different wavelengths, such as at 200 or more different wavelengths, such as at 300 or more different wavelengths, and includes measuring light collected at 400 or more different wavelengths.

[0220] In some embodiments, the method further includes adjusting the light emitted from the sample before detecting the light. For example, light from the sample source may pass through one or more lenses, mirrors, pinholes, slits, gratings, light refractors, and any combination thereof. In some instances, the collected light passes through one or more focusing lenses to, for example, narrow the light profile. In other instances, light emitted from the sample passes through one or more collimators to reduce beam divergence.

[0221] In some embodiments, image data is generated from measured light. In some instances, the generated data is image data. In some instances, the image data includes dark-field images of cells. In some instances, the image data includes light-field images of cells. In some instances, the image data includes fluorescently tagged images of cells. In some instances, the image data includes untagged images of cells. In some instances, the image data includes autofluorescence image data. The image data may include one or more images of cells, such as two or more images, such as three or more images, such as four or more images, such as five or more images, such as ten or more images, such as fifteen or more images, such as 25 or more images, and includes images of 50 or more cells. In some instances, the image data includes two or more images of different types (e.g., fluorescently tagged images, autofluorescence images, dark-field images, etc.), such as three or more images of different types, such as four or more images of different types, and includes five or more images of different types. In some embodiments, the method includes converting one type of image into another, for example, an image generated in a light-loss photodetector channel (e.g., a bright-field image) is inverted to create a dark-field image.

[0222] In some embodiments, the method includes generating an image based on detected light absorption, detected light scattering, detected light emission, or any combination thereof. In some instances, the method includes generating an image based on light absorption detected from a sample (e.g., from a bright-field detector). In some instances, the method includes generating an image based on light scattering detected from a sample, such as from a side-scattering detector, a forward-scattering detector, or a combination of both. In some instances, the method includes generating an image based on light emitted from the sample. In other instances, the method includes generating an image based on a combination of detected light absorption and detected light scattering.

[0223] In some embodiments, the method includes receiving two or more datasets associated with multiple data patterns. In some instances, the data patterns are image data patterns. In some instances, the data patterns are sequence data patterns (e.g., waveforms). In some instances, the data patterns are tabular data patterns.

[0224] In some embodiments, the multiple data modes include image data modes. In some embodiments, the method includes generating a single image for each cell in the sample based on each form of detection light. In other embodiments, the method includes generating multiple images for each cell, such as two or more, three or more, five or more, ten or more, and including 25 or more images per cell. For example, the method includes generating a first image of the cell based on fluorescence detected from a tagged cell; generating a second image of the cell based on detected light absorption; and generating a third image based on detected light scattering. In other embodiments, the method includes generating two or more images from each form of detection light, such as three or more, four or more, five or more, and including ten or more images or combinations thereof.

[0225] In some instances, the algorithm is a rule-based image processing algorithm. In some instances, the algorithm includes a neural network. In some instances, the method includes inverting the light loss channels of a dataset associated with an image data pattern, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to zero mean, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to unit standard deviation, or a combination thereof.

[0226] In some instances, the method includes generating image data, such as digital waveforms generated in one or more photodetector channels. In some instances, the method includes calculating image parameters directly from the waveform. In some instances, the method includes calculating image parameters from image data, the generated image, or a combination thereof. In some instances, the method includes determining cell parameters solely from the calculated image parameters (e.g., radial moment, eccentricity, etc.). In some instances, the method includes determining cell parameters solely from the particle image rather than from another data source (e.g., a data signal waveform). In some instances, the method includes determining cell parameters using a combination of a generated particle image and a data signal waveform generated in response to measurement light from the irradiated particles.

[0227] In some embodiments, the method includes generating frequency-coded data (e.g., frequency-coded spatial data) from measurement light of sample cells in a flowing stream. In some instances, the method includes generating one or more images from the frequency-coded data. The frequency-coded data may be generated in one or more detection channels, such as two or more, three or more, four or more, five or more, six or more, and including eight or more detection channels. In some embodiments, the frequency-coded data includes data components acquired (or derived) from light from different detectors, such as detected light absorption or detected light scattering. In some instances, the method includes phase correction of the frequency-coded data. In some instances, the method includes generating a phase-corrected image of the cells by performing a transform on the frequency-coded data. In one example, the method includes performing phase correction on the frequency-coded data by performing a Fourier transform (FT) on the frequency-coded data. In another example, the method includes performing phase correction on the frequency-coded data by performing a discrete Fourier transform (DFT) on the frequency-coded data. In yet another example, the method includes performing phase correction on the frequency-coded data by performing a short-time Fourier transform (STFT) on the frequency-coded data. In some embodiments, the method includes performing a transformation on the frequency-coded data without performing any mathematical imaginary calculations (i.e., only performing mathematical real calculations on the transformation) to generate an image from the frequency-coded data.

[0228] In some embodiments, the multiple data modes include a sequence data mode. In some instances, the sequence data includes waveforms generated in one or more different photodetector channels. In some instances, the sequence data includes raw waveforms. In some instances, the sequence data includes raw image data in the form of raw waveforms generated in imaging photodetector channels. In some instances, the sequence data includes raw waveforms, and the image data is generated in real time from the raw waveforms.

[0229] In some embodiments, multiple data modes include a tabular data mode. In some instances, tabular data is applied as a flow cytometry data file of tabular data to a neural network. In some instances, the tabular data is spectral tabular data generated in one or more fluorescence photodetector channels. In some instances, the spectral tabular data is generated from fluorescence measured under light in one or more spectral wavelength ranges, such as 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, twelve or more, sixteen or more, and includes light in 24 or more different spectral wavelength ranges. In some instances, the spectral tabular data is uncompensated or unprocessed spectral data. In some instances, the spectral tabular data is compensated or unmixed spectral data. Spectral tabular data can be unmixed by spectral analysis of the light from each fluorophore in the sample (e.g., using a weighted least squares algorithm or a generalized least squares algorithm). In some embodiments, overlap between each distinct fluorophore is determined, and the contribution of each fluorophore to the overlapping fluorescence is calculated. In some embodiments, spectral analysis of the spectral tabulation data is performed by calculating the spectral unmixing matrix of the fluorescence spectrum of each of the plurality of fluorophores with overlapping fluorescence detected by the photodetector system in the sample. For example, spectral unmixing of the spectral tabulation data may include the Mohr-Penrose inverse or pseudo-inverse of the spectral matrix. In some instances, the algorithm used for spectral unmixing is characterized by the Chollisky decomposition of the unmixing matrix. In some embodiments, unmixed spectral tabulation data is calibrated using a calibration scaling factor and fluorophore abundance.

[0230] In some instances, spectral table data from each fluorophore (e.g., calculating the spectral unmixing matrix for each fluorophore) can be used to estimate the abundance of each fluorophore in the sample. In some embodiments, the abundance of each fluorophore associated with the target particle can be determined. In some embodiments, the spectral table data is spectrally unmixed using algorithms such as those described in U.S. Patent No. 11,009,400, U.S. Patent Publication No. 2024 / 0192122, filed December 12, 2023, and U.S. Patent Application No. 18 / 986,295, filed December 18, 2024, the disclosures of which are incorporated herein by reference.

[0231] In some embodiments, the table data includes scattering table data. In some instances, the scattering table data is generated from measured forward-scattered light from the sample (i.e., the photodetector signal generated in the forward-scattering photodetector channel FSC). In some instances, the scattering table data is generated from measured side-scattered light from the sample (i.e., the photodetector signal generated from the side-scattering photodetector channel SSC).

[0232] In embodiments, the method includes applying algorithms to convert each of two or more datasets into feature vectors. Any suitable machine learning algorithm can be implemented to convert the generated image data, sequence data, and tabular data into feature vectors, wherein the machine learning algorithm of interest can include, but is not limited to, linear regression, logistic regression, Naive Bayes, k-nearest neighbor (kNN), random forest, decision tree, support vector machine, gradient boosting, and clustering algorithms. In some embodiments, the method includes applying neural networks, such as artificial neural networks, convolutional neural networks, or recurrent neural networks. In some instances, the method includes implementing a Python script. In some instances, the method includes applying artificial neural networks (e.g., convolutional neural networks, etc.) as described above, Bayesian statistics, learning automata, hidden Markov modeling, linear classifiers, quadratic classifiers, association rule learning, etc.

[0233] In some instances, the method involves applying a neural network to convert the generated data into feature vectors. In some instances, the neural network includes feature engineering layers for generating feature vectors from image data, sequence data, and tabular data. In some instances, the feature engineering layers convert image data into image parameters of the cells. In some instances, the feature vectors are quantitative image parameters. In some instances, the feature vectors are the radial moments of the cells. In some instances, the feature vectors are the size of the cells. In some instances, the feature vectors are the diffusion rate of the cells. In some instances, the feature vectors are the eccentricity of the cells. In some instances, the feature vectors are the pointillism of the cells. In some instances, the feature vectors are the shape of the cells. In some instances, the feature vectors are one or more morphological features of the cells.

[0234] In some instances, the feature engineering layer incorporates algorithms for extracting sequential data features from photodetector waveforms. In some instances, the algorithm includes a sequence learning machine learning model. In some instances, the feature engineering layer applies a machine learning model to extract features from waveforms.

[0235] In some instances, the feature-engineered layer includes algorithms for spectral unmixing of spectral tabular data. In some instances, it includes algorithms for calculating the abundance of biomarker molecules (e.g., surface biomarker fractions). In some instances, it includes algorithms for calculating fluorophore abundance. In some instances, it includes algorithms for calculating cellular physical measurements based on scattered light tabular data. For example, the feature-engineered layer may apply a machine learning model that converts the raw scattered light tabular data into cellular measurements, such as the size of a cell or cellular component in nanometers.

[0236] In one embodiment, the method includes applying a machine learning algorithm (e.g., a neural network) to determine one or more biological parameters of a cell based on a feature vector. Depending on the application, different analytical models may be applied to extract biological insights into the cell from the feature vector.

[0237] In some embodiments, the analytical model is an end-to-end model. In some instances, the end-to-end model is designed to identify predefined cell phenotypes. In some instances, the end-to-end model is a gated hierarchy or classification model, such as a single-classifier or multi-classifier model (e.g., a single-layer or multi-layer perceptron), a random forest classifier, or a support vector machine. In some embodiments, the analytical model is an exploratory model, such as a model capable of discovering new cell populations. In some instances, the applied exploratory model is a dimensionality reduction algorithm, such as Uniform Manifold Approximation and Projection (UMAP), t-Distributed Random Nearest Neighbor Embedding (TSNE), and Principal Component Analysis (PCA).

[0238] In some embodiments, the analytical model can select all or a subset of the feature vectors based on prior knowledge of which features are associated with the biological parameters of interest. For example, as described in more detail below, a survival classification task can be developed to identify live cells based on a combination of bright-field and dark-field image data from cells. In these instances, the analytical model can utilize feature vectors from light loss, side scattering, and forward scattering image data. In other instances, the TBNK classification task can be designed to classify cells into T cells, B cells, and natural killer (NK) cells using image data. In some instances, the analytical model applies unlabeled image data and autofluorescence image data to determine biological parameters from the feature vectors.

[0239] In some embodiments, the neural network includes a machine learning encoder layer. In some instances, the encoder layer is a rule-based image processing pipeline. In some instances, the encoder layer includes an image encoder. In some instances, the image encoder has an image recognition model architecture. In some instances, the image encoder has an encoder model architecture such as RestNet34, ShuffleNet V2, EfficientNet V2-S, Inception V3, and combinations thereof. In some instances, the method includes training the encoder layer using a supervised method with a weighted cross-entropy loss function. In some instances, the method includes reconstructing images in real time from data waveforms. In some instances, the trained encoder layer is validated based on the evaluated feature quality. In some instances, metrics such as precision, recall, and F-1 score are used to determine the evaluated feature quality.

[0240] In some embodiments, the method includes applying a neural network to data to classify cells based on one or more defined feature vectors. In some instances, the method includes applying a dimensionality reduction algorithm to the generated data. In some instances, the method includes applying a dimensionality reduction algorithm to high-dimensional feature vectors. In some instances, the dimensionality reduction algorithm is selected from uniform manifold approximation and projection (UMAP), t-distributed random nearest neighbor embedding (t-SNE), principal component analysis (PCA), and combinations thereof.

[0241] In some instances, the method involves applying a classification task to two or more datasets and applying one or more classification models to a feature vector based on the received classification task to determine the category. In some instances, the classification task is a survival classification task. In some instances, the classification task includes a survival classification task, with categories including a survival category, a death category, or an apoptosis category. In some instances, the method involves classifying the data using a cell type classification task. In some instances, the cell type classification task classifies cells into T cells, B cells, and natural killer (NK) cells. In some instances, the method involves applying a classification task that includes single-cell classification, with categories including single-cell categories, adherent two-cell categories, separated two-cell categories, three-cell categories, or fragment categories. In some instances, the classification task includes a whole blood classification task, with whole blood classification including granulocyte categories, monocyte categories, or lymphocyte categories. In some instances, the classification task includes a T cell activation classification task, with categories including an activated category or a non-activated category.

[0242] In some embodiments, the neural network includes a learning algorithm configured to train and optimize cell classification. In some instances, different datasets are combined and shuffled to ensure that each training batch contains samples from different datasets. In some instances, the dataset is divided into three distinct sets: 1) for training; 2) for validation; and 3) for testing. In some instances, the training dataset comprises 40% to 80% of the data, e.g., 50% to 75%, and includes 70% of the data for training. In some instances, the validation dataset comprises 10% to 30% of the data, e.g., 15% to 25%, and includes 30% of the data for validation. In some instances, the test dataset comprises 5% to 20% of the data, e.g., 7.5% to 15%, and includes 10% of the data for testing.

[0243] In some instances, one or more of image data patterns, sequence data patterns, and tabular data patterns are processed in real time during the application of the training algorithm. In some instances, images are reconstructed in real time from raw data waveforms during the application of the training algorithm. In some instances, the method includes transforming the image before applying it to the neural network. In some instances, data from the light loss channel is inverted to create a dark-field image. In some instances, all photodetector channels are zero-padded. In some instances, the pixel values ​​of each photodetector channel are scaled to have zero mean and unit standard deviation.

[0244] In some embodiments, the training algorithm includes a supervised method using a weighted cross-entropy loss function. In some instances, the weights are calculated as the product of class weights and task weights, where the class weights correspond to the class abundance in the task and the task weights correspond to the task abundance across all datasets. In some instances, the Adam optimizer with a predetermined learning rate and no weight decay is used. In some instances, the predetermined learning rate ranges from 1e... -1 up to 1e -5 For example, from 1e -2 up to 1e -4 And including 1e -3 The learning rate. In some instances, the learning rate does not decay with weights. In some embodiments, the learning rate is reduced by a predetermined factor if the validation loss remains constant, for example, if the validation loss remains constant for one or more time steps, such as two or more time steps, such as three or more time steps, such as five or more time steps, such as ten or more time steps, such as fifteen or more time steps, such as twenty or more time steps, such as twenty or five or more time steps, and including fifty or more time steps. In these embodiments, if the validation loss remains constant, the learning rate will be reduced by a factor of 2 or more, such as a factor of 3 or more, such as a factor of 4 or more, such as a factor of 5 or more, such as a factor of 10 or more, and including twenty or more. In some embodiments, if the validation loss remains constant for 10 time steps, the learning rate will be reduced by a factor of 10. The batch size used for training can vary, and the batch size can be 64 or more, such as 128 or more, such as 256 or more, and including 512 or more. In some instances, the batch size is 128. In other instances, the batch size is 512.

[0245] In some embodiments, the performance of the algorithm is evaluated. In some instances, the performance of the trained model is evaluated using validation or test datasets. In some instances, dimensionality reduction is applied to the high-dimensional feature vectors to evaluate feature quality. In some instances, applying dimensionality reduction to high-dimensional feature vectors can examine how known groups are separated in a low-dimensional space. In some instances, the performance of the algorithm is measured using one or more of the precision, recall, and F1-score for each class. In some instances, receiver operating feature curves are generated for each class. In some instances, the area under the curve (AUC) is evaluated to assess the performance of the feature vector algorithm.

[0246] In some embodiments, the method includes determining one or more sorting gates for classifying cells of a sample as described above. The term "gate" is used herein in its conventional sense, referring to a classifier boundary that identifies a subset of data of interest. In some instances, a gate can define a set of events of particular interest. Furthermore, "gating" can refer to the process of classifying data using defined gates for a given dataset, where a gate can be one or more regions of interest combined with Boolean logic. In some embodiments, gates identify particles exhibiting the same image parameters. Examples of gating methods, for instance, have been described in U.S. Patent Applications Nos. 4,845,653, 5,627,040, 5,739,000, 5,795,727, 5,962,238, 6,014,904, 6,944,338, and 8,990,047, the disclosures of which are incorporated herein by reference. In some embodiments, a gate defines a swarm of particles from one or more different samples that has been previously identified (e.g., by the user) as corresponding to the feature of interest.

[0247] In some embodiments, the method includes displaying a gating strategy on a graphical user interface. For example, the analysis algorithm may be one or more of a spectral compensation matrix, a clustering algorithm, and a t-distributed random nearest neighbor embedding (t-SNE) algorithm. In some instances, the analysis algorithm is applied to a particle swarm cluster by dragging its icon onto the cluster. In other instances, the particle swarm cluster is selected from a drop-down menu, and the analysis algorithm is applied. In some instances, the analysis algorithm is a spectral unmixing algorithm, such as the algorithm described in U.S. Patent No. 11,009,400 and International Patent Application No. PCT / US2021 / 46741, filed August 19, 2021, the disclosure of which is incorporated herein by reference. In some instances, a computational sorting algorithm is used to determine the gating strategy, such as that described in U.S. Patent No. 11,513,054, the disclosure of which is incorporated herein by reference. In some instances, computer software can be used to determine the sorting gate, such as HyperFinder (e.g., as described by Bonavia et al. in "Frontiers in Immunology" 2022; 13:1007016) and computational sorting using HyperFinder, FlowJo software, and BD FACSDiva software (Becton Dickinson, 2021), the disclosure of which is incorporated herein by reference. In some embodiments, the gating strategy is developed on a separate computing system (e.g., a different computer system or network) and transmitted to a flow cytometer to implement the gating strategy, such as a particle sorter using flow cytometry (e.g., with a sorting decision module).

[0248] In some embodiments, the method includes sorting one or more cells in a sample. For example, the method may include sorting a sample having two or more components, such as three or more components, such as four or more components, such as five or more components, such as ten or more components, such as fifteen or more components, and includes sorting a sample having 25 or more components.

[0249] In sorting particles, the method includes data acquisition, analysis, and recording (e.g., via a computer), wherein multiple data channels record data from each detector used to acquire overlapping spectra of multiple fluorophores associated with the particle. In these embodiments, the analysis includes spectrally resolving the light from the multiple fluorophores with overlapping spectra associated with the particle (e.g., by calculating a spectral unmixing matrix) and identifying the particle based on the estimated abundance of each fluorophore associated with it. This analysis can be transmitted to a sorting system configured to generate a set of digitized parameters based on particle classification.

[0250] In some embodiments, a method for sorting sample components includes sorting cells of a sample using a particle sorting module with deflection plates, such as that described in U.S. Patent Publication No. 2017 / 0299493, filed March 28, 2017, the disclosure of which is incorporated herein by reference. In some embodiments, a sorting decision module having multiple sorting decision units is used to sort sample cells, such as that described in U.S. Patent Publication No. 2020 / 0256781, the disclosure of which is incorporated herein by reference. In some embodiments, the subject matter system includes a particle sorting module with deflection plates, such as that described in U.S. Patent Publication No. 2017 / 0299493, filed March 28, 2017, the disclosure of which is incorporated herein by reference.

[0251] Non-transitory computer-readable storage medium

[0252] This disclosure also includes non-transitory computer-readable storage media having instructions for practicing the methods of this subject matter. Computer-readable storage media can be used on one or more computers to achieve full or partial automation of the system for implementing the methods described herein. In some embodiments, instructions according to the methods described herein can be encoded onto a computer-readable medium in a “programmed” form, wherein the term “computer-readable medium” as used herein refers to any non-transitory storage medium that participates in providing instructions and data to a computer for execution and processing. Examples of suitable non-transitory storage media include floppy disks, hard disks, optical disks, magneto-optical disks, CD-ROMs, CD-Rs, magnetic tapes, non-volatile memory cards, ROMs, DVD-ROMs, Blu-ray discs, solid-state drives, and network-attached storage (NAS), whether these devices are internal or external to a computer. Files containing information can be “stored” on a computer-readable medium, wherein “stored” means recording the information so that it can be accessed and retrieved by a computer later. The computer-implemented methods described herein can be executed using a program that can be written in one or more of any number of computer programming languages. For example, these languages ​​include Python, Java, JavaScript, C, C#, C++, Go, R, Swift, PHP, and many other languages.

[0253] A non-transitory computer-readable storage medium with algorithmic instructions is also provided. According to certain embodiments, the non-transitory computer-readable storage medium includes: an algorithm for receiving two or more datasets, wherein each of the two or more datasets is associated with one of a plurality of data patterns; an algorithm for applying the algorithm to convert each of the two or more datasets into a feature vector based on the associated data patterns; an algorithm for receiving a classification task concerning the two or more datasets; and an algorithm for applying one or more of a plurality of classification models to the feature vectors based on the received classification task to determine a category. In some instances, the non-transitory computer-readable storage medium includes an algorithm for illuminating a sample containing cells in a flowing stream with a light source, and an algorithm for measuring the light from the illuminating cells using a light detection system with a photodetector.

[0254] In some embodiments, the multiple data patterns include image data patterns. In some instances, the algorithm is a rule-based image processing algorithm. In some instances, the algorithm includes a neural network. In some instances, the memory includes instructions for inverting the optical loss channels of a dataset associated with the image data pattern, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to zero mean, scaling the pixel values ​​of each channel of the dataset associated with the image data pattern to unit standard deviation, or a combination thereof. In some instances, the algorithm includes a sequence learning machine model. In some instances, the algorithm includes a compensation model, a spectral unmixing model, or a combination thereof. In some instances, the algorithm includes a calibration model.

[0255] In some embodiments, the classification task includes a single-cell classification task, a survival classification task, a whole blood classification task, a T-cell activation classification task, or any combination thereof. In some instances, the classification task is a single-cell classification task, and the categories include single-cell categories, adherent two-cell categories, separated two-cell categories, three-cell categories, or fragment categories. In some instances, the classification task includes a survival classification task, and the categories include a survival category, a death category, or an apoptosis category. In some instances, the classification task includes a whole blood classification task, and the whole blood classification includes a granulocyte category, a monocyte category, or a lymphocyte category. In some instances, the classification task includes a T-cell activation classification task, and the categories include an activated category or a non-activated category.

[0256] In some embodiments, the multiple classification models include end-to-end models. In some instances, the end-to-end models include uniform manifold approximation and projection (UMAP) models, T-distributed random nearest neighbor embedding (TSNE) models, principal component analysis (PCA) models, or any combination thereof. In some instances, the multiple classification models are trained using supervised methods. In some instances, the multiple classification models are trained using a weighted cross-entropy loss function.

[0257] In some embodiments, image data includes dark-field images, light-field images, fluorescently tagged images, and combinations thereof. In some instances, image data includes untagged images of cells. In some instances, image data includes autofluorescence image data. In some embodiments, sequence data includes photodetector waveforms. In some instances, sequence data includes raw image data in the form of raw waveforms. In some instances, image data is generated in real time from raw waveforms. In some embodiments, tabular data is spectral tabular data. In some instances, tabular data is spectral data generated in response to different wavelengths of light emitted by cells. In some instances, spectral tabular data is compensated or unmixed spectral data. In some instances, tabular data is scattering tabular data generated in response to scattered light from irradiated cells.

[0258] In some embodiments, the neural network includes a feature engineering layer for generating feature vectors from image data, sequence data, and tabular data of cells. In some instances, the feature engineering layer converts image data into image parameters of the cells. In some instances, the image parameters include one or more of cell size, cell diffusion rate, cell eccentricity, cell punctate density, cell shape, and cell morphology. In some instances, the feature engineering layer has an algorithm for extracting sequence data features from photodetector waveforms. In some instances, the feature engineering layer has an algorithm for spectrally unmixing spectral tabular data. In some instances, the feature engineering layer has an algorithm for calculating the abundance of biomarker molecules (e.g., surface biomarker portions). In some instances, the feature engineering layer has an algorithm for calculating fluorophore abundance. In some instances, the feature engineering layer has an algorithm for calculating cell physical measurements based on scattered light tabular data.

[0259] In some embodiments, the neural network includes a machine learning encoder layer. In some instances, the encoder layer is a rule-based image processing pipeline. In some instances, the encoder layer includes an image encoder. In some instances, the image encoder has an image recognition model architecture. In some instances, the image encoder has an encoder model architecture such as RestNet34, ShuffleNet V2, EfficientNet V2-S, Inception V3, and combinations thereof. In some instances, the encoder layer is trained using a supervised method with a weighted cross-entropy loss function. In some instances, the image is reconstructed in real-time from the data waveform. In some instances, the trained encoder layer is validated based on the evaluated feature quality. In some instances, metrics such as precision, recall, and F-1 score are used to determine the evaluated feature quality.

[0260] In some embodiments, the non-transitory computer-readable storage medium includes an algorithm that applies a neural network to data to classify cells based on one or more determined biological parameters. In some instances, the non-transitory computer-readable storage medium includes an algorithm for applying a dimensionality reduction algorithm to the generated data. In some instances, the non-transitory computer-readable storage medium includes an algorithm for applying a dimensionality reduction algorithm to high-dimensional feature vectors. In some instances, the dimensionality reduction algorithm is selected from Uniform Manifold Approximation and Projection (UMAP), t-Distributed Random Nearest Neighbor Embedding (t-SNE), Principal Component Analysis (PCA), and combinations thereof. In some instances, the non-transitory computer-readable storage medium includes an algorithm for classifying cells based on a gating hierarchy. In some instances, the non-transitory computer-readable storage medium includes an algorithm for classifying cells using a classification model (e.g., a single-layer perceptron, a multilayer perceptron, a random forest classifier, a support vector machine, and combinations thereof). In some instances, the non-transitory computer-readable storage medium includes an algorithm for classifying data using a survival rate classifier. In some instances, the non-transitory computer-readable storage medium includes an algorithm for classifying data using a cell type classifier. In some instances, the cell type classifier classifies cells into T cells, B cells, and natural killer (NK) cells. In some instances, the non-transitory computer-readable storage medium includes algorithms for classifying data using a single-cell identification classifier. In some instances, the non-transitory computer-readable storage medium includes algorithms for classifying data using a whole blood classifier. In some instances, the non-transitory computer-readable storage medium includes algorithms for classifying data using an activation classifier.

[0261] In some instances, the non-transitory computer-readable storage medium includes algorithms for determining one or more sorting gates for classifying cells in a sample. In some instances, the non-transitory computer-readable storage medium includes algorithms for computing one or more sorting gates that capture clusters of target cells and exclude clusters of non-target cells. In some instances, the non-transitory computer-readable storage medium includes algorithms for computing sorting gates that maximize the inclusion rate of clusters of target cells. In some instances, the non-transitory computer-readable storage medium includes algorithms for computing sorting gates that maximize the exclusion of clusters of non-target cells.

[0262] In some instances, the non-transitory computer-readable storage medium includes algorithms for sorting cells in a sample into multiple sample containers. In some instances, the non-transitory computer-readable storage medium includes algorithms for sorting cells based on the presence, classification, or both. In some instances, the non-transitory computer-readable storage medium includes algorithms for sorting cells based on generated cell images. In some instances, the non-transitory computer-readable storage medium includes algorithms.

[0263] Non-transitory computer-readable storage media can be applied to one or more computer systems having a display and operator input devices. For example, operator input devices may be a keyboard, mouse, etc. The processing module includes a processor that can access memory on which instructions for performing the steps of the methods of this subject are stored. The processing module may include an operating system, a graphical user interface (GUI) controller, system memory, memory storage devices and input / output controllers, caches, data backup units, and many other devices. The processor may be a commercially available processor or one of other processors that are already available or not. The processor executes the operating system, which interacts with firmware and hardware in well-known ways and helps the processor coordinate and execute the functions of various computer programs, which may be written in various programming languages, such as those described above, other high-level or low-level languages, and combinations thereof, as known in the art. The operating system typically works with the processor to coordinate and execute the functions of other computer components. The operating system also provides scheduling, input / output control, file and data management, memory management, and communication control and related services, all conforming to known techniques.

[0264] kit

[0265] Various aspects of this disclosure also include kits, wherein the kits include one or more integrated circuits including those described herein. In some embodiments, the kits may also include instructions for programming the system of this subject matter, such as in the form of computer-readable media (e.g., flash drives, USB storage, optical discs, DVDs, Blu-ray discs, etc.) or for downloading programming from the Internet web protocol or a cloud server. The kits may also include instructions for practicing the methods of this subject matter. These instructions may exist in a variety of forms in the subject matter kits, one or more of which may exist in the kits. One form in which these instructions may exist is as printed information on a suitable medium or substrate (e.g., one or more sheets of paper with information printed on them), in kit packaging, in packaging inserts, etc. Another form in which these instructions are recorded on a computer-readable medium, such as a floppy disk, optical disc (CD), portable flash drive, etc. Yet another form in which these instructions may exist is a URL that can be used to access information on a removed site via the Internet.

[0266] practicality

[0267] The systems, methods, and computer systems described herein can be used for upstream cell evaluation, such as in applications requiring the use of cells with specific biological parameters, for example, molecular biology assays (e.g., gene expression analysis). In some instances, this disclosure provides high-quality and pure cells isolated from heterogeneous samples. Furthermore, the systems and methods described herein can be used in a variety of applications requiring the analysis and sorting of particle components in samples within a fluid medium. In some embodiments, the systems and methods described herein can be used for flow cytometry characterization of cell isolates. Embodiments of this disclosure are suitable for flow cytometers that desire to provide improved cell sorting accuracy, enhanced particle collection, particle charging efficiency, more precise particle charging, and enhanced particle deflection during cell sorting.

[0268] Embodiments of this disclosure are also applicable to applications where cells prepared from biological samples are desired for use in research, laboratory testing, or therapy. In some embodiments, the methods and apparatus of this subject matter can facilitate the preparation of single cells from a target fluid or tissue biological sample. For example, the methods and systems of this subject matter facilitate the acquisition of cells from fluid or tissue samples for use as research or diagnostic samples. Compared to conventional flow cytometry systems, the methods and apparatus of this disclosure allow for the separation and collection of cells from biological samples (e.g., organs, tissues, tissue fragments, fluids) with high efficiency and low cost.

[0269] Although the invention has been described in detail by way of illustrations and examples for the purpose of clarity, it will be apparent to those skilled in the art, based on the teachings of the invention, that certain changes and modifications may be made thereto without departing from the spirit or scope of the appended claims.

[0270] Therefore, the foregoing has only illustrated the principles of the invention. It should be understood that those skilled in the art will be able to design various devices that, although not explicitly described or shown herein, embody the principles of the invention and are included within its spirit and scope. Furthermore, all examples and conditional language listed herein are primarily intended to assist the reader in understanding the principles of the invention and the concepts contributed by the inventors to further developments in the field, and should be understood as not being limited to these specifically listed examples and conditions. In addition, all statements herein referencing the principles, aspects, and embodiments of the invention, as well as specific examples thereof, are intended to cover their structural and functional equivalents. Furthermore, such equivalents are contemplated to include both currently known equivalents and future-developed equivalents (i.e., any element developed that performs the same function, regardless of its structure). Moreover, nothing disclosed herein is intended to be offered to the public, whether or not such disclosure is expressly stated in the claims.

[0271] Therefore, the scope of the invention is not intended to be limited to the exemplary embodiments shown and described herein. Rather, the scope and spirit of the invention are embodied in the appended claims. In the claims, with respect to the limitation in the claims, 35U.SC §112(f) or 35U.SC §112(6) is explicitly defined as being invoked only when such limitation in the claims begins to refer to the exact phrase “means for…” or the exact phrase “step for…”; if such an exact phrase is not used in the limitation in the claims, then 35U.SC §112(f) or 35U.SC §112(6) is not invoked.

Claims

1. A system comprising: A light source, configured to illuminate the cells of a sample in a flowing stream; A light detection system, comprising a photodetector, for detecting light from an irradiated cell; A processor, comprising memory operatively coupled to the processor, wherein the memory includes instructions stored thereon, the instructions causing the processor, when executed by the processor, to: Receive two or more datasets, wherein each of the two or more datasets is associated with one of a plurality of data patterns; Apply algorithms to transform each of two or more datasets into a feature vector based on associated data patterns; Receive a classification task for the two or more datasets; and Based on the received classification task, one or more of multiple classification models are applied to the feature vector to determine the category.

2. The system according to claim 1, wherein, The multiple data modes include image data modes.

3. The system according to any one of claims 1 to 2, wherein, The algorithm includes rule-based image processing algorithms.

4. The system according to any one of claims 1 to 3, wherein, The algorithm includes a neural network.

5. The system according to any one of claims 2 to 4, wherein, The memory includes instructions to: Reverse the optical loss channel of the dataset associated with the image data pattern; Scale the pixel values ​​of each channel of the dataset associated with the image data pattern to zero mean; Scale the pixel values ​​of each channel of the dataset associated with the image data pattern to one standard deviation; or The combination of the above operations.

6. The system according to any one of claims 1 to 5, wherein, The multiple data modes include waveform data modes.

7. The system according to any one of claims 1 to 6, wherein, The algorithm includes a sequence learning machine model.

8. The system according to any one of claims 1 to 7, wherein, The multiple data modes include spectral data modes.

9. The system according to claim 8, wherein, The algorithm includes a compensation model, a spectral unmixing model, or both.

10. The system according to any one of claims 1 to 9, wherein, The multiple data modes include scattering data modes.

11. The system according to any one of claims 1 to 10, wherein, The algorithm includes a calibration model.

12. The system according to any one of claims 1 to 11, wherein, The classification tasks include single-cell classification tasks, survival rate classification tasks, whole blood classification tasks, T-cell activation classification tasks, or any combination thereof.

13. The system according to claim 12, wherein, The classification task includes the single-cell classification task, and the categories include single-cell categories, adherent two-cell categories, separated two-cell categories, three-cell categories, or fragment categories.

14. The system according to claim 12, wherein, The classification task includes the survival classification task, and the categories include survival category, death category, or apoptosis category.

15. The system according to claim 12, wherein, The classification task includes the whole blood classification task, and the categories include granulocyte category, monocyte category, or lymphocyte category.

16. The system according to claim 12, wherein, The classification task includes the T cell activation classification task, wherein the category includes an activated category or an inactivated category.

17. The system according to any one of claims 1 to 16, wherein, The multiple classification models include end-to-end models.

18. The system according to claim 17, wherein, The end-to-end model includes a gated hierarchical structure, a single-layer classifier, a multi-layer classifier, a random forest classifier, a support vector machine classifier, or any combination thereof.

19. The system according to any one of claims 1 to 16, wherein, The multiple classification models include exploratory models.

20. The system according to claim 19, wherein, The exploratory models include the uniform manifold approximation and projective UMAP model, the T-distributed random nearest neighbor embedding (TSNE) model, the principal component analysis (PCA) model, or any combination thereof.

21. The system according to any one of claims 1 to 20, wherein, One or more of the multiple classification models are trained using a supervised method.

22. The system according to any one of claims 1 to 21, wherein, One or more of the multiple classification models are trained using a weighted cross-entropy loss function.

23. A method comprising: Receive two or more datasets, wherein each of the two or more datasets is associated with one of a plurality of data patterns; Apply algorithms to transform each of two or more datasets into a feature vector based on associated data patterns; Receive a classification task for the two or more datasets; and Based on the received classification task, one or more of multiple classification models are applied to the feature vector to determine the category.

24. The method according to claim 23, wherein, The multiple data modes include image data modes.

25. The method according to any one of claims 23 to 24, wherein, The algorithm includes rule-based image processing algorithms.

26. The method according to any one of claims 23 to 25, wherein, The algorithm includes a neural network.

27. The method according to any one of claims 24 to 26, further comprising: Reverse the optical loss channel of the dataset associated with the image data pattern; Scale the pixel values ​​of each channel of the dataset associated with the image data pattern to zero mean; Scale the pixel values ​​of each channel of the dataset associated with the image data pattern to one standard deviation; or Any combination of the above operations.

28. The method according to any one of claims 23 to 27, wherein, The multiple data modes include waveform data modes.

29. The method according to any one of claims 23 to 28, wherein, The algorithm includes a sequence learning machine model.

30. The method according to any one of claims 23 to 29, wherein, The multiple data modes include spectral data modes.

31. The method according to claim 30, wherein, The algorithm includes a compensation model, a spectral unmixing model, or both.

32. The method according to any one of claims 23 to 31, wherein, The multiple data modes include scattering data modes.

33. The method according to any one of claims 23 to 32, wherein, The algorithm includes a calibration model.

34. The method according to any one of claims 23 to 33, wherein, The classification tasks include single-cell classification tasks, survival rate classification tasks, whole blood classification tasks, T-cell activation classification tasks, or any combination thereof.

35. The method according to claim 34, wherein, The classification task includes the single-cell classification task, and the categories include single-cell categories, adherent two-cell categories, separated two-cell categories, three-cell categories, or fragment categories.

36. The method according to claim 34, wherein, The classification task includes the survival classification task, and the categories include survival category, death category, or apoptosis category.

37. The method according to claim 34, wherein, The classification task includes the whole blood classification task, and the categories include granulocyte category, monocyte category, or lymphocyte category.

38. The method according to claim 34, wherein, The classification task includes the T cell activation classification task, wherein the category includes an activated category or an inactivated category.

39. The method according to any one of claims 23 to 38, wherein, The multiple classification models include end-to-end models.

40. The method according to claim 39, wherein, The end-to-end model includes a gated hierarchical structure, a single-layer classifier, a multi-layer classifier, a random forest classifier, a support vector machine classifier, or any combination thereof.

41. The method according to any one of claims 23 to 38, wherein, The multiple classification models include exploratory models.

42. The method according to claim 41, wherein, The exploratory models include the uniform manifold approximation and projective UMAP model, the T-distributed random nearest neighbor embedding (TSNE) model, the principal component analysis (PCA) model, or any combination thereof.

43. The method according to any one of claims 23 to 42, wherein, One or more of the multiple classification models are trained using a supervised method.

44. The method according to any one of claims 23 to 43, wherein, One or more of the multiple classification models are trained using a weighted cross-entropy loss function.

45. A non-transitory computer-readable storage medium comprising instructions stored thereon, wherein, The non-transitory computer-readable storage medium includes: An algorithm for receiving two or more datasets, wherein each of the two or more datasets is associated with one of a plurality of data patterns; An algorithm used to apply algorithms to transform each of two or more datasets into a feature vector based on associated data patterns; An algorithm for receiving a classification task about the two or more datasets; and An algorithm for determining a category by applying one or more of multiple classification models to a feature vector based on a received classification task.

46. ​​The non-transitory computer-readable storage medium according to claim 45, wherein, The multiple data modes include image data modes.

47. The non-transitory computer-readable storage medium according to any one of claims 45 to 46, wherein, The algorithm includes rule-based image processing algorithms.

48. The non-transitory computer-readable storage medium according to any one of claims 45 to 47, wherein, The algorithm includes a neural network.

49. The non-transitory computer-readable storage medium according to any one of claims 46 to 48, wherein, The memory includes instructions to: Reverse the optical loss channel of the dataset associated with the image data pattern; Scale the pixel values ​​of each channel of the dataset associated with the image data pattern to zero mean; Scale the pixel values ​​of each channel of the dataset associated with the image data pattern to one standard deviation; The combination of the above operations.

50. The non-transitory computer-readable storage medium according to any one of claims 45 to 49, wherein, The multiple data modes include waveform data modes.

51. The non-transitory computer-readable storage medium according to any one of claims 45 to 50, wherein, The algorithm includes a sequence learning machine model.

52. The non-transitory computer-readable storage medium according to any one of claims 45 to 51, wherein, The multiple data modes include spectral data modes.

53. The non-transitory computer-readable storage medium according to claim 52, wherein, The algorithm includes a compensation model, a spectral unmixing model, or both.

54. The non-transitory computer-readable storage medium according to any one of claims 45 to 53, wherein, The multiple data modes include scattering data modes.

55. The non-transitory computer-readable storage medium according to any one of claims 45 to 54, wherein, The algorithm includes a calibration model.

56. The non-transitory computer-readable storage medium according to any one of claims 45 to 55, wherein, The classification tasks include single-cell classification tasks, survival rate classification tasks, whole blood classification tasks, T-cell activation classification tasks, or any combination thereof.

57. The non-transitory computer-readable storage medium according to claim 56, wherein, The classification task includes a single-cell classification task, wherein the categories include single-cell categories, adherent two-cell categories, separated two-cell categories, three-cell categories, or fragment categories.

58. The non-transitory computer-readable storage medium according to claim 56, wherein, The classification task includes a survival classification task, wherein the categories include a survival category, a death category, or an apoptosis category.

59. The non-transitory computer-readable storage medium according to claim 56, wherein, The classification task includes the whole blood classification task, and the categories include granulocyte category, monocyte category, or lymphocyte category.

60. The non-transitory computer-readable storage medium according to claim 56, wherein, The classification task includes the T cell activation classification task, wherein the category includes an activated category or an inactivated category.

61. The non-transitory computer-readable storage medium according to any one of claims 56 to 60, wherein, The multiple classification models include end-to-end models.

62. The non-transitory computer-readable storage medium according to claim 61, wherein, The end-to-end model includes a gated hierarchical structure, a single-layer classifier, a multi-layer classifier, a random forest classifier, a support vector machine classifier, or any combination thereof.

63. The non-transitory computer-readable storage medium according to any one of claims 45 to 60, wherein, The multiple classification models include exploratory models.

64. The non-transitory computer-readable storage medium according to claim 63, wherein, The exploratory models include the uniform manifold approximation and projective UMAP model, the T-distributed random nearest neighbor embedding (TSNE) model, the principal component analysis (PCA) model, or any combination thereof.

65. The non-transitory computer-readable storage medium according to any one of claims 45 to 64, wherein, One or more of the multiple classification models are trained using a supervised method.

66. The non-transitory computer-readable storage medium according to any one of claims 45 to 65, wherein, One or more of the multiple classification models are trained using a weighted cross-entropy loss function.

Images