Cytometric analysis method

A density-based topological persistence method for cytometry data analysis addresses the challenges of reliability and reproducibility in automated cytometry, offering transparent and efficient segmentation of complex data sets.

JP2024543173APending Publication Date: 2024-11-19メタフォーラ バイオシステムス +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2024531549
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2021-11-25
Filing Date
2022-11-25
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

Current cytometry data analysis methods are unreliable, reproducible, and require significant human intervention, leading to cumbersome and subjective results, especially as the number of cell markers increases, making automated techniques difficult to adopt in clinical and research settings.

Method used

A computer-implemented method for analyzing cytometry data using a density-based approach that employs topological persistence to create a hierarchical structure, allowing for automatic segmentation and interpretation of complex multidimensional data sets, ensuring reproducibility and explainability.

Benefits of technology

The method provides reliable, reproducible, and easily interpretable cytometry data analysis, enabling robust automatic segmentation that meets clinical and research standards by maintaining transparency and reducing human subjectivity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024543173000001_ABST
    Figure 2024543173000001_ABST
Patent Text Reader

Abstract

The present invention relates to a device and a computer-implemented method for analyzing a dataset related to a plurality of biological objects selected from cells, cell-derived vesicles, cell-free microorganisms, and / or biofunctionalized materials, said dataset comprising N cytometric events each related to a biological object, each cytometric event being defined by at least two cytometric parameters measured for a corresponding biological object, such that the dataset is represented by a point cloud of N points in a D-dimensional space, said device and method being configured to output at least a hierarchical structure representing different classes of biological objects and their interrelationships.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to the general field of cytometric data analysis. [Background technology]

[0002] Protein sequencing, then DNA sequencing, and the advent of automated sequencers have revolutionized biology. Classical descriptive and reductionist approaches (one gene, one messenger RNA, one protein) have been succeeded by a more comprehensive understanding of biological systems based on the analysis of sets of biological elements ("-omes"). The basic idea associated with the "omics" approach is to grasp the complexity of the whole organism using a methodology that is as unrestricted as possible in its descriptive scope.

[0003] Such approaches primarily include genomics (study of genes), transcriptomics (analysis of gene expression and its regulation), proteomics (study of proteins), metabolomics (analysis of metabolic products), and metabonomics (study of in vivo metabolic profiles).

[0004] Genomics is divided into two branches: structural genomics, which focuses on whole genome sequencing, and functional genomics, which aims to determine the function and expression of sequenced genes. In functional genomics, techniques are applied in parallel to large numbers of genes, for example, the phenotype of mutants can be analyzed for entire gene families or for the expression of all genes in an entire organism.

[0005] Transcriptomics is the study of all messenger RNAs produced during the process of transcription of the genome. Transcriptomics relies on the quantification of all these messenger RNAs, providing a relative indication of the transcription rates of different genes under given conditions.

[0006] Proteomics is the analysis of all proteins in an organelle, cell, tissue, organ or organism under given conditions. Proteomics aims to identify proteins extracted from cell cultures, tissues or biological fluids, their localization in cellular compartments, their possible post-translational modifications, as well as the amount of proteins in a global manner. Proteomics makes it possible to quantify the variations in their expression levels, for example, depending on time, environment, developmental stage, physiological and pathological conditions, species origin, etc. Proteomics also studies the interactions of proteins with other proteins, DNA or RNA, or other substances.

[0007] Metabolomics studies all the metabolic substrates, intermediates, products, and products (sugars, amino acids, fatty acids, etc.) present in a cell, organ or organism, including hormones, other signaling molecules, and secondary metabolites.

[0008] Metabonomics involves monitoring in vivo metabolic profiles to provide information about drug toxicity, pathological processes, and gene function.

[0009] Cytomics corresponds to the analysis of the cytome, which is the collection of cellular components that contains morphological, antigenic and functional properties that allow to define or model the state and function of a cell at a given time. Cytome analysis makes it possible to describe the structural and functional heterogeneity of the various cells in an organism.

[0010] The aforementioned approaches provide a wealth of information regarding the response of cells and / or tissues to in vitro or in vivo exposure. The aforementioned approaches may be particularly useful for identifying and characterizing new biomarkers (for diagnosis, susceptibility, prognosis, exposure, effect), generating new mechanistic knowledge (mode of action), or developing new tools for predictive efficacy or toxicology to aid in the identification of new therapeutic targets or candidate drugs or vaccines.

[0011] The automation of sequencing techniques and the development of high-throughput technologies, especially enabled by the emergence of specialized technological platforms, have allowed the industrialization of data generation and the simultaneous analysis of multiple variables.

[0012] This results in vast amounts of data that must be processed, analyzed, visualized, and interpreted in the most informative way possible to extract the maximum amount of information about the biological process or system under study.

[0013] From a biostatistical point of view, the data obtained by "omics" approaches concern a large number of variables that need to be analyzed together. For example, transcriptome analysis makes it possible to study the expression of thousands of genes simultaneously.

[0014] It is therefore desirable to have powerful biostatistical and bioinformatics tools for processing, analyzing and interpreting the large amounts of data generated by "omics" approaches.

[0015] There are numerous techniques for acquiring "omics" data, such as mass spectrometry techniques, chromatography, sequencing, spectral cytometry, mass cytometry, flow cytometry, image cytometry, and the like.

[0016] Cytometry refers to a set of techniques for analyzing biological samples that contain a set of "biological elements" such as cells, vesicles, or particles in suspension.

[0017] Cytometric techniques are essential analytical techniques for the identification and characterization of populations of cells, vesicles, or particles. Moreover, as cytomics provides a single-cell view of more complex cellular conglomerates such as tissues of multicellular organisms, or communities in the case of unicellular bacteria, yeast, algae, etc., cytomics can address questions of intra- or extracellular processes and cell-cell interactions.

[0018] Although reliability and reproducibility of results are of paramount importance, a major source of variability in cytometry lies in data analysis.

[0019] Traditional cytometry data analysis typically involves sequential manual selection (or "gating") of regions of interest in what are usually two-dimensional scatter or contour plots, representing one parameter on each axis. This is the case, for example, when analyzing flow cytometry data. While the analysis is straightforward for three- or four-color immunofluorescence data, it becomes much more complex when the number of cell markers examined increases, increasing the variability and reproducibility issues associated with human operators.

[0020] Current state-of-the-art cytometry techniques can measure tens, hundreds, or even thousands of parameters, thus generating complex, multi-dimensional data sets that make manual analysis time-consuming, and therefore manual segmentation is subjective, tedious, unreliable, and expensive.

[0021] Segmentation can be automated without operator supervision, using software to define the segments. It can involve artificial intelligence, but the criteria for grouping cytometry events will no longer be understandable by the operator. Alternatively, it can rely on analytical algorithms.

[0022] Several comparative studies have evaluated unsupervised clustering methods based on their ability to reproduce manual selection (or segmentation or "gating") and detect rare cell populations, as well as their execution time. Although some methods may be effective for specific applications, none of them meets all the requirements to be reliably used by users on a daily basis.

[0023] Moreover, automated techniques allow for cytometric analysis free of manual variability, subjectivity, and segmentation bias, and thus new methods have been developed in this field over the past decade. However, many automated techniques still require parameterization of algorithms that require operator intervention, introducing a significant amount of human subjectivity and therefore resulting in non-reproducible results.

[0024] As a result, adoption of automated analysis among researchers in academia, biotechnology, pharmaceutical industry, or clinical research has been slow, and manual segmentation remains the default method and standard.

[0025] The main reasons why clinical centers do not use automated analysis have recently been identified as a lack of trust or understanding of the algorithms used, and a lack of resources.

[0026] For automated analysis techniques to replace manual segmentation, not only must the identification of cell populations correspond to those of an expert manual analysis, but the results from the algorithm must also be robust, meaning fully reproducible.

[0027] In the specific example of flow cytometry, a biological fluid containing cells, vesicles or particles to be analyzed is passed continuously past a detector, and for each cell, vesicle or particle that passes in front of the detector, the value of a cytometric parameter can be measured.

[0028] In general, a "cytometry event" is a particle-associated vector that regularly collects measurements of a cytometry parameter of that particle.

[0029] Conventionally, a cytometry event collects more than ten values ​​of a cytometry parameter for each particle. Each value of a cytometry parameter for a particle is called a "cytometry measurement."

[0030] To analyze all cytometric measurements, the points representing each cytometric event can be presented in one-, two- or three-dimensional space on a "presentation graph". Each dimension corresponds to a cytometric parameter. The graph representing the cytometric events is thus a "projection" that allows one, two or three cytometric measurements to be visualized for each cytometric event.

[0031] To facilitate the analysis, the point cloud needs to be "segmented", i.e. cytometric events having similar values ​​for some cytometric parameters are grouped together or "segments" are created. Analysis of the segments thus obtained, for example of their respective weights, makes it possible to characterize the biological fluid, for example to detect biological anomalies or pathological risks.

[0032] When segmentation is performed manually, the operator defines the contours of the segments on a presented graph.

[0033] Replacing manual segmentation with unsupervised automatic segmentation remains a major problem in the field of cytometry. To evaluate the quality of automatic segmentation methods, the following four criteria are commonly studied:

[0034] a) Correlation with manual segmentation:

[0035] The first criterion is a performance criterion, i.e., a strong correlation with manual segmentations resulting from expert consensus on a known dataset: experts must be able to easily find the main populations of interest in the automatic segmentation performed by the algorithm.

[0036] b) Considering all dimensions simultaneously:

[0037] The second criterion concerns the simultaneous consideration of all dimensions of the data, or a set of dimensions within them. Indeed, in manual segmentation, segments are created on the data projected, at most, on triplets, on marker pairs. By limiting to this type of representation, some of the richness and complexity of the data and the structures that make up the data are not taken into account. By considering multiple or all dimensions simultaneously, it becomes possible to take into account all the complexity of the data and, for example, to discover populations that are not identified by manual segmentation.

[0038] c) Segmentation Hierarchy:

[0039] When an expert manually segments a cytometry file, the expert proceeds recursively by identifying populations with increasing granularity. This hierarchical approach is particularly interesting because biology itself classifies cells according to a hierarchical approach. Hierarchical segmentation makes it possible to adapt the level of granularity according to the expert's needs.

[0040] d) Explainability and robustness of the algorithm:

[0041] Explainability of algorithms is a key point in artificial intelligence, and even more so in medical or research applications that require strong theoretical guarantees. Moreover, robustness and reproducibility of results on two identical data sets are also major challenges in medical and diagnostic applications.

[0042] Many methods exist for automatic and unsupervised segmentation of cytometry data, however none of them fully meet the four criteria outlined above and none have been implemented in solutions licensed for sale on regulated markets.

[0043] Regarding correlation with manual segmentation (a), most methods have some level of correlation with expert segmentation, but few achieve a stable level of performance and consistency depending on the existing diversity of the dataset. The methods that provide the best results generally do not meet the explainability requirements.

[0044] Regarding the simultaneous consideration of all dimensions (b), most algorithms use all dimensions simultaneously. However, some methods do not use all dimensions simultaneously, e.g., they simply project the data onto a given pair of markers. These methods have the advantage of closely matching what experts do, but provide little added value in discovering more related populations.

[0045] Regarding hierarchical segmentation (c), state-of-the-art algorithms rarely naturally generate hierarchical segmentations. In fact, in statistical learning in general, segmentation methods (clustering) are rarely hierarchical. This bias directly affects cytometric segmentation methods derived from statistical learning.

[0046] To overcome this problem, a hierarchical segmentation is generally generated after the initial segmentation. This approach consists in first resorting to an algorithm for a number of clusters (i.e. grouping), and then applying the hierarchical segmentation method directly to the clusters, and no longer to the points. This approach generates a hierarchical segmentation, but has the drawback of separating the procedure into two separate times, which tends to produce artifact effects. Furthermore, the stability of the hierarchical structure depending on the number of groups initially requested by the method becomes an issue.

[0047] Finally, with regard to the explainability and robustness of the algorithms (d), few methods still meet the criteria. On the one hand, many algorithms rely on random initialization and / or parameters. Therefore, it is usually necessary to rerun these algorithms several times to ensure the results. On the other hand, the explainability of algorithms remains a major issue, and in general, the most explainable algorithms are also the simplest and often the poorest performing.

[0048] Furthermore, regardless of the segmentation method, the segments resulting from the analysis of one projection should be comparable to the segments resulting from the analysis of other projections. For example, after isolating in a first projection those cytometric events that exhibit values ​​within a given range of a first and a second cytometric parameter, thus creating a segment of interest, it may be useful to specifically represent in a second projection the cytometric events of this segment of interest.

[0049] The efficiency of the analysis is closely related to the ability of the operator to make such comparisons. Summary of the Invention [Problem to be solved by the invention]

[0050] As a result, there is a need for a method for analyzing cytometric measurements that implements automated segmentation, provides reliable and reproducible results, is easy and robust to implement, and is human understandable in operation (which may allow for human intervention or user validation if necessary).

[0051] The present invention is directed, at least in part, to addressing this need.

[0052] The present invention also allows the user to obtain "on demand" the level of segmentation precision in particular regions of the data structure. [Means for solving the problem]

[0053] Accordingly, the present invention relates to a computer-implemented method for analyzing a dataset associated with a plurality of biological objects selected from cells, cell-derived vesicles, acellular microorganisms, and / or biofunctionalized materials, the dataset comprising N cytometric events each associated with a biological object, each cytometric event being defined by at least two cytometric parameters measured for a corresponding biological object, such that the dataset is represented by a point cloud (X) of N points in a D-dimensional space.

[0054] The method comprises: -For each point (x i ), the point (x i ) and the points to be considered (x i ) is a set of p points in the point cloud (X) that are closest to the neighboring point (V p (x i The density (w(x i ) determining -Point cloud (X) to modal segment (M j ), wherein each modal segment (M j ) locally exhibits a maximum density at the modal point (m j ), and the points in the point cloud are modal points (m j ) and the points in the basin are scaled by their density (w(x i )) recursively identifying -Each modal segment (M j ), determining a persistence (β) for Modal segment (M j ) adjacent to the modal segment (M t ) is the modal point (m j ) has no points with a density greater than the density of ; if two or more modal segments are adjacent when they share the same density saddle point greater than 0, then it is an infinite value, otherwise Modal segment (M j ) and at least one adjacent modal segment (M t ), which represents the depth of the density saddle point between Steps and -Modal segment (M d ) and The considered modal segment (M j ) and its adjacent modal segments (M t ) to identify the hierarchical link between the modal segment (M j ) and its adjacent modal segments (M t ) which corresponds to the highest basin, The considered modal segment (M j ) and the considered modal segment (M j ) and modal segments whose persistence is below a certain persistence threshold are merged to obtain a modal point (m j ) a parent modal segment having Repeat actions a) and b) iteratively by starting a new iteration from the parent modal segment, A step of determining a hierarchical structure defined based on persistence, the hierarchical structure (L k ) contains multiple levels, and each level (L k ) in multiple classes (C kj ), where each class has one or more modal segments (M j ), representing a group of multiple biological objects; - outputting at least a hierarchical structure representing different classes of biological objects and their interrelationships; Includes.

[0055] As will be seen in more detail in the following description, the method advantageously uses a deterministic approach. In fact, the method utilizes a density calculation based on the distance between the points of the point cloud representing the entire data set under study. The density calculated for a given point of the point cloud therefore represents the concentration of the neighboring points around the point. The method described herein is therefore different from other methods that use random metrics. Thus, for identical or nearly identical point clouds, substantially identical results are obtained. Thus, reproducibility and rigor of the results are guaranteed.

[0056] Moreover, unlike artificial intelligence models, also known as "black box", and other methods that have steps of a random nature or that use random metrics (such as a random point selection step), the algorithm implemented in the method of the present invention is based on the analysis of cytometric data density, allowing explainability and traceability of the proposed results (i.e. modal segments within a hierarchical structure), which is particularly suitable for clinical analysis, where the principle of transparency is a fundamental prerequisite.

[0057] Finally, tests have shown that the segmentation criteria of the proposed method are particularly relevant for cytometric analysis.

[0058] Furthermore, the method of the present invention makes it possible to perform the segmentation taking into account all dimensions present in the data (i.e. a set of cytometric data related to multiple biological objects). Advantageously, the hierarchical structure provided as output by the method allows for the representation of extreme variability in biological samples.

[0059] Furthermore, the method steps are designed to replicate the hierarchical and multi-scale aspects of the data using topological analysis, which advantageously allows obtaining a detailed representation of the native hierarchical structure of the cytometry-derived data, a representation that differs from, for example, simple neighborhood graphs, which aim to replicate the overall topology of the point cloud and incorporate the notion of proximity between clusters, but not the notion of hierarchy.

[0060] More specifically, the method described herein utilizes the calculated density values ​​to reproduce hierarchical and multi-scale aspects. Thus, in a hierarchical structure with multiple levels ranked from highest to lowest, high density points are represented, for example, at higher levels and low density points are represented at lower levels in the hierarchical structure. The density values ​​of the points are used and exploited to place the points within the hierarchical structure, and possibly within a representation of this hierarchical structure. Thus, in the method described herein, two points with very different densities are not placed at the same level. Typically, rare cells, i.e. cells with low density, are located at lower levels of the hierarchy. In other words, the hierarchical structure provided as an output of the method converts the parent-child relationships between the different levels of classes. This allows for the exploration of these different levels and the classes that make up each level.

[0061] The method described herein is not intended to ignore the relative density differences between points in a point cloud. On the contrary, the method utilizes these relative differences to transform the hierarchical (i.e., parent-child) relationships between different levels. The method thus aims to reproduce the global topology of the point cloud, for example by identifying clusters of points that exhibit similarity, but at the same time is distinct from and incompatible with other methods that erase the relative density differences between points of a point cloud, for example by density-adaptive sampling. In the method described herein, the density calculations performed are used to obtain a hierarchical structure, but not for sampling purposes.

[0062] By way of illustration, if the space is two-dimensional, the points lie in a plane. When representing the density on the third dimension orthogonal to the plane, each modal segment corresponds to a mountain, whose peak is the local maximum density associated with the modal point, and the valley between two adjacent mountains contains a density saddle point. A saddle point is the set of lowest points between two mountains that belong to the same ridge. The persistence of a mountain measures the difference in altitude between the bottom of the saddle point, which has a density greater than 0, and the peak of this mountain. If the persistence of a modal segment is low, it can generally be represented by a bulge that protrudes slightly from the neighboring mountain, for example on the side of this neighboring mountain. The merging step allows, for example, the merging of such a bulge with this neighboring mountain. As a result, the method of the present invention advantageously avoids placing importance on disturbances of the density function caused by noise. Noise is associated with low persistence values ​​and is therefore merging with neighboring modal segments due to the rational selection of the persistence threshold.

[0063] In one embodiment, the two parameters p and q are predetermined by heuristic studies on a large amount of data. Advantageously, the selection of these parameters is therefore operator independent, ensuring the method has operator-independent analytical robustness.

[0064] The threshold persistence (i.e., persistence threshold) associated with a level determines the number of classes at that level, i.e. the granularity of the representation. The value of the threshold persistence makes it possible to select the number of classes associated with each level of the hierarchical structure. In other words, the selection of the threshold persistence value makes it possible to simultaneously study or analyze different levels of the hierarchical structure, for example, representing the general population formed by all biological objects studied on the one hand, and subpopulations within the general population on the other hand. The described method makes it possible to explore the obtained hierarchical structure by defining different granularities for different areas of the hierarchical structure. In this aspect, the method described herein differs from other methods that allow only one representational granularity to be selected, i.e., a uniform granularity. Thus, the method described herein does not only aim to determine several homogeneous main groups, but to access multi-scale data, i.e. data that includes subsets of data represented at different scales on the same representation. Access to multi-scale data, as in the method described herein, makes it possible to access all data at any time without losing information. This allows the user to access at any time data that he or she selects and that corresponds to a subpopulation that is potentially more interesting and relevant to the user.

[0065] In one embodiment, the modal segment (M j The step of segmenting a point cloud (X) into vertices (x i ), proceeding from the point with the highest density to the point with the lowest density, -point(x i ) is the density of its q nearest neighbors (V q (x i )) and -Points to consider (x i ) is the density of its q nearest neighbors (V q (x i )), the point (x i ) to include the modal segment (M j ), otherwise, -Points to consider (x i ) to its q nearest neighbors (V q (x i )) in the modal segment that contains the point with the highest density Includes steps.

[0066] In one embodiment, the modal segment (M j ) adjacent to one or more modal segments (M t ), persistence (β) is the modal point (m j ) density and modal segment (M j ) and at least one adjacent modal segment (M t ) and the highest density of the points within the density saddle point.

[0067] In one embodiment, the method includes a preliminary step of receiving a dataset relating to a plurality of biological objects.

[0068] In one embodiment, the biological subject is - cells of animal, plant, fungal, protist, bacterial or archaeal origin; - a cell-origin vesicle selected from exosomes, ectosomes, microvesicles, microparticles, prostasomes, oncosomes, matrix / calcified vesicles, or apoptotic bodies; - acellular microorganisms selected from viruses, viroids and prions, and / or a biofunctionalized material comprising a synthetic or biological material selected from nanoparticles (e.g. nanobeads, nanospheres, or nanocapsules), microparticles (e.g. microbeads, microspheres, or microcapsules), lipid vesicles (e.g. unilamellar vesicles, multilamellar vesicles, lipoplexes, polyplexes, lipopolyplexes, liposomes, niosomes, cochleates, virosomes, immunostimulating complexes (ISCOM®)), wherein the synthetic or biological material is coupled to or coated with one or more peptides, proteins, antibodies, antibody fragments, receptors, cytokines, chemokines, toxins, oligonucleotides, colored or fluorescent molecules, amines, carboxyl or hydroxyl groups, bioactive molecules (e.g. immunomodulatory molecules, small molecule drugs, peptidomimetics, drugs), biotin, avidin or streptavidin molecules, or combinations thereof; It is.

[0069] In one embodiment, the cytometric parameters are selected from the size of the biological objects, their density, granularity, morphology, shape, refractive index, membrane composition, molecular content, molecular abundance, and / or molecular expression level.

[0070] In one embodiment, one of the cytometric parameters measured is the expression level of one or more proteins, receptors, markers, and / or the expression level of one or more nucleic acids, such as DNA or RNA.

[0071] In one embodiment, the cytometry data is obtained by flow cytometry (or FACS for "fluorescence activated cell sorting"), PCR-activated cell sorting (or PACS), microsphere affinity proteomics (MAP), mass spectrometry, chromatography, CYTOF, spectral cytometry, mass cytometry, imaging cytometry, gene expression microarrays, e.g., DNA (DNA-seq or single cell (sc)DNA-seq) or RNA (RNA-seq or single cell (sc)RNA-seq), in situ hybridization, and / or microscopy.

[0072] In one embodiment, the step of providing at least the hierarchical structure as output further comprises a display step. Indeed, as will be seen in more detail below, the hierarchical structure obtained by the present method provides a particularly advantageous flexibility for visualizing and interpreting data from cytometric measurements.

[0073] In one embodiment, the displaying of the hierarchical structure is configured to allow an operator to select one or more values ​​of the predefined persistence threshold associated with at least one or more levels of the hierarchical structure, thereby allowing the operator to select the level of granularity at which the cytometric data is displayed. The displaying of the hierarchical structure may further be configured to allow the operator to directly select which classes to display, regardless of membership level.

[0074] In one embodiment, the displaying step comprises: at least one selection graph, the selection graph being a graphical representation of classes and respective relationships of the classes in a hierarchical structure, the selection graph being adapted to select at least one class by specifying a corresponding class on the graphical representation; at least one two-dimensional or three-dimensional presentation graph, each axis of which is related to a respective cytometry parameter, the at least one presentation graph being configured to display points of the point cloud belonging to at least one selected class using a selection graph; The method includes the step of displaying the

[0075] Advantageously, this embodiment makes it particularly easy to navigate the hierarchical structure across all the classes located at different levels. It makes it possible to identify related classes that were very difficult to identify with previous techniques. Moreover, the selection graph allows the selection of classes at different levels and the presentation graph allows the visualization of the points simultaneously. In other words, the method of the invention makes it possible to visualize the "granularity" of classes at one level, but also allows "zooming in" of one or more selected classes or "zooming out" by selecting classes at a higher hierarchical level.

[0076] Moreover, in one embodiment, the selection graph further comprises a visualization means adapted to highlight at least one selected class on the graphical representation.

[0077] Moreover, such simultaneous display of both hierarchical structures (i.e., selection graphs) and presentation graphs advantageously presents the data set to the operator in a format that includes global information on the membership of one or more groups of biological objects to one or more classes, with close hierarchical links, and specific information on the distribution of the objects in the groups relative to two or three cytometric parameters. This form of display not only has the simple advantage of conveying information in a way that the observer finds intuitively particularly appealing, clear, or logical. The simultaneous display of global (i.e., selection graphs) and specific (i.e., presentation graphs) information provides the operator with contextual information that enhances the accuracy of the data analysis. This combination of information at different scales gives the operator a wealth of information about the cytometric data, including community relationships between populations of biological objects. This display not only allows easier and faster interpretation of the entire set of biological objects under analysis, but also, thanks to the additional information on the display of the remaining links between the subtypes of biological objects (i.e., the selection graph) and their cytological parameter values ​​(i.e., the presentation graph), the operator can directly draw conclusions regarding, for example, a clinical diagnosis, a prediction of the probability that the patient will develop a disease, an evaluation of the patient's response to a treatment, or monitoring the patient during treatment and the indication of a potential treatment. In general, and especially in the case of cytometric data relating to a patient, the cognitive content of the information presented to the operator relates to the patient's health status. The operator can therefore use this cognitive content to evaluate the patient's health status, make a diagnosis, suggest a treatment, or take any other action to improve the patient's health. In another example, where the cytometric data relates to a biofunctionalized material, the operator can use the cognitive content of the display of the present invention during the process of designing or testing the effectiveness of a candidate drug in an in vitro model.

[0078] Advantageously, the graphical representation can be dynamically altered so that an operator navigates between different views of the hierarchical structure by selecting a threshold associated with at least one or more levels of the hierarchical structure and by the operator selecting the classes to display. For example, at one point in time, the operator can visualize a first subpopulation of biological objects. Then, at a later point in time, the operator can visualize a second subpopulation of biological objects that is different from the first subpopulation. The operator can then return to visualizing the first subpopulation.

[0079] In one embodiment, the selection graph is presented in the form of a dendrogram, with each node representing a class, the nodes are arranged in a hierarchy, with each node representing a level of membership for the class represented by the node in the hierarchy, and the branches representing hierarchical links between classes at different levels.

[0080] In an alternative embodiment, the selection graph is presented in the form of a sunburst chart including concentric rings each representing a level, with the ring having a larger diameter the lower the hierarchical level it represents, and the classes of a ring level are fractions of the corresponding ring. The particular shape of the sunburst chart selection graph is particularly advantageous for representing and navigating among the different classes and levels present in the hierarchical structure.

[0081] In one embodiment, the length of a fraction is proportional to the number of points in the corresponding class.

[0082] In one embodiment, fractions representing classes that have direct hierarchical links (ie, classes that belong to the same expanded basin) are placed in the same angular sector of the sunburst chart.

[0083] In one embodiment, the graphical representation is a scatter plot graph with visualization means configured to distinguish points belonging to two or more selected classes.

[0084] In one embodiment, the at least one selection graph is a dynamic graphical representation of the classes and their arrangement in a hierarchical structure, the dynamic graphical representation configured to display sub-hierarchical classes when a grouping class is selected.

[0085] The method according to the invention further preferably comprises one or more of the following embodiments: p and / or q are greater than 20, preferably greater than 30, preferably greater than 40 and / or less than 80, preferably less than 70, preferably less than 60, preferably equal to 50; in step a), the distance represents a mathematical distance, in particular an average distance, between the first point and the neighboring points, the mathematical distance being preferably chosen from the Euclidean distance, the Minkowski distance, the Manhattan distance or the Mahalanobis distance, At the end of step a), the density undergoes a given transformation such that it is an increasing function, such as a logarithmic transformation.

[0086] In one embodiment, the method further includes receiving a selection of one or more classes from an operator using a selection graph.

[0087] In another embodiment, the method further comprises the step of providing a medical or research report as output, and / or contacting persons associated with the cytometry event, in particular the person from whom the biological fluid on which the cytometry measurement was performed, for example by sending a letter or email, and / or manufacturing a device based on the analysis.

[0088] The invention further relates to a data processing device comprising means for implementing a method according to any one of the above mentioned embodiments.

[0089] More specifically, the data processing device is configured for analysis of a dataset related to a plurality of biological objects selected from cells, cell-derived vesicles, cell-free microorganisms, and / or biofunctionalized materials, the dataset comprising N cytometry events each related to a biological object, each cytometry event being defined by at least two cytometry parameters measured for a corresponding biological object, such that the dataset is represented by a point cloud (X) of N points in a D-dimensional space, and the device: at least one input configured to receive a dataset related to a plurality of biological objects; at least one processor, -For each point (x i ), the point (x i ) and the points to be considered (x i ) is a set of p points in the point cloud (X) that are closest to the neighboring point (V p (x i The density (w(x i )) is determined, -Point cloud (X) to modal segment (M j ) and each modal segment (M j ) is the modal point (m j ), and the points of the point cloud are modal points (m j ) basin, and the points of the basin are compared by their density (w(x i )) and -Each modal segment (M j ), determine persistence (β), which is Modal segment (M j ) adjacent to the modal segment (M t ) is the modal point (m j ), and two or more adjacent modal segments share the same density saddle point greater than 0, then it is an infinite value, otherwise Modal segment (Mj ) and at least one adjacent modal segment (M t ) represents the depth of the density saddle point between -Modal segment (M d ) and a) The modal segment considered (M j ) and its adjacent modal segments (M t ) to identify the hierarchical link between the modal segment (M j ) and its adjacent modal segments (M t ) which corresponds to the highest basin, b) The modal segment considered (M j ) and the considered modal segment (M j ) and modal segments whose persistence is equal to or less than a certain persistence threshold are merged, and the points that have a higher basin and show maximum density among the points of the merged modal segments are designated as modal points (m j ) a parent modal segment having c) starting a new iteration from the parent modal segment and repeating actions a) and b) iteratively; Determine the hierarchical structure defined based on persistence, and k ) contains multiple levels, and each level (L k ) in multiple classes (C kj ), where each class has one or more modal segments (M j ) and represents a group of multiple biological objects. at least one processor configured to at least one output unit configured to provide at least a hierarchical structure representing different classes of biological objects and their interrelationships; Equipped with.

[0090] The invention further relates to a computer program product comprising instructions which, when executed by a computer, cause the computer to implement the method according to any one of the above embodiments.

[0091] The invention further relates to a computer-readable recording medium comprising instructions which, when executed by a computer, cause the computer to implement a method according to any one of the above-mentioned embodiments.

[0092] definition For purposes of the present invention, the following terms are defined as follows.

[0093] The term "processor" should not be construed as being limited to hardware capable of executing software, but generally refers to a processing device that may include, for example, a computer, a microprocessor, an integrated circuit, or a programmable logic device (PLD). A processor may also include one or more graphics processors (GPUs), whether utilized for graphics and image processing or other functions. Furthermore, instructions and / or data for performing the relevant functions and / or resulting functions may be stored in any processor-readable medium, such as, for example, an integrated circuit, a hard drive, an optical disk such as a CD (compact disk), a DVD (digital versatile disk), a RAM (random access memory), or a ROM (read only memory). Instructions may be stored in hardware, software, firmware, or any combination thereof.

[0094] "Cytometry" generally refers to the detection and / or measurement of cellular characteristics, basic structures, functions and biological units of living organisms. In addition, cytometry also refers to the detection and / or measurement of the characteristics of nano- or micrometer-sized particles such as vesicles of cell origin (such as extracellular or intracellular vesicles), or acellular microorganisms (e.g., viruses), or biofunctionalized materials (e.g., microspheres coated with biological molecules such as proteins or nucleic acids, or microspheres coated with bioactive molecules such as immunomodulatory molecules or drugs).

[0095] A "cell" may be of unicellular or multicellular organism, of prokaryotic or eukaryotic origin, of animal, plant, fungal, protist, bacterial or archaeal origin. Cells may be living, dead or fixed. Cells may be derived, for example, from solid or liquid tissues (e.g., bone marrow, blood or lymph, etc.) or from bodily fluids (e.g., cerebrospinal fluid, urine, bronchoalveolar fluid, etc.).

[0096] During the process of acquiring cytometric data, the cells, vesicles, or particles may be in suspension (aqueous or non-aqueous, biological or synthetic) or immobilized on a solid support (e.g., a vial, a cell culture dish with one or more wells, a plate or microplate, a glass slide, a chip, etc.). The cells, vesicles, or particles may have been subjected to prior treatment with one or more biological or chemical agents.

[0097] The cells, vesicles, or particles being analyzed may be present in one or more populations.

[0098] A "population" is a set of cells, vesicles, or particles that exhibit the same or similar properties. A population can be subdivided into different subpopulations that exhibit different secondary properties relative to each other.

[0099] "Cell-origin vesicles" are vesicles composed of membranes of cell origin. Cell-origin vesicles can be intracellular or extracellular vesicles. Extracellular vesicles are typically secreted or excreted by cells. "Cell-origin vesicles" include, for example, exosomes, ectosomes, microvesicles, microparticles, prostasomes, oncosomes, matrix / calcified vesicles, apoptotic bodies, and other subsets of cell vesicles.

[0100] "Acellular microorganisms" refer to microscopic organisms whose structure is not cellular. "Acellular microorganisms" include, for example, viruses, viroids, and prions.

[0101] "Biofunctionalized material" or "functionalized biomaterial" or "biofunctionalized device" or "biofunctionalized system" refers to any type of object that includes a material, either synthetic or of biological origin, coated on its surface, coupled or attached to one or more molecules or functional groups of biological origin.

[0102] Non-limiting examples include synthetic or biologically derived materials such as nanoparticles (e.g., nanobeads, nanospheres, or nanocapsules), microparticles (e.g., microbeads, microspheres, or microcapsules), unilamellar vesicles, multilamellar vesicles, lipoplexes, polyplexes, lipopolyplexes, liposomes, niosomes, cochleates, virosomes, immunostimulating complexes (ISCOMs), and other lipid vesicles.

[0103] Nanoparticles and microparticles, such as beads, spheres, or capsules, can be organic, inorganic, magnetic, or radioactive. For example, the nanoparticles or microparticles can be made of metal, silica, alumina, titanium, glass, ceramic, polystyrene, polymethylmethacrylate, melamine, polylactide, latex, dextran, oxide, graphene, magnetic materials, radioactive materials, or combinations thereof.

[0104] As non-limiting examples of what may be biofunctionalized, the material may be conjugated or coated with one or more peptides, proteins, antibodies, antibody fragments, receptors, cytokines, chemokines, toxins, oligonucleotides, colored or fluorescent molecules, amines, carboxyl or hydroxyl groups, biologically active molecules (e.g., immunomodulatory molecules, small chemical molecules, peptidomimetics, drugs), biotin, avidin or streptavidin molecules, or combinations thereof.

[0105] In the context of the present invention, the term "cytometry" when modifying data, parameters, measurements, or events may refer to particles such as cells, cell-derived vesicles, or acellular microorganisms or biofunctionalized materials. [Brief description of the drawings]

[0106] [Figure 1] FIG. 1 is a block diagram showing an example of an analysis method according to the present invention.

[0107] [Diagram 2] FIG. 2 shows an example of a point density representation obtained according to the method of FIG. 1. [Diagram 3] FIG. 2 shows an example of a point density representation obtained according to the method of FIG. 1.

[0108] [Figure 4] FIG. 4 is a diagram illustrating an example of a selection graph according to the first embodiment.

[0109] [Diagram 5] FIG. 13 is a diagram showing a modified example of a selection graph according to the second embodiment. [Figure 6] FIG. 13 is a diagram showing a modified example of a selection graph according to the second embodiment.

[0110] [Figure 7] FIG. 1 illustrates a representation of a graph according to one embodiment.

[0111] [Figure 8] 1 depicts an exemplary display including a selection graph and a suggested graph in accordance with the present invention.

[0112] [Figure 9] FIG. 13 depicts an example of using a selection graph to change a display on a presentation graph in accordance with the present invention.

[0113] [Figure 10] FIG. 2 is a functional diagram that generally represents a particular embodiment of a device for analyzing data sets relating to multiple biological subjects, configured to perform successive steps of the method of FIG. 1.

[0114] [Figure 11] FIG. 11 is a schematic diagram of a device incorporating the functionality of the analytical device of FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0115] This specification illustrates the principles of the present disclosure, and it will thus be appreciated that those skilled in the art will be able to devise various arrangements which, although not explicitly described or shown in this specification, embody the principles of the present disclosure and are included within its scope.

[0116] All examples and conditional language cited herein are for educational purposes to aid the reader in understanding the principles of the present disclosure and concepts provided by the inventors to advance the state of the art, and should not be construed as being limited to the specifically cited examples and conditions.

[0117] Moreover, all descriptions of principles, aspects, and embodiments of the present disclosure, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, these equivalents are anticipated to include both currently known equivalents as well as equivalents developed in the future, meaning all elements developed that perform the same function, regardless of their structure.

[0118] Thus, for example, those skilled in the art will appreciate that the schematic diagrams presented herein may represent conceptual views of illustrative circuitry implementing the principles of the present disclosure.Similarly, it will be appreciated that all flow charts, flow diagrams, and the like, whether explicitly represented or not, are inherently represented on a computer-readable medium and thus represent various processes that may be executed by a computer or processor.

[0119] The functions of the various elements shown in the figures may be provided through the use of dedicated hardware as well as hardware capable of executing software in combination with appropriate software. When provided by a processor, the functions may be provided by a single dedicated processor, by a single shared processor, or by a plurality of individual processors, some of which may be shared.

[0120] It will be understood that the elements shown in the figures may be implemented in various forms of hardware, software, or a combination thereof. Preferably, these elements are implemented in a combination of hardware and software on one or more appropriately programmed general-purpose devices, which may include a processor, memory, and input / output interfaces.

[0121] FIG. 1 shows an example of implementing the computational analysis method according to the invention.

[0122] The method is to find N points x in a D-dimensional space. i The method is configured to perform an analysis of a data set relating to a plurality of biological objects 21 (or cytometry data 21) that can be represented as a point cloud X of i represents a cytometry event and each dimension represents a cytometry parameter.

[0123] The method is configured to output a hierarchical structure 31 representing the different classes of biological objects represented in the input dataset and their interrelationships.

[0124] The hierarchical structure 31 is a multi-level structure L k and for each level L k defines the respective presentation of the point cloud and / or the data related to the points of the point cloud, i.e. the location of the points in space and the density of the points. The higher the level in the structure, the fewer the number of classes at that level. The structure is k Class C in kj The point is, - the next higher hierarchical level L k+1 The only Class C (k+1)j Also included in the level L k (if such a level exists above it), and therefore the class C considered kj The next level up is Class C. (k+1)j the daughter of, and the mother of, a class considered in itself; and -The next level down(k-1) One or more of Class C (k-1)j (The level L where such a level is considered.) k If present under In that sense, it is hierarchical.

[0125] The process may include a step 41 of receiving cytometry data sets 21 relating to a plurality of biological objects.

[0126] Biological subjects are - cells of animal, plant, fungal, protist, bacterial or archaeal origin, - a cell-origin vesicle selected from exosomes, ectosomes, microvesicles, microparticles, prostasomes, oncosomes, matrix / calcified vesicles, or apoptotic bodies; - acellular microorganisms selected from viruses, viroids and prions, and / or a biofunctionalized material comprising a synthetic or biological material selected from nanoparticles (e.g. nanobeads, nanospheres, or nanocapsules), microparticles (e.g. microbeads, microspheres, or microcapsules), lipid vesicles (e.g. unilamellar vesicles, multilamellar vesicles, lipoplexes, polyplexes, lipopolyplexes, liposomes, niosomes, cochleates, virosomes, immunostimulating complexes (ISCOM®)), said synthetic or biological material being coupled to or coated with one or more peptides, proteins, antibody fragments, receptors, cytokines, chemokines, toxins, oligonucleotides, colored or fluorescent molecules, amines, carboxyl, or hydroxyl groups, biologically active molecules (e.g. immunomodulatory molecules, small chemical molecules, peptidomimetics, drugs), biotin, avidin, or streptavidin molecules, or combinations thereof; may include.

[0127] The cytometry data 21 can be obtained by any method that can analyze, determine, or measure properties of a cell, vesicle, or particle. The analysis can be focused on a sample that includes a plurality of cells, vesicles, and / or particles. For example, the cytometry data 21 can be obtained from a biological sample from an individual, such as one or more organs, tissues, cells, or cell fragments from the individual.

[0128] In some cases, the process of acquiring cytometry data 21 may be preceded by a sample preparation process, including, for example, a step of isolating cells, vesicles and / or particles, or prior isolation of components of said cells.

[0129] The sample preparation process involving the biological objects associated with the cytometry data 21 may include, by way of non-limiting example, one or more steps of magnetic activated cell sorting (MACS), laser capture microdissection (LCM), manual cell harvesting or micromanipulation, microfluidic techniques, limiting dilution, optical tweezers separation, electrophoresis, bead separation, immunoprecipitation, immunopanning, labeling, immunostaining, immunofluorescence, multiplexing, and / or the use of biochips.

[0130] The cytometry data 21 may be omics or metaomics technology data such as genomics, epigenomics, transcriptomics, proteomics, metabolomics, lipidomics, etc. Thus, the cytometry data 21 may relate to, for example, the detection and / or quantification of DNA (e.g., genes), RNA, proteins (e.g., cytokines, receptors...), sugars, amino acids, and / or fatty acids present in or on a cell, vesicle, or particle, or any other molecule contained in the cell, vesicle, or particle. The cytometry data 21 may also relate to the detection of the presence or absence of a particular molecule or aggregate in or on a cell, vesicle, or particle, or the detection of an interaction between molecules present in or on a cell, vesicle, or particle, or the detection of an interaction between objects such as cells and / or vesicles, microorganisms, or biofunctionalized materials.

[0131] As a non-limiting and purely illustrative example, where the cytometric data 21 is proteomic in nature, the measured or analyzed cytometric parameters may be the expression levels of one or more intracellular or extracellular proteins, or the expression levels of one or more receptors or markers on the surface of a cell, vesicle or particle.

[0132] Other examples of cytometric parameters include the size of cells, vesicles, or particles, their density, granularity, morphology / shape, refractive index, membrane composition, molecular content (such as the presence or absence of intracellular or surface molecules) or their content (e.g. the intracellular content or content of ions, DNA, RNA, ...), and the redox state of a cell, the state of a cell in the cell cycle, the apoptotic state of a cell, or the phosphorylation level of a cell, ...

[0133] The cytometry data 21 can be obtained by single cell analysis techniques.

[0134] As non-limiting examples, the cytometry data 21 can be obtained by flow cytometry (or FACS for "fluorescence activated cell sorting"), PCR-activated cell sorting (PACS), microsphere affinity proteomics (MAP), mass spectrometry, chromatography, CYTOF, spectral cytometry, mass cytometry, imaging cytometry, gene expression profiling on chips (microarrays), sequencing, such as DNA sequencing (DNA-seq or single cell (sc) DNA-seq) or RNA sequencing (RNA-seq or single cell (sc) RNA-seq), in situ hybridization, and / or microscopy. The term "microscopy" includes atomic force microscopy (AFM), electrochemical detection (EC), scanning electron microscopy (SEM), transmission electron microscopy (TEM), surface plasmon resonance imaging (SPRi), Raman microspectroscopy...

[0135] Cytometric data 21 can also be obtained by combining several of these techniques. As a purely illustrative example, multiplex profiling of RNA and protein expression at the single cell level can be obtained simultaneously by combining the PLAYR technique (Proximity Ligation Assay for RNA), which measures the expression levels of multiple RNAs by flow or mass cytometry, with techniques for detecting proteins that are surface or internally labeled with antibodies.

[0136] Another purely illustrative example of a technique for acquiring cytometry data 21 is the profiling of individual cells isolated in droplets encapsulated in thin semi-permeable membranes (microcapsules) by high-throughput RNA cytometry, e.g. by multiplexed RT-PCR.

[0137] The cytometry data 21 includes a plurality of vectors, each of which includes D cytometry parameters measured for a cytometry event associated with a biological object. Each vector can be interpreted as the coordinates of a point in a D-dimensional space. Thus, the cytometry data 21 can be represented by a point cloud X of N points in the D-dimensional space.

[0138] The cytometric parameters of interest in the cytometric measurements preferably include one or a combination of the following cytometric parameters: the size of particles in the biological fluid in which the cytometric measurements are performed, the structure of said particles, the function of said particles, and any measured parameter that directly or indirectly reflects the properties and / or functions (qualitative and quantitative) of markers of interest, which may include proteins, nucleic acids, and molecules, alone or in complexes, and their interactions.

[0139] The number of dimensions D, i.e., the number of cytometry parameters associated with each cytometry event, may be greater than 1, preferably greater than 10, preferably greater than 20, preferably greater than 30, even more preferably greater than 40, and / or less than 200.

[0140] The number of points N, i.e. the number of biological objects contained in the biological fluid subjected to cytometric measurements, is preferably more than 10,000 and / or less than 10 9 .

[0141] The method may include a pre-processing step (step 42) of the cytometry data 21, particularly to facilitate subsequent analysis. Pre-processing may include compensating the cytometry data 21, converting the cytometry data 21 to a scale suitable for visualization and / or analysis, and / or normalizing the cytometry data 21.

[0142] For example, compensation can be achieved by multiplying each vector representing a point in the point cloud by the inverse of a compensation matrix of size D×D. The compensation matrix can be obtained by any method known to those skilled in the art. Thus, a set of vectors representing the points of the compensated point cloud can be obtained.

[0143] The data transformation may correspond to application of a mathematical function to the initial vectors or compensated vectors associated with the points of the point cloud, said mathematical function may include linear, logarithmic, bi-exponential, or hyperbolic arcsine "arcsinh" functions. In the illustrated embodiment, an "arcsinh" function is applied to the compensated point cloud.

[0144] Finally, data normalization can be performed in a known manner, in particular of the "min-max" type.

[0145] At the end of step 42, a compensated, transformed and / or normalized point cloud X is obtained (i.e., the vectors associated with the points of the point cloud are compensated, transformed and / or normalized).

[0146] The method of the invention aims to obtain a segmentation of a point cloud based on topological persistence, which studies the evolution of the topology of a multi-level set of the density function of the point cloud.

[0147] The method includes a step 43 of determining a density of extrema associated with each point in the point cloud. In this step, for each point x i For the considered point x i and a set V consisting of p points in the point cloud that are closest to the considered point. p (x i ) is inversely proportional to the sum of the distances between i ) is determined.

[0148] In the present disclosure, a point is considered to be near or adjacent to a point under consideration if it is one of the p points in the point cloud that is closest to the point under consideration. The proximity between two points is defined by the distance between these two points in the D-dimensional space of the point cloud. A point may or may not be a neighbor of the point under consideration, depending on the choice of the parameter p and the definition of the distance chosen. The proximity between two points can be calculated using Euclidean distance, Minkowski distance, or Manhattan distance. It is not beyond the scope of the present invention if other distances are used.

[0149] For density calculations, the denominator is preferably the number of points x considered i and a set V consisting of the p closest points of the point cloud to the point under consideration. p (x i ) The distance is preferably the average Euclidean distance between the considered point and its p neighbors. However, other types of distances can be used, such as the Minkowski distance, the Manhattan distance, or similar distances.

[0150] In the illustrated example, p is set to 50, but other values ​​of p can be chosen depending, inter alia, on the size of the point cloud, the nature of the cytometry data 21, or the intended application of the method.

[0151] In the preferred illustrated example, a point x in a point cloud X i For density w(x i ) is calculated up to a certain factor according to the following formula:

number

[0152] During the ceremony,

number

[0153] At the end of step 43, the density of the points in the point cloud w(x i ) is preferably logarithmically transformed, which advantageously attenuates relative density differences.

[0154] An example of a representation of the point density obtained in step 43 is shown in Figure 2. In the example shown, the space is two-dimensional, the points of the point cloud lie in a plane, and the values ​​of the density of said points are represented on a third dimension orthogonal to the plane.

[0155] The method then calculates a point x i To define groupings (more commonly called "clusters") of points X, we divide the points X into modal segments M j The step of segmenting the modal segment M jdefines a spatial region that contains a group of points from the point cloud. This set of points includes the modal point that has a density greater than the density of other points in the modal segment, and other points in space that are adjacent to the modal point and have a density less than that of the modal point. Thus, the modal point m j is related to the density called the modal density, which is the local maximum of the density, which is the modal segment M j This means that the density is higher than that associated with any other point in the

[0156] Modal segment M j can be naturally identified in the basin of the peak of the approximate density function. Intuitively, if we consider the density function as a landscape, the modal segment M j is the distribution of all points x that flow towards the same maximum (or peak) along the flow defined by the vector field of the gradient of the density function. i By identifying the density basin, we divide the point cloud into one or more modal segments M j Each modal segment M j is the density w(x i ) associated with the maximum of the modal point m j , all points in the point cloud belong to that basin. Thus, points in the basin are scaled by their density w(x i ) are identified recursively.

[0157] A modal segment adjacent to (or connected to) another modal segment is a modal segment that has the same density saddle point, said density saddle point being greater than 0. Conversely, two segments that have a null density value at their common boundary are not considered to be adjacent modal segments in the context of the disclosed method, since no points are found in this region of space. Locally, on both sides of the boundary that separates two adjacent modal segments, the points of each of these modal segments are associated with an increasing density as they move away from said boundary. In other words, the boundary defines the base of the density saddle point.

[0158] In other words, two segments are adjacent if there exists a point in the point cloud that contains at least one point in its neighborhood that belongs to both modal segments. A density saddle point (or boundary) is the point with the lowest density among the points that have neighbors in two adjacent modal segments.

[0159] In one embodiment, the following operations may be performed: for each point x in the point cloud i , proceeding in order of density, i.e., from the point with the highest density to the point with the lowest density, the density of the considered point is calculated by multiplying the density of its q nearest neighbors V q (x i ) is higher than each of the densities of the considered point x i If not, a modal segment is formed that includes the considered point x i are the q nearest neighbors V q (x i ) is included in the modal segment that contains the point with the highest density among all points in the point cloud. This operation is based on knowledge of the distances of pairs of points in the point cloud and the density values ​​corresponding to each point in the point cloud. Advantageously, the disclosed method requires knowledge only of the (approximate) distances between pairs of cytometric data points 21, and an approximate estimate of the density at these points. It is therefore applicable to virtually any metric space chosen to represent the point cloud. Furthermore, the complexity of the method remains reasonable, and although the size of the input distance matrix may be quadratic in the number N of points in the point cloud, the method uses only a linear amount of main memory in N.

[0160] The parameters p and / or q can be greater than 20, preferably greater than 30, preferably greater than 40, and / or less than 80, preferably less than 70, preferably less than 60, preferably equal to 50. The parameters p and q can be identical or different. In the illustrated example, the number q is equal to 50.

[0161] In the example shown in Figure 2, six modal segments M1-6 have been identified. As shown, each modal segment can resemble a mountain whose peak is a local maximum of density.

[0162] The method then performs the steps of: j The method includes a step 45 of estimating the persistence of

[0163] Modal segment M j The persistence β of is defined as: - its modal point m j At least one adjacent modal segment M has points associated with a density greater than the density of t has an infinite value if does not exist, or - otherwise, modal segment M j and the at least one adjacent modal segment M t represents the depth of the density saddle point between

[0164] A modal segment M having at least one adjacent modal segment j The persistence β of the modal point m j of the modal segment M j ) and modal segment M j and the at least one adjacent modal segment M t The density saddle point between the highest density value and the density value associated with the points of the point cloud that lie within the ...

[0165] For a simplified representation in which the points of the point cloud are related to a single cytometric parameter and the density w(x) corresponds to the values ​​on the vertical axis, an example of the persistence determination is shown in Fig. 3. In the example shown, the persistence β1 of the modal segment M1, i.e. the modal segment presenting the maximal density with the maximum value, is infinite.

[0166] The modal segment M3 is adjacent only to the modal segment M2. The persistence β3 of the modal segment M3 therefore corresponds to the difference between the value of its modal density, i.e. the maximum value of the density w(m3) of the basin of the modal segment M3, and the value of the minimum density between the modal segments M3 and M2. This persistence β3 corresponds to the distance along the vertical axis separating the two lines Δ′1 and Δ′2 in FIG. 3 and is equal to approximately 0.032 in the example considered.

[0167] In the example shown in Fig. 3, modal segment M2 has two adjacent modal segments M1 and M3 with a density value higher than that of modal segment M2. The persistence β2 of modal segment M2 corresponds to the difference between its modal density w(m2) and the highest density value among the two density minima between the adjacent modal segments, i.e., the difference between the peak of modal segment M2 and the highest saddle point between modal segments M2 and M1. In the illustrated case, persistence β2 corresponds to the distance along the vertical axis separating the two lines Δ1 and Δ2 in Fig. 3 and is equal to about 0.013 in the example considered.

[0168] The next step in the process, step 46, is to determine the modal segments M based on the persistence values ​​calculated in the previous step in order to build a hierarchical structure. j , and is configured to group points contained in a modal segment.

[0169] The hierarchical structure 31 defined by this step 46 has l levels L k where k ranges from 2 to 1. The number of levels 1 is preferably greater than 2, preferably greater than 3, preferably greater than 4, and / or less than 100, preferably less than 50, more preferably less than 30.

[0170] Each level L k H class C kh Define a segmentation of the point cloud X into each class C khis one or several modal segments M defined in step 44. j Group the points from all modal segments M j is associated with a class at some level, and no modal segment belongs to more than one class (i.e., no modal segment is shared between multiple classes).

[0171] The lower the level of the hierarchical structure 31, the greater the number of classes at that level. The base level L1 results from grouping all adjacent modal segments that can be grouped into a single common basin (i.e., a coarser segmentation). The final level L l represents the finest segmentation.

[0172] Level L k Class C kh To define the modal segment M j is the persistence β of each of the modal segments, whether global (i.e., has the same value at all levels) or at level L k A predefined persistence threshold that may be defined differently for each

number

[0173] Step 46 selects one or more adjacent modal segments M t Starting modal segment M j In an alternative embodiment, the starting modal segment M j is the modal point m j The starting modal segment M is selected based on the density value of j is the modal point m with the highest density value. j According to the method of the present invention, the modal segment M j and two or more adjacent modal segments (Mt1 , M t2 etc.), if the maximum depth of the density saddle point between j is a set of multiple adjacent modal segments (M t1 , M t2 etc.).

[0174] The starting modal segment M j Step 42 is to calculate the modal segment M j and its adjacent modal segment M t Which of the modal points m has the highest basin (i.e., the modal point m with the highest density value)? j The modal segment having a lower basin is considered within the modal segment having a higher basin, and the considered modal segment (M j , M t ) The term "height" of a basin refers to the density value of the modal points of the modal segments (i.e. density maxima) associated with said basin.

[0175] Next, the modal segment M j The persistence of the corresponding level L k A given persistence threshold for

number

number

number

number

[0176] In one example, the persistence threshold 22 is set to each modal segment M obtained in step 44. j , so all its points are at level L in the hierarchy. l Class C kl Level L l refers to the lowest hierarchical level in the hierarchical structure (i.e., the level with the highest granularity). The retrieved parent modal segment is the next higher hierarchical level L l Belonging to class C k(l-1) Therefore, Class C k(l-1) is the fused start segment M j and its adjacent modal segment M t Class C corresponding to kl In particular, Class C k(l-1) is the lower hierarchical level L l It can be considered the parent class of these two or more classes.

[0177] Finally, these determination and fusion steps are repeated iteratively using the newly defined parent modal segment as the starting modal segment. The same kind of hierarchical relationship as described above is defined for the parent modal segments created in subsequent iterations. As a result, fusion is performed in order of decreasing persistence.

[0178] By repeating these steps, it is possible to determine a hierarchical structure 31 based on persistence, which provides a representation of groups of biological objects contained within the cytometry data 21.

[0179] Level L k The number of classes in the threshold persistence associated with that level

number

number

number

number

[0180] In one embodiment, the step of outputting at least the hierarchical structure 31 further comprises at least a partial representation thereof via an operator accessible interface. Preferably, the hierarchical structure 31 is displayed on a screen. The hierarchical structure 31 may be stored in a computer memory before being presented.

[0181] The purpose of presenting the hierarchical structure 31 is to allow the operator to select classes of interest and to group only those points that the operator wishes to analyze.

[0182] The display step includes simultaneously displaying at least one selected graph 20 and at least one presented graph 40 .

[0183] At least one selection graph 20 is a graphical representation of the classes and their respective relationships in a hierarchical structure 31, and an operator can select at least one class C by specifying a corresponding class on the graphical representation. kh The present invention is adapted to enable a user to select

[0184] In the same display, the method further represents at least one or three-dimensional presentation graphs 40, each axis being related to a respective cytometric parameter. In the field of cytometric data analysis 21, it is particularly advantageous for the operator to simultaneously display multiple presentation graphs with different pairs of axes. The operator can select the axes of the presentation graphs 40 via a graphical interface. The presentation graphs 40 are displayed based on the selected class C using the selection graph 20. kh An example of a simultaneous display is shown in FIG. 4, where the left column contains the suggested graphs 40 and the right column contains the corresponding selected graphs 20.

[0185] Preferably, the selection graph 20 visually represents the classes by grouping them according to levels. For example, a class may be visually further from a reference point 22 on the selection graph if it belongs to a level further away from the base level. The reference point may be, for example, the apex of a pyramid, where each level corresponds to a layer of the pyramid as shown in Figure 4, or the center of a circle, where each level corresponds to a ring centered on the nucleus, as depicted in Figure 5.

[0186] In practice, an operator can use a dynamic cursor (ie, a mouse) via a graphical interface to click and thus select one of the classes represented on the graphical representation of selection graph 20 .

[0187] In one embodiment, the graphical representation of the selection graph 20 is presented in the form of a dendrogram, which is a diagram representing a tree commonly used to show the arrangement of groups produced by hierarchical clustering. In this dendrogram, each node 12 is a member of class C kh Branch 14 represents Class C kh Represents a hierarchical link between the same level L k The nodes 12 belonging to can be arranged hierarchically.

[0188] In the example shown in FIG. 4, the selection graph 20 takes the form of a dendrogram, with levels L 1~5 represents different branch levels. Each node 12 is of class C kh Each branch 14 represents a mother class C kh and daughter class C (k+1)h Represents a hierarchical link between.

[0189] In one embodiment, the selection graph 20 is presented in the form of a sunburst chart. The chart is depicted as a disk divided into a number of sectors or "quadrants", each sector representing, and only representing, all points (and / or associated data) of a respective class from the highest base level. Preferably, the angle of a sector is proportional to the number of points in the class from the base level it represents. The sunburst chart includes concentric rings, each ring representing a level of the hierarchical structure 31. In particular, there is a core associated with level L1, having l-1 concentric rings 24, each of which is associated with another level L having a lower hierarchy than level L1. k (k ranges from 2 to l-1). In this sunburst chart, the rings have increasing diameters as they represent lower hierarchical levels.

[0190] Same level L k Class C kh can be represented as fractions of the corresponding ring, where each fraction extends within a sector that contains classes from the previous higher hierarchy before the direct or indirect hierarchical link. Two classes have a direct hierarchical link if only one branch connects them, and they have an indirect hierarchical link if the two classes are connected by at least two branches.

[0191] Preferably, the fractions representing the classes extend around the centre of the disk as circular arcs whose opening angle is proportional to the number of points in the class represented by the fraction.

[0192] FIG. 5 shows an example of a sunburst chart, where the levels are concentric rings 24, and the more concentrated the class it represents, the closer it is to the center of the concentric ring, and the lower the level L k Class C kh That level L k is the fraction of the corresponding ring 24.

[0193] In this preferred embodiment, the diameter of the ring 24 is at a level L1 further away from the base level L1. 2~5 It is so large that it represents

[0194] Preferably, for all levels, the mother class C kh Daughter Class C (k+1)h The fraction of rings from the lower hierarchical level (i.e., daughter) representing the mother class C in the ring associated with the higher hierarchical level (i.e., mother) is kh In this way, the daughter class C (k+1)h The fraction of mother class C kh overlap externally with the fraction associated with

[0195] As an illustrative and non-limiting example, if the analyzed object is a cell, the mother class may correspond to a cell population of lymphocytes. This lymphocyte population (mother class) can be subdivided into different cell subpopulations (corresponding to daughter classes of the mother class), such as, for example, T lymphocyte subpopulations and B lymphocyte subpopulations, and these two types of subpopulations exhibit different secondary cytometry parameters. In the sunburst selection graph 20, the different cell subpopulations are represented by "daughter" rings surrounding the "mother" ring.

[0196] Furthermore, these T and B lymphocyte subpopulations can be further subdivided into other subpopulations. For example, within the T lymphocyte subpopulation, analysis of cytometric parameters can distinguish between a CD4+ T lymphocyte subpopulation (T lymphocytes expressing the CD4 protein on their surface) and a CD8+ T lymphocyte subpopulation (T lymphocytes expressing the CD8 protein on their surface). The selection graph 20 can present the entire hierarchical structure 31, or preferably only a portion of the hierarchical structure 31.

[0197] The selection graph 20 may further include a visualization tool adapted to highlight the selected classes on the graphical representation. The visualization tool may be configured to, for example, change the color, shape, or size of the class representations (i.e., nodes 12 or fractions) on the selection graph 20.

[0198] In one embodiment, selection graph 20 is further configured to provide a dynamic graphical representation of classes and their placement within the hierarchical structure, which may be configured to display subhierarchical classes when a class that groups the subhierarchical classes is selected on selection graph 20.

[0199] In this embodiment, the selection graph 20 displays base-level classes by default. In this way, when an operator clicks on a representation of a "mother" class from the base level, the interface presents representations of the daughter classes of that mother class. More generally, an operator can select classes from any level L displayed on the screen. k From "Mother" Class C kh When you click on the expression, the computer will see its mother class C kh Daughter Class C (k+1)h Present the following.

[0200] Thanks to this embodiment, the operator can select any class C presented by the selection graph 20 and that does not belong to the last level of the hierarchy. kh "Zooming in" on a class means presenting its daughter classes. Thus, the operator can focus on cytometric events that are of interest to them (see Figure 6), which is particularly advantageous for analysis.

[0201] Preferably, the selection graph 20 presents only base level classes and classes that evolve from said base level classes.

[0202] Conversely, thanks to this embodiment, the operator can collapse all daughter classes of a mother class presented by the selection graph. By collapsing a collection of daughter classes, it is meant to erase the representation of said daughter classes from the screen. Thus, the operator can return to a more macroscopic view of the cytometry events of the daughter classes (see FIG. 6).

[0203] 6 shows, from left to right, an example of class expansion, and from right to left, an example of class folding according to the invention. In this example, classes C 2,j and C 4,k By selecting 3,j1 , C 3,j2 , C 5,k1 and C 5,k2 will be accessible.

[0204] Similarly, Class C 3,j2 and C 5,k1 When we expand them, their daughter class C 4,j3 , C 4,j4 , C 6,k3 and C 6,k4 is displayed.

[0205] Similarly, Class C 4,j3 and C 6,k3 By selecting 5,j5 , C 4,j6 , C 7,k5 and C 6,k6 can be accessed.

[0206] Advantageously, an operator can easily select classes of interest across different levels of the hierarchical structure.

[0207] The expandable class-by-class selection graph 20, rather than level-by-level, advantageously identifies a portion of a hierarchy, or a sub-hierarchy of interest, consisting of only classes of interest selected from different levels, as shown in FIG. 6, without having to display the same granularity for other classes that do not contain cytometry events relevant to the use case.

[0208] Thus, the graphical representation of the classes and their respective relationships in the hierarchical structure 31 of the selection graph 20 is used to select points of only those classes that the operator wants to visualize in the presentation graphs 40. Once a class is selected on the selection graph 20, all of its points are shown on each presentation graph 40.

[0209] The presentation graph 40 can be represented in the form of a scatter plot. A two-dimensional or three-dimensional scatter plot is simply a projection of selected points of the point cloud onto a plane or 3D space that is associated with two or three selected cytological parameters, respectively, with respect to the plane or 3D space.

[0210] The number of presented graphs is preferably greater than 1, 2, 3 and / or less than 200, 100, 50, 30, or 10.

[0211] All points of the same class can be represented in the same way. These presentation graphs 40 therefore preferably include visualization means configured to distinguish points belonging to two or more selected classes (FIG. 7). These visualization means allow points of the same class to be presented identically and specific to said class on different presentation graphs, allowing the operator to directly identify them. For example, points of the same class have a shape, size and / or color specific to that class.

[0212] If points in the point cloud are represented differently depending on class membership, e.g. having a particular color for the class, then the level-related representation shows the classes as groups of points of different colors. The granularity is coarser the higher the level represented in the hierarchical structure 31. Conversely, classes belonging to levels located at the bottom of the hierarchical structure 31 provide a finer-grained representation of the populations and subpopulations present in the cytometry data 21. Thus, by selecting a level, the operator can select the scale at which he wishes to visualize the cytometry events.

[0213] In the example shown in FIG. 7, points of the same class have the same color specific to that class.

[0214] FIG. 8 shows an example of a display according to the invention in which the selected graph 20 is a sunburst graph and four representation graphs 40 of the scatter type are shown on the same screen of the operator interface.

[0215] If the operator wants to "zoom in on a mother class to reveal its daughter classes", the points must be represented with an appearance specific to the daughter class to which they belong. The visualization means is therefore configured such that when the operator selects a daughter class on the selection graph 20, it changes the appearance of the points of the mother class such that each group of points belonging to one of the daughter classes has the same characteristic appearance of said class (i.e. different from the other daughter classes).

[0216] FIG. 9 shows an example of a change in the display after an operator's interaction with the selection graph 40. The left column of this figure shows a first configuration of the display in which a class associated with a fraction of the ring 41 of the selection graph has been selected by the operator, and thus the two presentation graphs 20 show points associated with the biological objects contained in the selected class. In this example, the points are displayed in the same color as the fraction of the ring 41. The right column shows how the display has been changed by the operator choosing to "zoom in" on the mother class associated with the fraction of the ring 41 in order to visualize the two daughter classes represented by the two fractions of the rings 42 and 43. This action of zooming in by selecting the two fractions of the rings 42 and 43 also makes it possible to change the display of the points associated with the two daughter classes at or near the same time as the selection of the daughter classes. The display of the points on the presentation graph 20 is changed in such a way that it is possible to distinguish between the points belonging to the two classes. In this example, the color of the points has been changed (black and grey on the presentation graph 20 in the right column). Thanks to this display, the operator simultaneously obtains information regarding the community relationships between selected subtypes of biological objects (i.e., hierarchical links between mother and daughter classes) and specific information regarding their cytometric parameters.

[0217] Advantageously, the method of the invention allows the operator to select, in at least one presentation graph, classes of interest whose points he wants or does not want to visualize by interacting with the selection graph. It is thus advantageously possible in particular not to display points considered irrelevant, for example noise consisting of cell debris, matrix or crystals. In conclusion, the presentation modes implemented by the method of the invention help the operator to carry out, by a human-machine dialogue process, the technical task of assessing the state of the system from which the cytometry data 21 originates (i.e. the health state of the patient).

[0218] The invention relates inter alia to a data processing device 1 arranged to implement the steps of the method described above (FIG. 10).

[0219] While device 1 described herein is versatile and has several functions that can be performed alternately or in any cumulative manner, other implementations within the scope of this disclosure include devices having only a portion of the functions described herein.

[0220] Advantageously, device 1 is an apparatus, or a physical part of an apparatus, designed, configured and / or adapted to perform the aforementioned functions and produce the aforementioned effects or results. In alternative implementations, device 1 is realized in the form of a set of apparatuses or physical parts of apparatuses, whether grouped on the same machine or on different, possibly remote, machines. Device 1 can, for example, be distributed on a cloud infrastructure and have functions available to users as cloud-based services or have remote functions accessible via an API.

[0221] In the following disclosure, modules should be understood as functional entities rather than physically separate components. Thus, modules can be embodied either as grouped together in the same tangible concrete component or distributed across several of these components. Similarly, each of these modules may itself be shared between at least two physical components. Furthermore, modules are implemented in the form of hardware, software, firmware, or any mixed form thereof.

[0222] The device 1 includes a module 11 for receiving cytometry data 21 and predefined hyperparameters (i.e. p, q, and / or predefined persistence thresholds) for the method 22 stored in one or more local or remote databases 10. The latter may take the form of storage resources available from any kind of suitable storage means, which may include in particular RAM or EEPROM (Electrically Erasable Programmable Read Only Memory), such as flash memory in a SSD (Solid State Disk) in some cases. Alternatively, the hyperparameters 22 are received via a communication network.

[0223] The device 1 optionally includes a module 12 for pre-processing the cytometry data 21 configured to perform one or more of the operations associated with step 42 of the method.

[0224] Furthermore, the device 1 includes a module 13 for calculating the hierarchical structure 31 configured to perform one or more of the operations related to steps 43, 44, 45 and 46 of the method.

[0225] The device 1 interacts with an operator interface 14, through which a user can input and retrieve information. The operator interface 14 includes any suitable means for inputting or retrieving data, information or instructions, including visual, tactile and / or auditory capabilities, which may include one or more of the following means known to those skilled in the art: a screen, a keyboard, a trackball, a touchpad, a touchscreen, a speaker, a voice recognition system.

[0226] The device 1 may also include a display module 15 configured to perform a display step using the operator interface 14 .

[0227] The particular device 9 visible in Figure 11 implements the above-mentioned device 1. Device 9 may correspond, for example, to a workstation, a laptop computer, a tablet, a smartphone, or a head-mounted display (HMD).

[0228] This device 9 is suitable for analyzing data sets relating to multiple biological subjects. The device 9 comprises the following elements interconnected by an address and data bus 95, which also carries a clock signal: -Microprocessor 91 (or CPU), - a graphics card 92 including multiple graphics processing devices (or GPUs) 920 and a graphics random access memory (GRAM) 921; - ROM type non-volatile memory 96, -Random Access Memory (RAM) 97, - one or more input / output (I / O) devices 94, such as a keyboard, mouse, trackball, webcam (other methods of command entry, such as voice recognition, are also possible); - Power supply 98, and -High Frequency Unit 99.

[0229] In a variant, the power supply 98 is external to the device 9 .

[0230] Device 9 also includes a direct display screen type display device 93 that is directly connected to graphics card 92 and displays composite images calculated and composited within the graphics card. The use of a dedicated bus to connect display device 93 to graphics card 92 offers the advantage of having a much higher data transmission rate and therefore reduced latency for displaying images composed by the graphics card.

[0231] In a variant, the display device is external to the device 9 and is connected to the device 9 by cable or wirelessly for transmitting the display signal. The device 9 comprises a transmission or connection interface adapted to transmit the display signal to an external display means, for example an LCD or plasma screen or a projector, for example via a graphics card 92. In this regard, the radio frequency unit 99 can be used for wireless transmission.

[0232] It should be noted that the term "register" as used below in the description of memories 97 and 921 can designate low-capacity memory areas (small amounts of binary data) and large-capacity memory areas (which can store entire programs or allow all or part of the data representing the data to be calculated or displayed) in each of the aforementioned memories. Similarly, the registers represented for RAM 97 and GRAM 921 can be arranged and organized in any manner, each of which does not necessarily correspond to adjacent memory locations and can be distributed differently (particularly covering the case where a register includes several smaller registers).

[0233] Microprocessor 91 loads and executes instructions of a program contained in RAM 97 upon power up.

[0234] As will be appreciated by those skilled in the art, the presence of a graphics card 92 is not essential and may be replaced by a complete processing and / or simpler visualization implementation by a central unit.

[0235] Additionally, device 1 may be implemented differently than standalone software, and a device or set of devices including only a portion of device 9 may be operated via API calls or a cloud interface. EXAMPLES

[0236] A number of algorithms are tested in Weber, LM and Robinson, MD (2016), Comparison of clustering methods for high-dimensional single-cell flow and mass cytometry data. Cytometry 2016, 89(12):1084-1096 (doi:10.1002 / cyto.a.23030).

[0237] However, the method according to the invention offers a better compromise between reliability and acceptability than these algorithms.

[0238] In particular, we compared the reliability obtained with the different algorithms tested by Weber and the reliability obtained with the method according to the invention on different data sets (Levine32, Levine13, Samusik01, SamusikAll).

[0239] Reliability was measured by the F-score, a classic measure of agreement between an algorithm's results and the "ground truth."

[0240] Table 1 below summarizes the F-scores obtained. [Table 1]

[0241] The table shows that the method according to the invention provides a reliability comparable to the best algorithms from the prior art.

[0242] However, the method according to the invention not only provides a reliable segmentation, but also results that are acceptable to any cytometric analysis expert, in particular for the following reasons: -These results are particularly robust, meaning that they are reproducible. - These results do not derive from the implementation of an artificial intelligence-based algorithm, the conclusions of which are difficult to explain, but are based on density, allowing an objective description of cytometric event groups within classes, - These results do not require a subjective initial parameterization of the process, -These results can be organized in a way that is well suited to cytometry, namely in a hierarchical morphology, while cells themselves are traditionally classified hierarchically.

[0243] The method according to the invention is therefore believed to offer the best compromise between reliability and acceptability for experts in cytometric analysis (resulting especially from good interpretability).

[0244] As is apparent, the present invention provides -Provide relevant segmentation, - be easily interpretable, verifiable, and acceptable to any cytometric analysis expert; -Allows highly flexible navigation between cytometry events, - Allows for rapid selection of the representation of cytometric events, A solution for analyzing cytometric data is provided.

[0245] Thus, the analysis of biological fluids offers an improved compromise between reliability, efficiency and acceptability.

Claims

1. 1. A computer-implemented method for analyzing a dataset associated with a plurality of biological objects selected from cells, cell-derived vesicles, cell-free microorganisms, and / or biofunctionalized materials, the dataset comprising N cytometry events each associated with a biological object, each cytometry event being defined by at least two cytometry parameters measured for the corresponding biological object, such that the dataset is represented by a point cloud of N points in a D-dimensional space, the method comprising: - determining, for each point of said point cloud, a density inversely proportional to the sum of the D-th powers of the distance between said point and a set of neighbors consisting of the p points of said point cloud that are closest to said considered point; - segmenting the cloud of points into modal segments, each modal segment comprising a modal point locally exhibiting a maximum density, the points of the cloud of points belonging to the basin of the modal points, the points of the basin being identified recursively based on their density relative to the densities of their q nearest neighbors; - for each modal segment, determining a persistence, said persistence being - no modal segment adjacent to said modal segment has a point with a density greater than the density of a modal point of said modal segment; if two or more modal segments are adjacent if they share the same density saddle point greater than 0, then an infinite value, otherwise represents the depth of a density saddle point between the modal segment and the at least one adjacent modal segment; Steps and - Starting with the modal segment, a) determining which of the considered modal segment and its adjacent modal segments corresponds to a highest basin in order to identify hierarchical links between the considered modal segment and its adjacent modal segments; b) fusing the considered modal segment with the modal segments whose persistence with the considered modal segment is less than or equal to a predefined persistence threshold to define a parent modal segment having a higher basin and having as modal point the point that presents a maximum density among the points of the fusing modal segment; c) iteratively repeating operations a) and b) starting from each new iteration from said parent modal segment; determining a hierarchical structure defined based on persistence, the hierarchical structure comprising a number of levels, each level defining a segmentation of the point cloud into a number of classes, each class including all points of one or more modal segments and representing a group of a number of biological objects; - outputting at least said hierarchical structure representing different classes of biological objects and the interrelationships of said classes; A method comprising:

2. the step of segmenting the point cloud into modal segments comprises, for each point, proceeding from the point having the highest density to the point having the lowest density: - comparing said density of said point with the densities of its q nearest neighbors, define a modal segment that includes the considered point if the density of the corresponding point is greater than the densities of all of its q nearest neighbors, otherwise - include the considered point in the modal segment that contains the point with the highest density among its q nearest neighbors The method of claim 1 further comprising the steps of:

3. 2. The method of claim 1 , wherein in the presence of one or more modal segments adjacent to the modal segment, the persistence is determined as the difference between the density of the modal points and the highest density of the densities of the points in the point cloud within the density saddle point between the modal segment and the at least one adjacent modal segment.

4. providing at least the hierarchical structure as an output, at least one selection graph, said selection graph being a graphical representation of said classes and their respective relationships within said hierarchical structure, said selection graph being adapted to select at least one class by specifying said corresponding class on said graphical representation, at least one two-dimensional or three-dimensional presentation graph, each axis of which is related to a respective cytometric parameter, said at least one presentation graph being configured to display said points of said point cloud belonging to said at least one selected class using said selection graph; The method of claim 1 , further comprising the step of displaying:

5. The graphical representation: a dendrogram, in which each node represents a class, said nodes are arranged in a hierarchy each representing a respective level of membership of the class represented by the node of the hierarchy, and in which branches represent the hierarchical links between classes at different levels, or - Each has a level (L k hierarchical levels, with the rings having larger diameters the lower the hierarchical level they represent, and the classes at a level are fractions of the corresponding rings at that level. The method of claim 4 , presented in the form:

6. The method of claim 4 , wherein the selection graph further comprises a visualization means configured to highlight the at least one selected class on the graphical representation.

7. The method of claim 4 , wherein the graphical representation is a scatter plot graph with visualization means configured to distinguish between the points belonging to two or more selected classes.

8. 5. The method of claim 4, wherein the at least one selection graph is a dynamic graphical representation of the classes and their arrangement in the hierarchical structure, the dynamic graphical representation configured to display a sub-hierarchical class when the class that groups the sub-hierarchical class is selected.

9. The biological subject is - animal, plant, fungal, protist, bacterial or archaeal cells, - cell-origin vesicles selected from exosomes, ectosomes, microvesicles, microparticles, prostasomes, oncosomes, matrix / calcified vesicles, or apoptotic bodies; - a cell-free microorganism selected from viruses, viroids and prions, and / or - a biofunctionalized material comprising a synthetic or biological material selected from nanoparticles (e.g. nanobeads, nanospheres, or nanocapsules), microparticles (e.g. microbeads, microspheres, or microcapsules), lipid vesicles (e.g. unilamellar vesicles, multilamellar vesicles, lipoplexes, polyplexes, lipopolyplexes, liposomes, niosomes, cochleates, virosomes, immunostimulating complexes (ISCOMs®)), said synthetic or biological material being coupled to or coated with one or more peptides, proteins, antibodies, antibody fragments, receptors, cytokines, chemokines, toxins, oligonucleotides, colored or fluorescent molecules, amines, carboxyl, or hydroxyl groups, biologically active molecules (e.g. immunomodulatory molecules, small chemical molecules, peptidomimetics, drugs), biotin, avidin, or streptavidin molecules, or combinations thereof. The method of claim 1, wherein

10. 2. The method of claim 1, wherein the cytometric parameters are selected from the size of the biological object, the density of the biological object, the granularity, the morphology, the shape, the refractive index, the membrane composition, the molecular content, the abundance of a molecule, and / or the expression level of a molecule.

11. The method of claim 10, wherein one of the measured cytometric parameters is the expression level of one or more proteins, receptors, markers, and / or the expression level of one or more nucleic acids, such as DNA or RNA.

12. 12. The method of any one of claims 1 to 11, wherein the cytometry data 21 is obtained by flow cytometry (or FACS for "fluorescence activated cell sorting"), PCR-activated cell sorting (or PACS), microsphere affinity proteomics (MAP), mass spectrometry, chromatography, CYTOF, spectral cytometry, mass cytometry, imaging cytometry, gene expression profiling on a chip (microarray), sequencing (e.g. scDNA-seq or scRNA-seq), in situ hybridization and / or microscopy.

13. A computer program product comprising instructions which, when said program is executed by a computer, cause said program to implement a method according to any one of claims 1 to 12.

14. A data processing device comprising means for implementing the method according to any one of claims 1 to 12.