Flow cytometry method for determining physically interacting cells

Abstract

The invention relates to a method for determining physically interacting cells in a cell population. The method comprises (a.) providing a flow-cytometry data set of said cell population, said data set comprising data on scattering properties and expression of cell-type specific surface markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward and / or side scatter, (b.) applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific surface markers, (c.) wherein the expression of one or more cell-type specific surface marker(s) and / or combinations of cell-type specific surface markers in a cluster is indicative of a specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster. The invention further relates to software configured for a method to determine physically interacting cells in a cell population from a flow-cytometry data set and a computer-readable storage device, comprising the software.

Description

[0001] FLOW CYTOMETRY METHOD FOR DETERMINING PHYSICALLY INTERACTING CELLS DESCRIPTION The invention is in the field of biology, flow cytometry and computer implementation of biological methods, in particular flow cytometry-based methods for the characterization and quantification of physically interacting cells in a biological sample. The invention relates to a method for determining physically interacting cells in a cell population. The method comprises (a.) providing a flow-cytometry data set of said cell population, said data set comprising data on scattering properties and expression of cell-type specific surface markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward and / or side scatter, (b.) applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific surface markers, (c.) wherein the expression of one or more cell-type specific surface marker(s) and / or combinations of cell-type specific surface markers in a cluster is indicative of a specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster. The invention further relates to the method performed on multiple flow-cytometry data sets obtained from cell populations of multiple subjects, wherein each subject has received a (medical or other) treatment, and wherein the method comprises comparing the states of each cell population and wherein the state of a cell population is indicative of therapeutic success of a medical treatment. The invention further relates to software configured for a method to determine physically interacting cells in a cell population from a flow-cytometry data set and a computer-readable storage device, comprising the software. BACKGROUND OF THE INVENTION Novel immunotherapies, that leverage a patient's own immune system, have transformed the treatment landscape for various cancers and autoimmune-related disorders and offer significant potential for the creation of innovative treatments for a wide array of other diseases (Waldman et al., 2020). However, therapy resistances and poor response rates remain major limitations of immunotherapies and cell-based medicines such as CAR-T cells. Moreover, immunotherapies are exorbitantly expensive, imposing a major financial burden on healthcare systems world-wide (Schaft et al., 2023). Consequently, precision diagnostics capable of accurately identifying patients who would truly benefit from specific immunotherapies can direct more effective treatments to resistant patients, mitigate treatment-related burden, and simultaneously save hundreds of millions of Euros currently spent on ineffective therapies. Given the recent surge in the development of new immunotherapies for both cancer and non-cancer conditions, innovative solutions that facilitate personalized treatment approaches will become critically important. While few biomarkers have been identified that are associated with response to immunotherapies, current predictors remain largely insufficient (Pilard et al., 2021). The limited success of predictive biomarkers can likely be attributed to current strategies that focus particularly on clinical parameters, and on the abundance of individual markers, cell types, and states, such as quantities of T cell subsets or their phenotypic (dys)functionality (Pilard et al., 2021; Duell et al., 2017). However, the mechanistic action of all immunotherapies is based on enabling, blocking, or modulating interactions among immune cells or between immune and cancer cells, factors that are currently not taken into consideration in predicting therapy response. Not only immunotherapies, but many fundamental processes in life are shaped by physical interactions between cells, including the orchestration of organismal development, tissue homeostasis and immunity (Armingol et al., 2021; Armingol et all., 2024; Cooper et al., 2000; Bechtel et al., 2021). Notably, the immune system is one of the most dynamic biological systems in mammals, operating through an exceptionally complex network of intercellular signaling mediators and cell-cell interactions. During immune responses, a highly ordered sequence of antigen-dependent and antigen-independent interactions among various immune cells collectively orchestrates a comprehensive response of the immune system (Yu et al., 2019). In this process, transient cellular interactions act as central hubs for information processing and decision making, collectively driving the outcome of immune responses in diverse physiological and pathological states. Thus, cellular interactions play a decisive role in regulating the function of the immune system. Depending on the type of stimulation of the immune cells by cytokines or other immune cells, immunosuppressive or activating states of the target cells can arise. This is relevant, for example, for the development of therapies for the treatment of solid tumors by immunotherapies such as therapeutic immune cells, as immunosuppressive endogenous cells in the so-called tumor microenvironment may impede the efficiency of the therapeutic immune cells such as therapeutic T cells against the solid tumor. While single-cell genomic technologies have significantly advanced our understanding of cellular ecosystems in health and disease, the spatial context of cells in tissues is lost when using these technologies. To overcome this limitation, spatial transcriptomic and high-plex imaging technologies have been developed (Chen et al., 2015; Borm et al., 2023; Shah et al., 2016; Gyllborg et al., 2020; Biancalani et al., 2021; Sountoulidis et al., 2020; Rodriques et al., 2019; Stickels et al., 2021; Stahl et al., 2016; Goltsev et al., 2018; Viratham et al., 2020; Baccin et al., 2020). However, these technologies can only investigate cellular interactions indirectly and only in solid tissues, are difficult to scale up and are very expensive. In addition, these technologies require maximally invasive biopsies. A few companies have recently commercialized such imaging techniques for cellular interactions. However, these methods further focus only on interactions between a few preselected cell types that form in a culture dish ex vivo. These methods are therefore limited with regards to a number of pre-known cell types and their physical interactions, have a low throughput and are associated with high costs. Consequently, these technologies are unsuitable for the study of complex, transient and dynamically changing cellular interactions that occur in all tissues. In recent years, specialized technologies to study cellular interactions through single-cell transcriptomic profiling of physically interacting cells, such as physically interacting cells – sequencing (PIC-seq) have been developed (García-Guerrero et al., 2018; Andrews et al., 2021; Wheeler et al., 2023; Giladi et al., 2020; Halpern et al., 2018; Manco et al, 2021; Boisset et al.; 2018). However, these technologies as well focus on the interaction of pre-defined cell types and are thus limited by the pre-defined cell types and their cellular throughput and costs. PIC-seq such as disclosed in Giladi et al., 2020, combines the isolation of interacting cells using FACS sorting with a subsequent scRNA-seq transcriptome analysis and computer-assisted modelling of said transcriptome data in order to analyze cellular interactions and their influence on gene expression. A similar PIC-seq approach is disclosed in WO 2021 / 161310 A1. Therein, cells derived from a tissue are stained for two mutually exclusive markers, measured by flow cytometry and subsequently sorted. The cells are assigned and sorted into two singlet populations and one interacting population (“cluster”) by manual gating according to the two mutually exclusive markers, wherein the interacting population is double positive for the two markers. After identifying the interacting cell population in the sample, a transcriptome analysis is performed and an algorithm for analyzing said transcriptome data of the interacting cells is applied in order to identify the cell type, cell state and influence of interaction on gene expression of the interacting cells. However, these PIC-seq approaches can only be used to analyze known interactions of pre-defined cell-types and are thus not suitable for an evaluation of transient, dynamically changing and potentially unknown cellular interactions of a high number of cell types in different types of tissues. In parallel, approaches using murine reporter mouse lines have been developed that track past interactions upon transient cellular engagement (Pasqual et al., 2018; Nakandakari-Higa et al., 2023; Zhang et al., 2022; Medaglia et al., 2017). However, these technologies are dependent on complex mouse models and are not applicable to study human samples. Other technologies for investigating cellular interactions include the ‘LIPSTIC’ technology, which allows cell interactions in mice to be observed directly in vivo (Schiepers et al., 2023). However, this technology is also not applicable to human samples. US 2016 / 152716 describes methods for identifying immunobinders, such as scFv antibodies, capable of specifically binding to cell surface antigens. The method described therein relies on a traditional two-marker gating strategy to identify one specific kind of predetermined interaction, without employing any form of clustering algorithm. Bono et al ("A flow cytometric procedure for the quantification of cell adhesion in complex mixtures of cells", Journal Of Immunological Methods, vol.223, no.1, February 1999, 27-36) discloses a non-radioactive cytometry-based assay that permits the simultaneous quantitation of cell adhesion of distinct subsets of cells contained in a mixture without any previous fractionation. The method is based on counting the absolute number of cells, and does not employ a clustering algorithm. Serda ("Abstracts of the 12th European Society for Clinical Cell Analysis (ESCCA) Euroconference", Cytometry Part B, Clinical Cytometry, no.6, 2012-09-07, p.376, abstract LE-4- 01) discloses a method in which cellular association with particles causes an increase in orthogonal light scatter, providing a metric for particle adhesion using flow cytometry. Triana et al ("Single-cell proteo-genomic reference maps of the hematopoietic system enable the purification and massive profiling of precisely defined cell states", Nature Immunology, vol.22, no.12, 2021-11-22, 1577-1589) teach an analysis of single-cell proteogenomic data, using a dataset that does not include cellular scatter properties. Despite these technologies known in the prior art, improved or alternative means for characterizing cellular interactions are required. In particular, means are required that are applicable to human samples, and allow to quantitatively map a high number of cellular interactions among all relevant cell types of a given biological sample or system at low cost, less resource requirements and fast processing times. Thus, means are required that enable evaluation of the dynamic interaction of cells for example across entire organs, organisms and patient cohorts, thereby providing improved assessment and prediction of therapy efficiency and therapy response, such as for immunotherapies. SUMMARY OF THE INVENTION In light of the prior art, the technical problem underlying the present invention is to provide improved and / or alternative means for identifying and quantifying physically interacting cells in a biological sample that overcome the disadvantages of the prior art. Another problem underlying the present invention is to provide improved and / or alternative means for identifying and quantifying physically interacting cells in a biological sample that are cost-efficient, fast, have a high precision and can be performed at a high number of different cell types and at high-scale. Another problem underlying the present invention is the provision of improved or alternative computer-implemented means for the assessment of a high number of complex and dynamic physical interactions of cells in a biological sample. Another problem underlying the present invention is the provision of improved or alternative means for the assessment of intracellular signaling in physically interacting cells in a biological sample. Another problem underlying the present invention is the provision of improved or alternative means for the assessment of intracellular signaling in response to physical cellular interactions. Another problem underlying the present invention is the provision of improved and / or alternative means for diagnostics, disease monitoring and / or therapy decision guiding in a subject by characterizing physical interactions of cells in a sample of the subject. Another problem underlying the present invention is the provision of improved and / or alternative means for diagnostics, disease monitoring and / or therapy decision guiding in a subject by characterizing physical interactions of cells and intracellular signaling in response to the physical interactions in a sample of the subject. Another problem underlying the present invention is the provision of improved and / or alternative means for assessing therapy response and therapy efficacy in a subject by characterizing physical interactions of cells in a sample of the subject. Another problem underlying the present invention is the provision of improved and / or alternative means for assessing therapy response and therapy efficacy in a subject by characterizing physical interactions of cells and intracellular signaling in response to the physical interactions in a sample of the subject. These problems are solved by the features of the independent claims. Preferred embodiments of the present invention are provided in the dependent claims. In one aspect the invention relates to a method for determining physically interacting cells in a cell population, comprising a. providing a flow-cytometry data set of said cell population, said data set comprising data on scattering properties and expression of cell-type specific surface markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward and / or side scatter, b. applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific surface markers, c. wherein the expression of one or more cell-type specific surface marker(s) and / or combinations of cell-type specific surface markers in a cluster is indicative of a specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster. As described herein, a method and related aspects are therefore provided that allow the identification and quantification of complex cellular landscapes, such as the immune system, and their physical interactions. The present invention advantageously provides a universal and flexible cytometry-based method allowing to accurately discriminate between single cells and physically interacting cells in a sample based on flow cytometry data comprising data on scattering properties and the expression of cell-type specific surface markers. Thereby, the method of the present invention advantageously allows to evaluate millions of cellular events in complex biological samples, while providing efficient discrimination and quantification of single and physically interacting cells in said samples based on an automated multiparametric clustering algorithm. Compared to single-cell genomics- and transcriptomics- based workflows such as disclosed WO 2021 / 161310 A1, the ultra-high cellular throughput, rapid processing time and low costs associated with the presented cytometry-based approach advantageously enable the analysis of millions of cellular events within short time periods by simple means of flow cytometry. In one embodiment the flow cytometry data set additionally comprises data on the expression of one or more intracellular signaling markers, preferably one or more cell-type specific intracellular signaling markers. In one embodiment the flow cytometry data set additionally comprises data on the presence of one or more intracellular signaling markers, preferably one or more cell-type specific intracellular signaling markers. In one embodiment the expression of one or more intracellular signaling markers in a cluster comprising two or more physically interacting cells is indicative of a changed intracellular signaling in response to the physical cellular interaction. In one embodiment the presence of one or more intracellular signaling markers in a cluster comprising two or more physically interacting cells is indicative of a changed intracellular signaling in response to the physical cellular interaction. The method of the present invention advantageously and surprisingly not only allows to identify and quantify complex physical interactions between cells in a sample, e.g., between different types and subtypes of immune cells, but simultaneously also allows to assess the influence of physical interaction on the intracellular signaling and thus status and functionality of the cells. The present invention thereby advantageously provides a tool by which by simple, straightforward and rapid means, i.e. based on flow cytometry data, a comprehensive, detailed and accurate analysis of cellular interaction, cell status and functionality can be obtained. Based on this analysis the method of the present invention advantageously allows the systematic and detailed identification of cellular mechanisms underlying therapy response and the identification of novel therapy targets. In one embodiment the one or more intracellular signaling markers, preferably the cell-type specific intracellular markers, comprise at least a phosphorylation marker. In one embodiment the one or more intracellular signaling markers, preferably the cell-type specific intracellular markers, comprise at least a phosphorylated intracellular molecule, preferably an intracellular protein. In one embodiment the one or more intracellular signaling markers, preferably the cell-type specific intracellular markers, are selected from the group consisting of Nuclear factor kappa B (Nfkb) (pSer529), Extracellular-signal Regulated Kinase 1 and 2 (ERK1 / 2) (Thr202 / Tyr204), Src homology region 2 domain-containing phosphatase-2 (SHP-2) (pTyr542), phosphorylated protein kinase B (pAkt) (pSer473), Signal transducer and activator of transcription 5 (STAT5) (pY694), Phospholipase C gamma 1 (PLCγ1) (pY783), Phospholipase C gamma 2 (PLCγ2) (pY759), Phospho-S6 Ribosomal Protein (pRPS6) (Ser235 / 236), CD247 (pY142), Signal transducer and activator of transcription 3 (STAT3) (pY705), Phosphatase and Tensin homolog (PTEN), T cell factor 1 (TCF1), Nuclear receptor 4A1 (NUR77) and Ki67. In one embodiment the term in parentheses refers to the position of phosphorylation of the intracellular marker. In one embodiment the term in parentheses (e.g., pSer529) refers to the position of phosphorylation of the intracellular molecule, preferably the intracellular protein Another aspect of the invention relates to a method for determining physically interacting cells in a cell population, comprising a. providing a flow-cytometry data set of said cell population, said data set comprising data on scattering properties and expression of cell-type specific markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward and / or side scatter, b. applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific markers, c. wherein the expression of one or more cell-type specific marker(s) and / or combinations of cell-type specific markers in a cluster is indicative of a specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster. In one embodiment the cell type specific markers are a cell-type specific surface marker and / or an intracellular signaling marker. In one embodiment the cell type specific markers are a cell-type specific surface marker and / or a cell-type specific intracellular signaling marker. Another aspect of the invention relates to a method for determining physically interacting cells in a cell population, comprising a. providing a flow-cytometry data set of said cell population, said data set comprising data on scattering properties and expression of cell-type specific intracellular signaling markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward and / or side scatter, b. applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific intracellular signaling markers, c. wherein the expression of one or more cell-type specific intracellular marker(s) and / or combinations of cell-type specific intracellular signaling markers in a cluster is indicative of a specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster. In one embodiment, intracellular signalling be determined as consequence of cellular interactions. In one embodiment, intracellular signalling comprises phosphorylation of molecules, preferably intracellular proteins. In one embodiment, the method comprises a high-plex cytometry panel. In one embodiment, the high-plex cytometry panel comprises an antibody to detect phosphorylation of intracellular molecules, including proteins. In one embodiment the phosphorylated intracellular molecule is selected from the group consisting of Nfkb (pSer529), ERK1 / 2 (Thr202 / Tyr204), SHP- 2 (pTyr542), pAkt (pSer473), STAT5 (pY694), PLCγ1 (pY783), PLCγ2 (pY759), pRPS6 (Ser235 / 236), CD247 (pY142), STAT3 (pY705), PTEN, TCF1, NUR77 and Ki67. In one embodiment, the method comprises the determination of phosphorylation of molecules, preferably the phosphorylation of intracellular proteins. By applying the automated clustering algorithm to the flow cytometry data set according to the method of the present invention, it is advantageously and surprisingly possible to analyze physical interactions even within highly complex biological samples comprising large numbers of different cell types and cell states with high accuracy without relying on transcriptome analysis. In one embodiment the flow cytometry data set does not comprise transcriptome data of said cell population. In one embodiment the flow cytometry data-set does not comprise data obtained by transcriptome analysis of said cell population. In one embodiment the automated clustering algorithm does not assign the cells to a cluster according to transcriptome data of said cell population. In one embodiment the flow cytometry data the automated clustering algorithm is applied to does not comprise transcriptome data of said cell population. It is surprisingly not necessary to have knowledge of every single cell type or cell state present in said sample prior to performing the method. Thus, even unknown interactions and cellular signaling within complex samples can be analyzed by the method of the present invention based on a flow cytometry data set comprising data on scattering properties and expression of cell-type specific surface markers in a time-efficient and accurate manner. As disclosed in the examples below, the inventors further demonstrate the utility of the method of the present invention in deciphering the kinetics, mode-of-action and response mechanisms of immunotherapies, and for the quantitative dissection of complex, organism-wide immune cell interaction networks in vivo. Thereby, in contrast to methods for mapping physical interactions of cells in the prior art, the method of the present invention advantageously shows high throughput, cost effectiveness, low processing times, low requirements with regards to technical prerequisites and ease of implementation. The method of the present invention can be employed based on any flow cytometry data such as data obtained by any multi-color fluorescence flow cytometer such as a conventional or a spectral flow cytometer. Further, the method of the present invention can advantageously be applied to identify and quantify cellular interactions both in newly acquired and pre-existing cytometry datasets. In contrast to recently developed technologies that map past cellular encounters using transgenically engineered mouse models (Pasqual et al., 2018; Nakandakari-Higa et al., 2023; Zhang et al., 2022; Medaglia et al., 2017), the method of the present invention can be readily applied to any cellular suspension that is compatible with flow cytometry analysis, and does not rely on reporter mouse lines. Other methods for identifying physically interacting cells, such as US 2016 / 152716, Bono et al, 1999, and Serda, 2012 (see above), also fail to teach or suggest important features of the present method. For example, none of these methods employs a clustering algorithm as described herein in order to detect the interacting cells. By contrast, the inventive approach described herein, in embodiments, applies simultaneous clustering across multiple cellular parameters, including both scatter properties and surface markers. This allows for a high- dimensional and data-driven identification of all cell-cell interactions in a sample, that extends beyond the limitations of predefined gating. This enables an unbiased identification of cell-cell interactions, avoiding the disadvantages of traditional gating strategies that are susceptible to errors introduced by the user. Furthermore, the methods in the prior art are only configured to detect antigens specifically pre-selected, according to cell types to be identified, or to predefined queries. None of these methods provide information about extracellular or intracellular proteins presented by the cells beyond analysis of specifically pre-selected antigens. Triana et al, 2021 (see above), employs a clustering algorithm. However, the data analyzed in Triana et al is single-cell proteogenomic data, which is not compatible with the method described herein. In embodiments of the present invention, the algorithm employed requires specific input parameters such as cellular scatter properties that are not present in single-cell proteogenomics data, and therefore not present in the dataset described in Triana et al. Given the fundamental differences between the types of data to be analyzed, when comparing the prior art mentioned above and the present invention, a specialist in the field would not reasonably combine these approaches to develop the procedure of the invention disclosed herein. Further advantages are evident in the present invention when compared to earlier strategies, even those employing clustering algorithms. For example, the present invention provides, as shown in the examples below, advantages that arise from the use of the automated clustering algorithm on both scatter properties and surface markers. To test whether the data for scatter properties or surface markers may be effective on their own, we performed Louvain clustering on cell type markers only, followed by FSC ratio-based classification into singlet and multiplet clusters (Figure 15H-K). While this approach outperformed FSC ratio only classification, it remained inferior to using all important features (Figure 1K). In contrast, incorporating both cell type markers and scatter properties, as defined for the present invention, into the clustering, followed by FSC ratio-based classification into singlet and multiplet clusters, yielded results comparable to those achieved when all important features including image-based features were utilized (Figure 1G,H,K,L, Figure 15L-N). This result was reproducible across various cluster-resolutions (Figure 1I). A comparison between different clustering methods showed Louvain clustering, alongside others, as a most accurate approach (Figure 15O) but other clustering methods are also suitable. Of particular note are Fig.1K, 15O and 15A. In contrast to the present invention, Triana et al, 2021 (see above), states in their methods, that the resulting MOFA dimensions were used to construct a shared nearest neighbor graph and modularity-based clustering using the Louvain algorithm. This means that the input was RNA-seq data and the sequencing proxy of the surface proteome readout was preprocessed via MOFA, which is a data type different from flow cytometry data (as used herein). In summary, although earlier approaches have used Louvain clustering, as in Triana et al, 2021 (see above), the present invention uses flow cytometry data as an input, whereas Triana et al use proteogenomic data (e.g., from CITE-seq or AB-seq). The present invention therefore employs, in some embodiments, scatter properties and their ratio (FSCratio), which are not present or not available in the data of in Triana et al. Further, in some embodiments, clustering is only one part of the algorithm employed in the examples of the invention. In addition to clustering, the invention may also use thresholding (in particular Otsu), which is not taught or suggested in Triana et al, 2021. The present invention differs from Triana et al, 2021 due to the different nature of the underlying data, and additionally with regard to non-random sampling (before the clustering), and / or normalization (after clustering). Further, the inventors have demonstrated that physical interactions of cells are advantageously stable upon freeze-thawing and can be stabilized by chemical fixation, enabling the implementation of the method of the present invention for the study of bio-banked patient material. In one embodiment, the method can be used on a publicly available cytometry dataset. In one embodiment the cytometry data set is publicly available. In one embodiment, the automated clustering algorithm is applied to a publicly available cytometry dataset. In one embodiment, the interaction of cells is analyzed using ImageStream-based flow cytometry. In one embodiment the flow-cytometry data set is obtained by image-based flow cytometry, preferably by ImageStream-based flow cytometry. In one embodiment, the interaction of cells is derived from in vivo settings or ex vivo settings. In one embodiment the sample is derived from an in vivo setting or an ex vivo setting. In one embodiment the sample is an in vivo sample or an ex vivo sample. In one embodiment, stability and quantifiability of cellular interactions is investigated in ex vivo performance experiments. As shown in the example below CytoStimTM-treated PBMCs were split, labelled with two different fluorescently labelled CD45 antibodies, recombined and processed under different cell concentrations, processing times and fixation methods. Newly acquired double-positive interactions are minimal or absent, especially compared to single- positive interactions. Prolonging the incubation time does not lead to a significant increase in newly acquired interactions, and fixation has only a negligible effect on the interactions or is not present at all. Technically, the experiments demonstrate the stability of ex vivo induced cellular interactions and provide quantitative insights into the effects of experimental settings on cellular interactions. They also allow effective quantification of cellular interactions under the conditions studied. Notably, the costs for cellular interaction mapping using the method of the present invention is orders of magnitudes lower when compared to single-cell genomics-based technologies, while its throughput is orders of magnitudes higher. The method of the present invention thus enables the study of cellular interactions and associated intracellular signaling in currently unexplored settings, such as high throughput screens, extensive time course experiments, organism-wide studies and large patient cohorts. Jointly, the aforementioned features render the method of the present invention broadly applicable to any research field where alterations in cellular frequencies, interactions and intracellular signaling may place decisive roles. These encompass, but are not limited to, basic immunology, autoimmune diseases, cancer research, infectious diseases, age-related conditions including neurodegenerative disease, drug development and personalized medicine. The inventors have thus demonstrated the utility of the method of the present invention for the characterization of cellular states and interactions induced by immunotherapies, including CAR-T cells and bispecific antibodies. The examples below illustrate how kinetics, mode-of-action and mechanisms governing therapy response of immune therapies can be quantified at ultra-high precision and cellular resolution by the method of the present invention. Due to its high scalability and low costs, the method of the present invention can be readily implemented into any cytometry-based assay and large-scale screens to identify or prioritize candidate immunotherapy drugs. Major bottlenecks in the current implementation of immunotherapies are heterogenous response rates and the development of therapy resistance. Here, the inventors have demonstrated the utility of the method of the present invention for the systematic identification of cellular mechanisms underlying therapy response to, by way of example, Blinatumomab, a commonly used immunotherapeutic drug. The inventors have further demonstrated the utility of the method of the present invention for predicting binding between immune cells and / or interactions between immune cells, e.g. between B and T cells, mediated through an immunotherapeutic drug or an immunotargeting drung and thus predicting the effectiveness of an immunotherapy or an immunotargeting drug. The approach according to the present invention confirmed known parameters associated with therapy response and identified novel, independent parameters, including cellular interactions, as new biomarkers. Due to its rapid turnaround time and low costs, the method of the present invention, entailing ex vivo treatment of primary patient samples followed by cellular landscaping and interaction mapping, provides an ideal basis for the development of a companion diagnostic test for personalized response prediction. As the mode- of-action of immunotherapies is broadly based on the redirection of cellular interactions among immune and cancer cells, the method of the present invention advantageously provides a tool for the exploration, therapeutic development, and personalized response prediction of immunotherapies. In one embodiment, co-expression of the B cell marker CD19 and the T cell markers CD3 and CD4 within an interacting cluster indicates an interaction of a B and T cell The extremely high throughput of the method of the present invention enables the quantitative dissection of complex interaction networks across entire organ systems and organisms. In this context, within the examples below the cellular interaction networks in response to virus infection in mice across distinct time points and immune organs were mapped. The method of the present invention revealed organ-specific shifts in single-cell landscapes and cellular interaction networks underlying antiviral immune responses and identified fundamental differences in cellular interaction dynamics between primary and secondary lymphoid organs. Importantly, these data confirmed previously known and identified new cellular interaction patterns and provide a first quantitative framework for a systems level understanding of how complex cellular interaction networks cooperate across immune organs to jointly orchestrate immune responses. The method of the present invention is thus of great utility to decipher fundamental principles of multi-layered immune cell crosstalks underlying complex (patho-) physiological processes, such as age-related decline of the immune system and the associated development of diseases, or cancer-immunity. Collectively, the method of the present invention advantageously represents a highly versatile and scalable cytometry-based framework that can be readily implemented for the joint mapping of cellular immune landscapes and their physical interactions with a wide range of applications across a variety of research and medical fields including basic biology, advanced immunology, cancer research and applied biomedicine. In one embodiment the expression of one or more cell-type specific surface marker(s) and / or combinations of cell-type specific surface markers in a cluster is indicative of the number and / or frequency of one or more specific cell type(s) and / or cell state(s) in the cluster. In one embodiment, the flow-cytometry data set comprises data on the expression of at least 5 cell-type specific surface markers (preferably at least 10 cell-type specific surface markers, more preferably at least 25 cell-type specific surface markers), such as at least 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38,39, 40, 45, 50 or 60or more cell-type specific surface markers. In one embodiment, the flow-cytometry data set comprises data on the expression of at least one intracellular marker, preferably a cell type specific intracellular marker such as at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38,39, 40, 45, 50 or 60 or more intracellular markers. In one embodiment, the flow-cytometry data set comprises data on the presence of at least one intracellular marker, preferably a cell type specific intracellular marker such as at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38,39, 40, 45, 50 or 60 or more intracellular markers. In particular, by applying the automated clustering algorithm, the method of the present invention advantageously allows to identify and quantify physically interacting cells even in complex samples comprising a high number of different cells. It is particularly advantageous to determine the expression of a large number of cell-type specific surface markers and / or intracellular markers in such complex samples such as by using a by a marker panel with at least 5 markers (selected from cell-type specific surface markers and / or intracellular markers) in spectral flow cytometry in order to enhance accuracy of the identification and quantification of physically interacting cells. In conventional flow cytometry and the method of the prior art such as WO 2021 / 161310 A1, and others mentioned above, manual gating strategies are commonly used to analyze specific cell populations based on a certain predefined set of a low number of markers (such as two markers in WO 2021 / 161310 A1). However, manual gating strategies are unsuitable and unfeasible when analyzing complex samples and flow cytometry data comprising data on the expression of a high number of cell-type specific surface markers. The method of the present invention thus advantageously proved means for analyzing complex samples in a cost- and time- efficient manner with high precision. In addition, manual gating strategies can rely on subjective decisions, including on which position to place the gates, sample-to-sample variations and / or experiment-to- experiment variations, so objectivity may be lacking. In one embodiment, the method does not comprise manual and / or predefined gating for a particular cellular set or subset, as is typically employed in traditional flow cytometry techniques, for example in WO 2021 / 161310 A1. Manual gating is limited to a predefined set of a small number of markers, is error prone due to the subjective nature of gating strategies, and thus works against the unbiased analysis of a high number of cell-type specific markers and / or different cell subsets. In one embodiment, the method does not comprise analysis of cell-interactions caused by cell- bound scFv antibodies interacting with target antigens on the surface of other cells, for example, as described in US 2016 / 152716. In one embodiment, the method does not comprise analysis of cell-cell interactions caused by cell-bound scFv antibodies interacting with a target antigen. In one embodiment the expression of at least two distinct cell-type specific surface markers and / or combinations of cell-type specific surface markers in a cluster, said cluster having a scattering property above a threshold value, is indicative of the presence of at least two physically interacting cells of different cell types and / or cell states in the cluster. In one embodiment, said scattering properties comprise forward scatter area (FSC-A), forward scatter width (FSC-W), side scatter width (SSC-W) and / or forward scatter height (FSC-H). In one embodiment, the scattering properties comprise a ratio of the signal intensity of FSC-A and FSC-H (FSC-ratio). It is particularly advantageous that by the method of the present invention physically interacting cells can be identified and quantified by using flow cytometry data without relying on transcription data such as the methods of the prior art. In particular, the scattering properties of the cells can be obtained easily and quickly by flow cytometry without additional steps such as staining or labeling of the cells. The FSC-ratio enables a particularly good differentiation between individual cells and physically interacting cells within a cluster to which the cells were assigned by the automated clustering algorithm according to the present invention. Although the FSC-ratio has been used previously to identify and exclude physically interacting cells in flow cytometry analysis, in embodiments of the present approach, the method uniquely employs the selection of physically interacting cells based on scattering properties, such as the FSC-ratio, for further characterization using the clustering described herein. In one embodiment the automated clustering algorithm is an unsupervised clustering algorithm. In one embodiment the automated clustering algorithm is a semi-supervised clustering algorithm. One decisive advantage of the method of the present invention, in particular of the automated clustering algorithm, is the unsupervised or semi-supervised and thus unbiased analysis of physically interacting cells in a sample meaning that not only pre-defined cells and interactions can be analyzed but also unknown interactions within complex samples thus enabling to analyze even unknown interactions in complex cellular landscapes. No or almost no prior knowledge of the samples to be analyzed is required. In contrast, the prior art only assesses pre-known interactions between cells and the effects of their interaction on the gene expression patterns andfunctions of the two cells by virtue of transcriptome analyses. Thus, the present invention allowsthe analysis of all cell types and states, not just pre-selected specific types, and maps the interactions between all cell types, not just those of a cell type of interest. In one embodiment, applying the automated clustering algorithm comprises performing a connectivity model, a centroid model, a distribution model, a density model, a graph-based model, a self-organizing map, hierarchical clustering, K-means clustering and / or a neural model. In one embodiment, applying the automated clustering algorithm comprises performing a connectivity model, a centroid model, a distribution model, a density model, a graph-based model, a self-organizing map, and / or a neural model. In one embodiment, the automated clustering algorithm comprises performing a Louvain clustering, Leiden clustering, FlowSOM clustering, Phenograph clustering and / or K-means clustering. In one embodiment, the automated clustering algorithm comprises performing a Louvain clustering. In one embodiment K-means clustering is also termed kmeans clustering. In one embodiment, the automated clustering algorithm comprises performing a Louvain clustering, Leiden clustering, FlowSOM clustering, FlowMeans clustering, Phenograph clustering, HDBSCAN clustering, Immunoclust clustering, Rclusterpp clustering and / or K-means clustering. In one embodiment, the automated clustering algorithm comprises performing a Louvain clustering, Leiden clustering, FlowSOM clustering, FlowMeans clustering, Phenograph clustering, HDBSCAN clustering, Immunoclust clustering and / or Rclusterpp clustering. In one embodiment, the automated clustering algorithm does not comprise analysis of data obtained by manual gating. In one embodiment, the automated clustering algorithm is configured for characterizing physically interacting cells, and / or cells based on a cell-cell interaction. In one embodiment, the automated clustering algorithm is configured to assess a combination of scattering properties and expression of one or more cell-type specific surface markers, wherein said scattering properties comprise at least one measurement of forward and / or side scatter. In one embodiment, the automated clustering algorithm is configured to perform, or performs, simultaneous clustering of cells based on scattering properties and expression of one or more cell-type specific surface markers. In one embodiment, the clustering of cells based on scattering properties and marker expression does not occur sequentially. In embodiments, the automated clustering algorithm advantageously maps any interaction among all cell types in a given sample at the same time. As described herein, the simultaneous clustering of cells based on scattering properties and marker expression leads typically to a higher accuracy, for example leading to fewer false positives and / or fewer false negatives. Embodiments of the invention that employ a clustering algorithm, configured to cluster cells simultaneously based on both scattering properties and marker expression, may exhibit these advantages. Additional clustering algorithms are known to a skilled person and may be selected and employed in the present method accordingly. In one embodiment, the scattering properties comprise the FSC-ratio and a threshold value for the FSC-ratio is determined by an automated thresholding method. In one embodiment the automated thresholding method is a global thresholding method. In one embodiment an FSC-ratio below the threshold value indicates the presence of a single cell in the cluster and an FSC-ratio equal or above the threshold value indicates the presence of two or more physically interacting cells in the cluster. In one the scattering properties comprise the FSC-ratio and a threshold value for the FSC-ratio is determined by Gaussian Mixture Model thresholding, Huang thresholding, Intermodes thresholding, IsoData thresholding, kmeans thresholding, Li thresholding, manual gating, mean thresholding, Otsu thresholding, Renyi Entropy thresholding, Shanbhag thresholding and / or triangle thresholding. In one the scattering properties comprise the FSC-ratio and a threshold value for the FSC-ratio is determined by Gaussian Mixture Model thresholding, Huang thresholding, Intermodes thresholding, IsoData thresholding, kmeans thresholding, Li thresholding, mean thresholding, Otsu thresholding, Renyi Entropy thresholding, Shanbhag thresholding and / or triangle thresholding. In one embodiment, the scattering properties comprise the FSC-ratio and a threshold value for the FSC-ratio is determined by Otsu thresholding. In one embodiment K-means thresholding is also termed kmeans thresholding. In one embodiment, the automated clustering algorithm comprises performing a Louvain clustering, the scattering properties comprise the FSC-ratio and a threshold value for the FSC- ratio is determined by Otsu thresholding. In one embodiment an F1 score indicates the accuracy of the thresholding method determining a threshold value for the FSC-ratio. In one embodiment the F1 score is between 0.40 to 1,00, such as 0,40, 0.42, 0.44, 0.45, 0.46, 0.48, 0.50, 0.52, 0.54, 0.55, 0.56, 0.58, 0.60, 0.62, 0.64, 0.65, 0.66, 0.68, 0.70, 0.72, 0.74, 0.75, 0.76, 0.78, 0.80, 0.82, 0.84, 0.85, 0.86, 0.88, 0.90, 0.92, 0.94, 0.95, 0.96, 0.98, 0.99 or 1.00, preferably between 0.45 and 1.00, more preferably between 0,50 and 1.00. In one embodiment the F1 score is used to distinguish singlets from multiplets, wherein the F1 score is depending on the thresholding method used. In one embodiment, the method comprises the steps of acquiring flow cytometry data, performing FSC ratio analysis and applying the Otsu threshold method to discriminate between single cells and multiplets. In one embodiment, clustering-based approaches for simultaneous multiplet discrimination and annotation are used for improved classification and identification of interacting cells. In another embodiment, Louvain clustering is additionally used for improved classification and annotation of interacting cells. In one embodiment, annotation of interacting cells based on the co-expression of mutually exclusive lineage-defining markers is enabled by classifying clusters based on the FSC ratio into singlets versus multiplets. This is advantageous compared to the use of scatter properties only. In one aspect the invention relates to a cytometry-based analysis comprising the steps of data acquisition, multiplet identification using the FSC ratio and the Otsu thresholding method, and cluster analysis for comprehensive annotation and quantification of cell interactions. The use of FSC ration analysis in combination with the Otsu threshold method and Louvain clustering enables precise differentiation between single cells and multiplets, resulting in high accuracy of classification. By applying clustering methods and using specific marker expressions, physical cell interactions and associated intracellular signaling can be effectively identified and analyzed. The framework of the present invention is flexible and can be adapted to different analysis methods and parameters to enable broad application. The methods and procedures described herein offer a high reproducibility of results, which increases the reliability of the analysis. By integrating cytometric and image-based data, more comprehensive and in-depth analyses of cell interactions can be performed, leading to a better understanding of cellular mechanisms. In one embodiment, Louvain clustering is preferably used as clustering method. Louvain clustering delivers advantageously accurate results. In another embodiment, Otsu thresholding of the FSC ratio and Louvain clustering are preferably applied. In one embodiment, the method of the present invention comprises a normalization step. In one embodiment, the normalization step comprises the normalization of the frequency of a cellular interaction to all live events, indicating the prevalence of the interactions in relation to all cells and other interactions. In one embodiment, normalization step comprises the normalization of the frequency of a given type of interaction to all interactions, indicating how the relative composition of the type of cellular interaction changes across conditions. In one embodiment, the normalization step comprises calculation of the harmonic mean indicating the expected interaction frequency based on singlet frequencies. The expected interaction frequency can then be compared to the observed interactions to assess relative enrichment. The normalization method is selected depending on the biological questions to be addressed. In one embodiment, the normalization step is applied separately. In one embodiment, the normalization step is applied subsequent to applying the automated clustering algorithm. In one embodiment, the normalization step is applied in combination with the steps of the method of the present invention. The normalization method is selected depending on the biological questions to be addressed. The normalization approaches described offer increased precision in the analysis of cell interactions. This enables a more precise analysis and interpretation of cell interaction data. Analyzing the relative frequencies of a particular interaction type among all interactions shows how the composition of cell interactions changes under different conditions and allows the dynamics and changes in cell populations to be understood. Using the harmonic mean in scenarios with unbalanced or rapidly changing frequencies of interaction partners results in the calculation of the expected interaction frequency based on singlet frequencies. This allows comparison with observed interactions to assess relative enrichment and provides a robust way to assess the relative frequency and importance of cell interactions. In one embodiment, the method additionally comprises a non-uniform sampling of said data set prior to applying the clustering algorithm preferably by applying a sampling algorithm based on dictionary learning. Due to the enormous quantity of cells typically analyzed in flow cytometry experiments the use of downsampling techniques such as a sampling algorithm based on dictionary learning advantageously enables efficient data management. Commonly used random downsampling approaches often do not account for rare events (such as rare cell types within the cell population analyzed), thus leading to potentially not taking these cells into account and thus leading to potential gaps in the analysis of the cell population with regards to physically interacting cells. To address this, novel methods such as sampling algorithms based on dictionary learning have been developed that are particularly effective in preserving rare events, including cellular interactions. In one embodiment the method additionally comprises spectral compensation, transformation and scaling of said data set prior to performing non-uniform sampling. In one embodiment the method additionally comprises adjusting the clustered data for a number of free interacting partners, preferably comprising calculation of the expected rate of physically interacting cells within the cell population by calculating the harmonic mean of the rates of the cell types present in said cell population. In one embodiment providing the flow cytometry data set comprises a. obtaining a sample of a subject comprising a cell population, b. preparing a cell suspension from the sample, c. addition of multiple labels (preferably fluorophore-coupled binding reagents), each binding a distinct cell-type specific surface marker (preferably adding at least 5, 10, or 15 labels, more preferably at least 20 labels for distinct cell-type specific surface markers), such as 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 45, 50, 60, 70, 80 or 100 cell-type specific surface markers and d. determining scattering properties of the cells and intensities of said labels by flow cytometry, wherein the intensity of said labels is indicative of the expression of surface markers of said cell population. In one embodiment providing the flow cytometry data set additionally comprises addition of one or more labels (preferably fluorophore-coupled binding reagents), binding an intracellular signaling marker preferably a cell-type specific intracellular signaling marker, such as 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 45, 50, 60, 70, 80 or 100 intracellular signaling markers. In one embodiment, the cell-type specific surface markers are selected from the group consisting of CD45, CD8, CD16, CD4, CD141, CD1c, HLA-DR, TCRab, CD27, CD3, CD20, CD19, CD14, CD11b, CD11c, CD33, CD56, CD138, CD24, CD45RO, CD16, CD69, CD34, CD154, CD45RA, CD197, CD123, MHCII, CD45.2, CD25, CD45.1, CD86, CD44, TCRvb5, CD8a, Ly6C, CD41, F4.80, CD117, NK1.1, CD93, CD115, Ly6G, IgM, CD150, TCRgd, CD16.32, B220, CD28, CD7, CD38, CD94, CD71, CD297, CD10, CD39, CD90.1, ICOS, CD357, CD68, CD317, CD172 and TCRb, CD62L, CD43, CD64, CD57, IgD, CD30L, CD160, CD127, Tigit. In one embodiment the one or more intracellular signaling markers, preferably the cell-type specific intracellular markers, comprise at least a phosphorylation marker. In one embodiment the one or more intracellular signaling markers, preferably the cell-type specific intracellular markers, comprise at least a phosphorylated intracellular molecule, preferably an intracellular protein. In one embodiment the intracellular signaling marker is selected from the group consisting of Nfkb (pSer529), ERK1 / 2 (Thr202 / Tyr204), SHP-2 (pTyr542), pAkt (pSer473), STAT5 (pY694), PLCγ1 (pY783), PLCγ2 (pY759), pRPS6 (Ser235 / 236), CD247 (pY142), STAT3 (pY705), PTEN, TCF1, NUR77 and Ki67. In one embodiment, the sample is or comprises a body fluid from a subject, or is derived therefrom. In one embodiment, the sample is a blood sample, or sample derived from blood, preferably a sample comprising human peripheral blood mononuclear cells (PBMC). In one embodiment, the sample is derived from a tissue or an organ of a subject such as from the bone marrow, the spleen, a lymph node, the lung, the liver, the kidney, the heart, the skin or the pancreas. In one embodiment the cell population comprises immune cells. In one embodiment providing the flow cytometry data set comprises a. obtaining a sample of a subject comprising a cell population, b. preparing a cell suspension from the sample, c. addition of multiple fluorophore-coupled binding reagents, each binding a distinct cell-type specific surface marker (preferably adding at least 5, 10, 15, more preferably at least 20 fluorophore-coupled binding reagents for distinct cell-type specific surface markers), and d. determining scattering properties of the cells and fluorescence intensities of said fluorophore-coupled binding reagents by flow cytometry, wherein the fluorescence intensity of said labels is indicative of the expression of surface markers of said cell population. In one embodiment providing the flow cytometry data set additionally comprises addition of one or fluorophore-coupled binding reagents binding an intracellular signaling marker preferably a cell- type specific intracellular signaling marker. In one embodiment, said flow-cytometry data set is obtained by conventional and / or spectral flow cytometry. Advantageously, the method of the present invention can be employed based on any flow cytometry data such as data obtained by any multi-color fluorescence flow cytometer such as a conventional or a spectral flow cytometer thus showing high throughput, cost effectiveness, low processing times, low requirements with regards to technical prerequisites and ease of implementation. Further, the method of the present invention can advantageously be applied to identify and quantify cellular interactions both in newly acquired and pre-existing flow cytometry datasets such as publicly available flow cytometry data sets. In one embodiment, the physical interaction of the cells comprises a physical interaction between immune cells. In one embodiment, the physical interaction of the cells comprises a drug-induced interaction or a drug-mediated interaction. In one embodiment, the physical interaction of the cells comprises an antigen-antibody and / or receptor-ligand interaction. In one embodiment the physical interaction of the cells comprises an antigen-antibody, receptor-ligand interaction and / or ligand-ligand interaction. In one embodiment, the physical interaction of the cells comprises an immunotherapy-induced interaction, preferably induced by a therapeutic immune cell such as a CAR-T cell and / or a therapeutic binding agent antibody, preferably an antibody or fragment thereof or a bispecific antibody. In one embodiment the physical interaction of the cells comprises an immunotherapy- induced interaction, preferably induced by a CAR-T cell, an antibody or a bi-specific antibody. In one embodiment, the physical interaction of the cells comprises an infection-induced interaction, preferably induced by a viral infection. In one embodiment, the assignment of the cells to clusters according to the scattering properties and the expression of cell-type specific surface markers represents a state of the cell population (“cellular interaction landscape”). In one embodiment, the method is performed on multiple flow-cytometry data sets obtained from cell populations of multiple subjects, wherein each subject has received a (medical or other) treatment, and wherein the method comprises comparing the states of each cell population. In one embodiment, the state of a cell population is indicative of therapeutic success of a medical treatment. In one embodiment the state of a cell population is indicative of therapeutic success of a medical treatment in a subject. In one embodiment, the state of a cell population is indicative of the subject having a disease. In one embodiment, the state of a cell population is indicative of the subject being at risk of having a disease. In one embodiment, the medical treatment comprises administering an immunotherapy to a patient. In one embodiment, the medical treatment comprises administering a bispecific binding agent (e.g., antibody), and the composition and state of a cell population is indicative of therapeutic success of said medical treatment. In one embodiment, the medical treatment comprises administering a therapeutic immune cell (e.g., a CAR-T cell), and the composition and state of a cell population is indicative of therapeutic success of said medical treatment. The method of the present invention enables a skilled person to analyze cellular interactions in currently unexplored settings, such as high throughput screens, extensive time course experiments, organism-wide studies and large patient cohorts. As the method of the present invention enables the profiling of large sample collections at low costs across entire cellular ecosystems, it is ideally suited to identify novel biomarkers in large patient cohorts. In the context of the present invention, biomarkers may include the expression of individual surface markers, cell type frequencies, cellular interactions or their frequencies. These parameters can be read out simultaneously at ultra-high scale using the method of the present invention. Such biomarkers identified may be used for prognostic, diagnostic or predictive purposes. In one embodiment, the frequencies of physical cellular interactions are reported as the relative frequency among all live events or the relative frequency among all interacting cells. In another embodiment, interaction frequencies are normalized by taking the frequency of the respective interaction partners into account. Similar to biomarkers, therapy targets can be systematically identified by analyzing large patient cohorts using the method of the present invention. Targets may include surface markers specific for malignant cells or cellular interactions that underlie disease pathology and could be therapeutically targeted such as by immunotherapies or small molecules. Along the same lines, the method of the present invention can be used to validate, explore, specify or prioritize targets at ultra-high scale and low costs. Jointly, the aforementioned features render the method of the present invention broadly applicable to any technical field where alterations in cellular frequencies and interactions may place decisive roles. These encompass, but are not limited to, basic immunology, autoimmune diseases, cancer research, infectious diseases, age-related diseases, drug development and personalized medicine.In one aspect the invention relates to a software configured for performing a method to determinephysically interacting cells in a cell population from a flow-cytometry data set, the method comprising: a. receiving a flow-cytometry data set of said cell population, comprising data on scattering properties and expression of cell-type specific surface markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward or side scatter, b. applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific surface markers, c. wherein the expression of one or more cell-type specific surface marker(s) and / or combinations of cell-type specific surface markers in a cluster is indicative of one or more specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster, and d. transmitting and optionally displaying an output of the software to a user interface. In one embodiment the flow cytometry data set additionally comprises data on the expression of one or more intracellular signaling markers, preferably one or more cell-type specific intracellular signaling markers. In one embodiment the flow cytometry data set additionally comprises data on the presence of one or more intracellular signaling markers, preferably one or more cell-type specific intracellular signaling markers. In one embodiment the expression of one or more intracellular signaling markers in a cluster comprising two or more physically interacting cells is indicative of a changed intracellular signaling in response to the physical cellular interaction. In one embodiment the presence of one or more intracellular signaling markers in a cluster comprising two or more physically interacting cells is indicative of a changed intracellular signaling in response to the physical cellular interaction.In one aspect the invention relates to a software configured for performing a method to determinephysically interacting cells in a cell population from a flow-cytometry data set, the method comprising: a. receiving a flow-cytometry data set of said cell population, comprising data on scattering properties and expression of cell-type specific markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward or side scatter, b. applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific markers, c. wherein the expression of one or more cell-type specific marker(s) and / or combinations of cell-type specific markers in a cluster is indicative of one or more specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster, and d. transmitting and optionally displaying an output of the software to a user interface. In one embodiment the cell type specific markers are a cell-type specific surface marker and / or an intracellular signaling marker. In one embodiment the cell type specific markers are a cell-type specific surface marker and / or a cell-type specific intracellular signaling marker. In one embodiment the expression of one or more intracellular signaling markers in a cluster comprising two or more physically interacting cells is indicative of a changed intracellular signaling in response to the physical cellular interaction. In one embodiment the presence of one or more intracellular signaling markers in a cluster comprising two or more physically interacting cells is indicative of a changed intracellular signaling in response to the physical cellular interaction.Another aspect the invention relates to a software configured for performing a method todetermine physically interacting cells in a cell population from a flow-cytometry data set, the method comprising: a. receiving a flow-cytometry data set of said cell population, comprising data on scattering properties and expression of cell-type specific intracellular signaling markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward or side scatter, b. applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific intracellular signaling markers, c. wherein the expression of one or more cell-type specific intracellular signaling marker(s) and / or combinations of cell-type specific intracellular signaling markers in a cluster is indicative of one or more specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster, and d. transmitting and optionally displaying an output of the software to a user interface. In one aspect the invention relates to a computer-readable storage device, comprising the software according to the present invention. In one aspect the invention relates to a kit configured for performing the method of the invention. In one embodiment, the kit comprises a computer-readable storage device as described herein. In one embodiment, the kit comprises the software as described herein. In some embodiments, the kit comprises software with the clustering algorithm described herein. In some embodiments, the kit comprises means for determining the expression of cell-type specific surface markers for the cells of a cell population. In one aspect the invention relates to a flow cytometer (or flow cytometer system) configured for performing the method of the invention. In one embodiment, the flow cytometer comprises a computer-readable storage device as described herein. In one embodiment, the flow cytometer comprises the software as described herein. In some embodiments, the flow cytometer comprises software with the clustering algorithm described herein. In some embodiments, the flow cytometer comprises means for determining the expression of cell-type specific surface markers for the cells of a cell population. In some embodiments, the flow cytometer comprises means for determining the scattering properties of the cells of a cell population, preferably for determining at least one measurement of forward or side scatter. All features described in the present specification may be employed to define any other embodiment or aspect of the invention. For example, features used to describe the method may be used to describe the software, the computer-readable storage device, the kit, and the flow cytometer, and vice versa. Despite covering various embodiments or aspects, these various means of the invention are preferably unified by their unique and related ability to effectively and accurately determining physically interacting cells in a cell population as described herein. DETAILED DESCRIPTION The invention relates broadly to a method for determining physically interacting cells in a cell population. The method comprises multiple steps, each of which may be used to define the invention, without necessary limitation to all other method steps disclosed herein. The method comprises providing a flow-cytometry data set of said cell population, said data set comprising data on scattering properties and expression of cell-type specific surface markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward and / or side scatter. The flow-cytometry data set of said cell population may further comprise data on the expression of one or more intracellular signaling markers. A skilled person is aware of such surface markers and intracellular signaling markers in the context of flow cytometry approaches. Suitable non-limiting examples are disclosed herein. A skilled person is aware of measuring forward and / or side scatter in the context of flow cytometry approaches. Suitable non-limiting examples are disclosed herein. The method comprises applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific surface markers. Suitable non-limiting clustering algorithms are disclosed herein and are known to a skilled person. According to the inventive method, the expression of one or more cell-type specific surface marker(s) and / or combinations of cell-type specific surface markers in a cluster is indicative of a specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster. The combination of using cell-type specific surface markers and scattering properties of a cluster represents one preferred and unique aspect of the present invention. Non-limiting examples of how this combination enables determining physically interacting cells and optionally determining a cell state based on the clustering of the invention are disclosed herein. According to the invention the expression of one or more intracellular signaling markers marker(s) in a cluster comprising two or more physically interacting cells is indicative of a changed intracellular signaling in response to the physical cellular interaction. Cells and cellular interactions The term “physically interacting cells” according to the present invention refers to two or more cells that are in direct physical contact with each other, for example by physical contact of their cell membranes and / or by molecules present on the surface of the cells, such as by receptor- receptor interactions, ligand-receptor interactions, by junctions including tight junctions, glycolipids, gap junctions and desmosomes, or other structural proteins or cellular contact structures. In one embodiment the direct physical contact of the cells with each other may be temporal or permanent. In some embodiments the physical interaction between two or more cells may comprise the transfer of signal molecules, ions, or other molecules between the interacting cells. The physical interaction or contact between two or more cells may result in a change in cell state of one or more of the interacting cells. In some embodiments the two or more physically interacting cells may be cells of the same cell type and / or cell state. The physical interaction between two or more cells of the same cell type and / or cell state is referred to as “homotypic interaction” or “homotypic physical interaction”. In some embodiments the two or more physically interacting cells may be cells of at least two different cell types and / or cell states. The physical interaction between at least two different cell types and / or cell states is referred to as “heterotypic interaction” or “heterotypic physical interaction”. The term “cell type” refers to a category or class of cells characterized by features and functions distinctive from other cell types. These features can include without limitation structural, molecular, genetic, and functional properties that distinguish a particular cell or cell population from other cells or cell populations. Cell types can further be classified based on their physiological origin, physiological role in the organism, morphology, gene or protein expression profile. Cell type specific features for example include the expression of one or more markers on the cell surface (“cell surface markers”) or combinations of cell surface markers. A cell type may comprise several “sub cell types” or “sub types”, for example a T cell may be further distinguished into several subtypes such as, without limitation, T helper type 1 (Th1) cells, T helper type 2 (Th2) cells, T helper type 9 (Th9) cells, T helper type 17 (Th17) cells, T helper type 22 (Th22) cells, Follicular helper T (Tfh) cells, Regulatory T (Treg) cells, Natural killer T (NKT) cells and CD8+ cytotoxic T lymphocytes (CTLs). The term "cell state" refers to the functional and physical condition of a cell. This includes a variety of properties that reflect the status and activity of the cell, such as, but not limited to, its metabolic activity, cellular functions, phosphorylation of intracellular markers and proteins, marker or gene expression, protein synthesis, signaling pathways, morphology, interactions with its environment, cell division and senescence and its degree of differentiation. Cells of a particular cell type may have different cell states. For example, a T cell may be without limitation immature, mature, resting or activated. The "cell state" can be determined and characterized by various biological, chemical, and physical parameters that together provide a comprehensive picture of the cellular function and activity. Methods for determining the cell state known in the prior art include without limitation metabolomics by using for example mass spectrometry (MS) and nuclear magnetic resonance (NMR), transcriptomic analysis using RNA sequencing (RNA-Seq) and microarrays, proteomic analysis using methods such as MS and gel electrophoresis, flow cytometry, immunohistochemistry (IHC), microscopy including fluorescence microscopy, confocal microscopy, electron microscopy and live-cell imaging, epigenomic analysis such as by fluorescence in situ hybridization (FISH) for analysis the chromosomal composition and to identify genetic abnormalities influencing the cell state (Zeng, What is a cell type and how to define it?, Cell, 2022). In some embodiments of the present invention, the cell state is determined by applying an automated clustering algorithm to data obtained by flow cytometry comprising at least data on scattering properties and expression of cell type specific surface markers and optionally data on the expression and / or presence of one or more intracellular signaling markers such as phosphorylated intracellular proteins. In embodiments the expression of one or more cell-type specific surface markers and / or combinations thereof is indicative of the number and / or frequence of one or more specific cell states in a sample. In embodiments the expression of one or more cell-type specific surface markers and / or combinations thereof and the expression and / or presence of one or more intracellular signaling markers and / or combinations thereof is indicative of the number and / or frequence of one or more specific cell states in a sample. As used herein, “cell-type specific surface markers” or “cell-type discriminating marker” or “cell- type distinguishing” may be used interchangeably and refers to markers that are expressed on the surface of a cell type and thereby distinguish this cell type from others not expressing this surface marker. A cell-type specific surface marker must not necessarily be indicative or unique to one specific cell type on its own. Combinations of cell-type specific surface markers may also lead to characterization of a cell type. Some cell-type specific surface markers may be unique to one cell type or expressed in multiple cell types. The terms “surface marker”, “cell surface marker”, “biomarker” or “marker” refer to molecules that associate with and / or may be used to characterize a cell type and / or cell state, and preferably refer to a protein presented (expressed) on the surface of a cell, or in some cases to other cell surface molecules, such as carbohydrates attached to a cell membrane. Surface markers are often specific for certain cell types (“cell -type specific surface marker”). The most common cell surface markers are cluster of differentiation (CD) molecules. CD molecules are mostly membrane-bound glycoproteins, some of which are expressed in a cell-specific manner, e.g., CD3, CD4, CD8 and CD25 are commonly used to identify T cells. CD molecules can have a wide variety of functions including without limitation receptor or signaling functions, enzymatic activity and intercellular communication. By determining the expression of surface markers on the cell surface it is possible to distinguish specific cell populations from other cell populations expressing a different type of surface markers on their surface. In some embodiments the cell-type specific surface markers are selected from the group consisting of CD45, CD8, CD16, CD4, CD141, CD1c, HLA-DR, TCRab, CD27, CD3, CD20, CD19, CD14, CD11b, CD11c, CD33, CD56, CD138, CD24, CD45RO, CD16, CD69, CD34, CD154, CD45RA, CD197, CD123, MHCII, CD45.2, CD25, CD45.1, CD86, CD44, TCRvb5, CD8a, Ly6C, CD41, F4.80, CD117, NK1.1, CD93, CD115, Ly6G, IgM, CD150, TCRgd, CD16.32, B220, CD28, CD7, CD38, CD94, CD71, CD279, CD10, CD39, CD90.1, ICOS, CD357, CD68, CD317, CD172, and TCRb, CD62L, CD43, CD64, CD57, IgD, CD30L, CD160, CD127, Tigit. The term “expression of a surface marker” or “expression of a cell-type specific surface marker” refers to the process wherein a gene is transcribed into RNA and / or translated into protein, i.e., a surface marker. In some embodiments, the term “expression” or “expression level” relates to the presence or absence of a surface marker on the surface of a cell. Herein, a cell showing expression (+) also termed “positive expression” or “positive expression level” has a detectable amount of the respective surface marker on its surface such as by flow cytometry and a cell showing no expression (-) also termed “negative expression” or “negative expression level” has no or only negligible amounts of the surface marker on its surface. A skilled person is capable of determining whether any given cell expresses a surface marker and to what extent, and can differentiate between “+” and “-“ cell types without undue effort. In some embodiments the expression of a cell-type specific surface marker is determined by flow cytometry such as by labeling said surface marker with a fluorescence-coupled binding reagent and determining the fluorescence intensity after labeling. As used herein, “intracellular signaling marker” refers to biomolecules within a cell that indicate and / or intracellular signaling events. These markers preferably include phosphorylated proteins, which are proteins that have undergone phosphorylation. Within phosphorylation a phosphate group is added to one or more specific amino acids of the intracellular protein (such as serine, threonine, or tyrosine) in response to cellular signals. The phosphorylation acts as a regulatory signal, activating or inactivating the protein and thereby modulating cellular processes such as proliferation, differentiation, apoptosis, and immune responses. In one embodiment the intracellular signaling marker is cell-type specific. The term “expression of an intracellular signaling marker” or “expression of a cell-type specific intracellular signaling marker” relates to the presence or absence and / or the level of an intracellular signaling marker on the surface of a cell. A skilled person is capable of determining whether any given cell expresses an intracellular signaling marker and to what extent without undue effort. In some embodiments the expression of a intracellular signaling marker is determined by flow cytometry such as by labeling said marker with a fluorescence-coupled binding reagent and determining the fluorescence intensity after labeling. In one embodiment intracellular signaling markers include without limitation Nfkb (pSer529), ERK1 / 2 (Thr202 / Tyr204), SHP-2 (pTyr542), pAkt (pSer473), STAT5 (pY694), PLCγ1 (pY783), PLCγ2 (pY759), pRPS6 (Ser235 / 236), CD247 (pY142), STAT3 (pY705), PTEN, TCF1, NUR77 and Ki67 An "immune cell" is any immune cell of the immune system of a subject such as a human subject. An "immune effector cell" is any cell of the immune system that has one or more effector functions (e.g., cytotoxic cell killing activity, secretion of cytokines, induction of antibody- dependent cellular cytotoxicity (ADCC) and / or complement-depending cytotoxicity (CDC)). Immune cells, such as T cells and NK cells, can be autologous / autogeneic ("self) or non- autologous ("non-self," e.g., allogeneic, syngeneic or xenogeneic). "Autologous", refers to cells from the same subject. "Allogeneic", refers to cells of the same species that differ genetically to the cell in comparison. "Syngeneic", as used herein, refers to cells of a different subject that are genetically identical to the cell in comparison. "Xenogeneic", as used herein, refers to cells of a different species to the cell in comparison. The term ‘leukocytes’ refers to white blood cells, which are cells of the immune system of a subject (immune cells). Leukocytes originate from hematopoietic stem cells in the bone marrow and circulate through the body within both the blood and the lymphatic system. Leukocytes preferably comprise T Cells, B Cells, natural killer (NK) cells, dendritic cells, granulocytes (basophils, eosinophils, and neutrophils), innate lymphoid cells (ILCs), megakaryocytes, monocytes / macrophages, and / or thymocytes. Lymphocytes are a type of leukocytes, comprising T cells, B cells and Innate lymphoid cells (ILCs). T cells are a type of lymphoid cells (lymphocytes), more precisely a type of white blood cells (leukocytes). T cells play an essential role in the adaptive immune response and the immune system. T cells may be characterized by the presence of a T cell receptor (TCR) on their cell surface. Specific, differentiated T cell subtypes have various functions in the control and modification of the immune response. An important function is the initiation of immune-mediated cell death, which is carried out by two major subtypes: CD8+ killer T cells and CD4+ helper T cells. The CD4+ (CD4 positive / expressing) population of T cells function as helper T cells (TH). CD4+ TH cells function by further activating memory B cells and cytotoxic T cells, leading to an increased immune response. CD8+ T cells are also termed killer T cells and are cytotoxic. Cytotoxicity in T cells refers to the ability to kill, for example, pathogen (e.g., virus)-infected cells or cancer cells. In addition, CD8+ T cells secrete cytokines (signalling molecules), to recruit other cell types to modulate the immune response. Another specific population of T cells provide the important mechanism of immune tolerance, wherein immune cells are competent in distinguishing between invading cells and the own cells of a subject. This ability of the so termed ‘suppressor T cells’ is important as it prevents immune cells from falsely attacking a body’s own cells, which is also known as autoimmune reaction. T cells may be further characterized into different differentiation stages by their gene, protein and / or surface (marker) protein expression. T cell ‘differentiation states’ may be defined as early differentiated cellular states that are endowed with a high self-renewing capacity. Representative early differentiated T cell subsets may be defined by T cell memory markers as CCR7+CD45RA+ naïve-like T cells and CCR7+CD45RA- central memory T cells (TCM). In contrast, late differentiated cellular states are considered to execute immediate T cell effector function. Representative T cell subsets may be CCR7-CD45RA- effector memory T cells (TEM), and CCR7-CD45RA+ terminally differentiated effector-memory T cells (TEMRA). Natural killer cells, also known as NK cells or large granular lymphocytes (LGL), are a type of cytotoxic lymphocyte critical to the innate immune system that belong to the rapidly expanding family of known innate lymphoid cells (ILC) and represent 5–20% of all circulating lymphocytes in humans. The role of NK cells is analogous to that of cytotoxic T cells in the vertebrate adaptive immune response. NK cells provide rapid responses to virus-infected cell and other intracellular pathogens acting at around 3 days after infection, and respond to tumor formation. Typically, immune cells detect the major histocompatibility complex (MHC) presented on infected cell surfaces, triggering cytokine release, causing the death of the infected cell by lysis or apoptosis. NK cells are unique, however, as they have the ability to recognize and kill stressed cells in the absence of antibodies and MHC, allowing for a much faster immune reaction. NK cells can be identified by the presence of CD56 and the absence of CD3 (CD56+, CD3−). The method described herein can be used to study intracellular signaling in response to cellular interactions. In one embodiment, the method comprises a high-plex cytometry panel. In one embodiment, the high-plex cytometry panel comprises an antibody to detect the phosphorylation (pY142) of the intracellular CD3 zeta protein (CD247). This protein acts as a transmembrane signalling adaptor and is phosphorylated during T cell receptor signalling and T cell activation. This panel was used to investigate intracellular TCR signalling in cellular interactions induced in human PBMCs by CytoStim (cross-linked antigen-presenting cells with T cells) and Blinatumomab (cross-linked B cells with T cells). In one embodiment, specific induction of B - T cell interactions upon Blinatumomab treatment and broader myeloid and B cell interactions with T cells upon CytoStim treatment can be observed. These specific interactions were confirmed by the observed phosphorylation of the CD3 zeta domain in the T cells, demonstrating functional T cell receptor activation in the interacting T cells. In one embodiment, CytoStim-induced interactions led to a strong phosphorylation of the intracellular CD3 zeta domain in both T*B and T* myeloid interactions as well as in T*B* myeloid triplets. For example, blinatumomab caused a specific increase in phosphorylation of the intracellular CD3 zeta domain in T cells involved in interactions with B cells, but to a much lesser extent in interactions not involving B cells, as shown in the examples. The use of a highly complex cytometry panel is particularly advantageous for the detailed investigation of T cell receptor signalling and their activation in specific cellular interactions. This leads to an improved analysis of the functionality and dynamics of immune responses in different experimental conditions. The specific induction of cell interactions by CytoStim and Blinatumomab and targeted activation of immune cells can be determined particularly well. This is particularly advantageous for therapeutic control of the immune response and analysis of binding properties of candidates for bi-specific molecules. The detection of specific phosphorylation events provides valuable insights into the molecular mechanisms of the immune response and enables the development of more precise and targeted immunotherapeutic approaches. In vivo derived interactions: In one embodiment, ImageStream-based flow cytometry is used to distinguish between single and interacting cells. In one embodiment, congenic mouse models are used. The congenic mouse models are also used to show that newly acquired interactions during sample preparation are not random but biologically relevant and representative of in vivo conditions. In one embodiment, the method of the present invention combines morphological imaging and fluorescence intensity data. This allows the type of cell interactions to be identified particularly well and their changes during an infection to be assessed. In one embodiment, various cellular interactions during the processing of samples from in vivo and ex vivo environments are analysed, with specific interactions being recorded during sample preparation. By additionally using ImageStream-based flow cytometry, a precise distinction between single and interacting cells can be made, increasing the accuracy of the analyses. The ,method of the present invention enables a comprehensive characterisation of cellular interactions by using both morphological and fluorescence intensity data, leading to a detailed and reliable assessment of infection processes. Examination of sample processing shows that newly acquired interactions during sample preparation are biologically relevant and reflect actual in vivo conditions, strengthening the validity of research results. Finally, the congenic mouse models used in the example below prove that the observed interactions are not random but represent targeted biological effects, which underpins the validity and applicability of the method. Application to existing cytometry datasets: In one embodiment, the automated clustering algorithm of the method of the present invention is applied to cytometric datasets such as publicly available datasets. This is demonstrated in the examples below for a detailed analysis of cell interactions in juvenile idiopathic arthritis (JIA). In one embodiment, FCS files of the publicly available dataset are downloaded and preprocessed as described in the examples below. In one embodiment, a FlowAI QC algorithm is ran on the FCS files to exclude those with anomalous flow rates from further analysis and wherein said processed FCS files are then further processed using the method of the present invention. By downloading and pre-processing the FCS files and excluding anomalies using the FlowAI QC algorithm, a highly accurate data basis is created. In one embodiment, the automated clustering algorithm of the method of the present invention is used to quantify differences in cell interaction between blood samples from patients with active and inactive JIA and between blood and synovial fluid samples. This enables the detection of qualitative and quantitative changes in cell interactions. In one embodiment, T cells in inactive JIA patients exhibit a regulatory FoxP3-expressing T cell phenotype, whereas an inflammatory, non-regulatory phenotype predominates in active JIA patients. In one embodiment, the automated clustering algorithm of the method of the present invention is used to evaluate significant qualitative differences in the interactions between CD4 T cells and monocytes, particularly between blood and synovial fluid samples from patients with active JIA. Using the automated clustering algorithm of the method of the present invention to analyse previously generated cytometric data enables highly precise and detailed analysis of cell interactions, especially from publicly available datasets, which is particularly valuable for understanding autoimmune diseases such as JIA. The automated clustering algorithm can thus advantageously also reveal qualitative and quantitative changes in the interactions of immune cells, which can lead to new insights into disease progression. Thus, the method of the present invention provides a basis for the identification of target structures for therapeutic interventions by visualising changes in the phenotype of T cells and their interactions with other cells. Finally, the method reveals significant differences in cell interactions between different samples, which contributes to a better understanding of the pathophysiological processes in JIA. According to the examples, the method, for example, can be particularly advantageous because patients do not have to be asked to provide a sample again and thus do not have to be burdened, and money and resources are saved because repeated measurements are not necessary or only to a small extent. Flow cytometry: In general, ‘flow cytometry’ is a technique used to analyze and quantify characteristics of individual cells and / or particles. Flow cytometry enables the analysis of cells, particularly suspensions / mixtures comprising different types of cells, on a single cell-basis by using a laser beam that scans / measures each cell regarding various properties of interest, such as cell (or particle) size, shape, granularity, and / or fluorescence by simultaneously detecting / analyzing laser light scattering (“scattering properties” of the cells) and / or fluorescence (e.g., induced by laser irradiation of a respective dye). According to the present invention the term “flow cytometry data set” refers to data obtained for a sample measured by flow cytometry. Flow cytometric measurement is used for e.g. cell counting, cell sorting, determining cell characteristics and function, detecting microorganisms, detecting a biomarker, diagnosing disorders such as blood cancer. In this process, a sample containing cells or particles is suspended in a fluid and injected into a flow cytometer instrument. The sample is focused to ideally flow one cell at a time through a laser beam, where the light scattered is characteristic to the cells and their components. Cells are often labeled with fluorescent markers, so light is absorbed and then emitted in a band of wavelengths. Tens of thousands of cells can be quickly examined, and the data gathered are processed by a computer. Modern flow cytometers allow to analyze many thousands of particles per second, in "real time" and, if configured as cell sorters, can actively separate and isolate particles with specified optical properties at similar rates. Commonly, flow cytometry instruments produce scattered and fluorescent light signals that are read by detectors. In some embodiments dyes, such as fluorescein, Alexa Fluor 488, Texas Red (325 D), Alexa Fluor 647 (1464 D), Pacific Blue and Cy5 (762 D) may commonly be used for conjugation to a binding reagent such as an antibody binding to a cell surface marker of a cell of interest. Samples are commonly prepared for fluorescence measurement by transfection and / or expression of fluorescent proteins (e.g., GFP), staining with fluorescent dyes (e.g., Propidium Iodide (PI), phycoerythrin (PE), allophycocyanin (APC) and peridinin chlorophyll protein (PerCP)) and / or staining with fluorophore-conjugated binding reagents such as antibodies (e.g., anti-CD4 FITC (FITC = fluorescein isothiocyanate)). In some embodiments the cells are stained by using fluorophore-conjugated binding reagents binding to a cell-type specific surface marker expressed by the cells. Cell proliferation can be determined by flow cytometry, e.g., by providing cells with BrdU (bromodeoxyuridine) and subsequent anti-BrdU antibody and a DNA dye staining. In flow cytometry, specific cell populations / types and / or cell states may be distinguished and optionally sorted or enriched according to gene expression and / or surface availability of certain proteins (‘markers’). Most immune cells express characteristic CD proteins (cell type specific markers) that define them as a cell population. Such markers are termed “lineage markers” or “cell type specific markers”, e.g., T cell lineage markers (CD3, CD4, CD8), and can be employed to characterize, isolate, count and / or enrich specific cell populations from a sample. During flow cytometry single cells (also termed “singlets”) and interacting cells such as doublets (comprising two interacting cells), triplets (comprising three interacting cells) and higher order multiplets (comprising more than three interacting cells) may be characterized and sorted / analyzed not only depending on (surface and / or intracellular) marker expression, but also by their scattering properties. The term “scattering properties” refers to parameters that are measured when cells or particles pass through a focused light beam (typically a laser) in a flow cytometer and scatter the light of the beam. Scattering properties according to the present invention include without limitation forward scatter (FSC) and side scatter (SSC). FSC is a measure of light scattering at a small angle (below 90 degrees) relative to the direction of the incident light beam, providing information on the size and volume of the cells or particles passing the beam. SSC is a measure of light scattering at a 90-degree angle or at an about 90-degree angle to the direction of the light beam, providing information on the granularity and internal structure of the cells or particles. The FSC or SSC signal can be processed as height (H), area (A) or width (W) of the signal detected. The height (H) of the FSC signal (FSC-H) or SSC signal (SSC-H) refers to peak maximum of the signal when the cell passes the laser beam. The width (W) of the FSC signal (FSC-W) or SCC signal (SSC-W) represents the time duration of the signal when the cell passes the laser beam. The area (A) of the FSC signal (FSC-A) or the SCC signal (SSC-A) refers to the area under the curve of the FSC or SSC signal detected and is calculated based on the height (H) measurements during the time (W) the cell passes through the laser beam. During a flow cytometry measurement, cells with markers and / or characteristics of interest are usually selected in real time by selecting a region (gate) around a (in the analysis software) shown population of cells (gating) and applying that region to other parameters within the experiment. Different parameters can be applied to other conditions, such that cells may be characterized and / or sorted according to multiple characteristics. Examples of how to perform flow cytometry sorting may be found in standard literature and scientific review articles and the skilled person is familiar with standard methods and workflows of flow cytometry and cell sorting, such as FACS. A flow cytometer typically has five main components: a flow cell, a measuring system, a detector, an amplification system, and a computer for analysis of the signals. The flow cell has a liquid stream (sheath fluid), which carries and aligns the cells so that they pass single file through the light beam for sensing. The measuring system commonly uses measurement of impedance (or conductivity) and optical systems - lamps (mercury, xenon); highpower water-cooled lasers (argon, krypton, dye laser); low-power air-cooled lasers (argon (488 nm), red-HeNe (633 nm), green-HeNe, HeCd (UV)); diode lasers (blue, green, red, violet) resulting in light signals. The detector and analog-to-digital conversion (ADC) system converts analog measurements of forward-scattered light (FSC) and side-scattered light (SSC) as well as dye-specific fluorescence signals into digital signals that can be processed by a computer. “Spectral Flow Cytometry” is an advanced form of flow cytometry that utilizes a full spectrum of fluorescence emission to analyze multiple fluorescent markers simultaneously. Unlike “conventional flow cytometry”, which detects signals from discrete optical filters, and which is thus limited to a low number of labels that can simultaneously be detected, spectral flow cytometry captures the entire fluorescence emission spectrum for each particle or cell. This technique involves the use of a spectrograph and an array of detectors, allowing for the deconvolution of overlapping fluorescence signals from multiple dyes. As a result, spectral flow cytometry enables to use a higher number of labels and thus higher-dimensional, more detailed and accurate analysis of cells or particles. As used herein, the term “sample” is a biological sample that is obtained or isolated from the patient or subject. A sample may, without limitation refer to a biopsy sample, a sample of body fluid or tissue obtained for the purpose of diagnosis, prognosis, or evaluation of a subject of interest, such as a patient. In some embodiments, the sample is a tissue sample, a tissue biopsy, a sample of a body fluid, such as blood, serum, plasma, cerebrospinal fluid, synovial fluid, saliva, urine, pleural effusions, cells, a cellular extract, and the like. Particularly, the sample is blood, or blood plasma, blood serum or a PBMC sample. The term “labeling” or “staining” shall mean delivering a binding agent to a cell, for example on a cell surface or into the cell of a microorganism. A labeled sample may refer to a sample in which one or more phenotypic parameters of the cells are labeled with a suitable label, or binding agent. In some embodiments, a “binding agent” that specifically recognizes a surface marker expressed by a cell may be selected from, without limitation, an antibody, protein, peptide, nucleic acid, or other small molecule that specifically binds a surface marker or any fragment(s) thereof to identify, track or capture its target molecule. As used herein, an "antibody" generally refers to a protein consisting of one or more polypeptides substantially encoded by immunoglobulin genes or fragments of immunoglobulin genes. Where the term “antibody” is used, the term “antibody fragment” may also be considered to be referred to. The recognized immunoglobulin genes include the kappa, lambda, alpha, gamma, delta, epsilon and mu constant region genes, as well as the myriad immunoglobulin variable region genes. Light chains are classified as either kappa or lambda. Heavy chains are classified as gamma, mu, alpha, delta, or epsilon, which in turn define the immunoglobulin classes, IgG, IgM, IgA, IgD, and IgE, respectively. The basic immunoglobulin (antibody) structural unit is known to comprise a tetramer or dimer. Each tetramer is composed of two identical pairs of polypeptide chains, each pair having one "light" (L) (about 25 kD) and one "heavy" (H) chain (about 50-70 kD). The N-terminus of each chain defines a variable region of about 100 to 110 or more amino acids, primarily responsible for antigen recognition. The terms "variable light chain" and "variable heavy chain" refer to these variable regions of the light and heavy chains, respectively. Optionally, the antibody or the immunological portion of the antibody, can be chemically conjugated to, or expressed as, a fusion protein with other proteins. "Single-chain Fv" or "scFv" antibody fragments comprise the VH and VL domains of an antibody, wherein these domains are present in a single polypeptide chain and in either orientation (e.g., VL-VH or VH-VL). Generally, the scFv polypeptide further comprises a polypeptide linker between the VH and VL domains which enables the scFv to form the desired structure for antigen binding. In preferred embodiments, a CAR contemplated herein comprises an antigen-specific binding domain that is a scFv and may be a murine, human or humanized scFv. Single chain antibodies may be cloned from the V region genes of a hybridoma specific for a desired target. "Single-chain Fv" or "scFv" antibody fragments comprise the VH and VL domains of an antibody, wherein these domains are present in a single polypeptide chain and in either orientation {e.g., VL- VH or VH-VL). Generally, the scFv polypeptide further comprises a polypeptide linker between the VH and VL domains which enables the scFv to form the desired structure for antigen binding. A label capable of detection in a flow cytometer such as a conventional flow cytometer or a spectral flow cytometer may be elected by a skilled person without undue effort. In some embodiments, the label is a fluorescent label. In embodiments, the fluorescent label is selected from the group consisting of cyanines, Janelia Fluors or rhodamines. Non-limiting examples of cyanines are Cy3 (Cyanine-3), Cy5 (Cyanine-5) and Cy7 (Cyanine-7). Non-limiting examples of rhodamines are TMR, SiR, TMR12 and SiR-d12. Non-limiting examples of Janelia Fluors are Janelia Fluor549-NHS Ester, Janelia Fluor646-NHS Ester, Janelia Fluor585-NHS Ester, Janelia Fluor635-NHS-Ester and Janelia Fluor669-NHS Ester. Further examples of fluorescent labels include, without being limited to, rhodamine and derivatives, lissamine, fluorescein, 5- bromomethylfluorescein and derivatives, DAPI, Hoechst 33258, R-phycocyanin, B-phycoerythrin, R-phycoerythrin, Lucifer Yellow, IAEDANS, 7-Me2N-coumarin-4-acetate, 7-OH-4-CH3-coumarin- 3-acetate, monobromobiman, Pyrene trisulfonates such as Cascade Blue and monobromotrimethyl ammoniobiman, Texas Red, Rhodamine Green, Oregon Green 30488, Oregon Green 514, 7-NH2-4CH3-25-coumarin-3-acetate (AMCA), FAM, TET, CAL Fluor Gold 540, JOE, VIC, Quasar 570, CAL Fluor Orange 560, NED, Oyster 556, TMR, CAL Fluor Red 590, HEX, ROX, LC Red 610, CAL Fluor Red 610, LC Red 610, CAL Fluor Red 610, LC Red 640, CAL Fluor Red 635, LC Red 670, Quasar 670, Oyster 645, LC Red 705, BODIPY FL, Cal Gold, BODIPY R6Gj, Yakima Yellow, Cal Orange, BODIPY TMR-X, JOE, HEX, Quasar-570, TAMRA, Rhodamine Red-X, Redmond Red, BODIPY 581 / 591, Cy3.5, Cy5, Cy5.5, Cal Red / Texas Red, BODIPY 630 / 665-X, BODIPY TR-X , Quasar-670 / Cy5, Pulsar-650, Dy590, Dy490, Dy636, Dy682, Atto-488, Atto-532, Atto-Rho-6G, Atto-Rho101, Atto-647N, Atto-680, BMN-488, BMN-505, BMN-536, BMN-562. The term “gating” refers to a basic conventional principle of flow cytometric analysis, which is the sequential identification and refinement of a cellular population of interest using one or more surface markers that are visualized, e.g., by fluorescence in a unique emission spectrum. The process of gating in flow cytometry typically refers to manually selecting an area on the scatter plot generated during the flow experiment that determines which cells are used for further analysis and which cells are not. That means, that cells that are not of interest will be “gated out”. Clustering algorithm and computer implementation The term “automated clustering algorithm” refers to an algorithm assigning items, events or other defined elements or objects (such as the cells of a cell population, or a digital representation of said cells) to a cluster or group according to a degree of similarity of each item (for example, based on the scattering properties of the cells and the expression of cell type-specific surface markers). An automated clustering algorithm avoids or reduces manual user input (such as manual gating), for example by performing an (unbiased) algorithm comprising a connectivity model, a centroid model, a distribution model, a density model, a graph-based model, a self- organizing model or a neural model. In one embodiment, an automated clustering algorithm excludes manual user input. In one embodiment, an automated clustering algorithm excludes manual gating. In some embodiments, the automated clustering algorithm is unbiased, for example no predefined subset of the data or items in the data set is included or other items excluded, such as obtained through manual gating. Rather, in embodiments, all items or data from the data set are analyzed and clustered without manual intervention by a user. The term “biased” refers to an analysis or dataset in which systematic errors or distortions are present that favor or disadvantage certain outcomes, that are not representative of the entire population, in which certain genes or sequences are preferentially selected, or in which measurement instruments or methods systematically give incorrect values. An analysis or data is considered "unbiased" if no or only negligible systematic errors or distortions are present and all data points are considered equally upon analysis or within the dataset. The term “clustering algorithm” refers to a computer implemented method for grouping a plurality of items, events or other defined elements or objects of a data set (“data points”) (such as cells) into clusters (also termed “groups”) based on a degree of similarity of one or more characteristics of the items (such as scattering properties and expression of cell specific surface markers). Thereby, the clusters are formed in such a way that objects within a cluster are similar, preferably as similar as possible, concerning certain characteristics (high degree of similarity), while objects from different clusters are dissimilar, preferably as dissimilar as possible (low degree of similarity). Clustering algorithms typically use mathematical and statistical models to measure the similarity or distance, also termed “degree of similarity”, between data points concerning certain characteristics and use this information to organize the data points into clusters. By way of example, a “degree of similarity” refers to a distance metric calculated between the data points of a dataset based on certain characteristics of the data points. In some embodiments, the degree of similarity of the scattering properties and expression of cell-type specific surface markers and / or combinations thereof is calculated, and the cells are assigned to a cluster based on the calculated degree of similarity. “Degree of similarity” distance metrics commonly used for clustering algorithms and well known to a person skilled in the art include without limitation Euclidean distance, Manhattan distance and correlation distance. Typical models used for clustering methods are known to a person skilled in the art and include, without limitation, connectivity models (such as Louvain clustering and Leiden clustering), centroid models (such as K-means clustering), distribution models, density models, graph-based models (such as Phenograph clustering), self-organizing maps (SOM, such as FlowSOM clustering), hierarchical clustering and neural models. Louvain and Leiden clustering are two algorithms for detecting communities (community detection) in large and complex networks, and often used for gene clustering. “Louvain clustering” refers to a clustering algorithm that finds non-overlapping communities in large networks by optimizing modularity, a measure of the quality of network partitioning. The Louvain clustering can be broken into two phases: (i) the algorithm tries first to maximize the modularity of the graph by moving nodes between communities (maximization of modularity) and (ii) the algorithm aggregates second the nodes in the same community into a single node and creates a new graph (aggregation). The process is repeated on the simplified network. “Leiden clustering” refers to a clustering algorithm that improves the Louvain method by ensuring that all communities are well connected and that modularity is further optimized. Leiden consists of three phases: (i) local displacement of nodes, (ii) refinement of the partitioning and (iii) aggregation of the network based on the refined partitioning. In contrast to Louvain, Leiden only visits nodes whose neighborhood has changed, which makes the algorithm more efficient and faster. By way of example, two different types of clustering are known in the art: “hard clustering” and “soft clustering”. “Hard clustering” refers to assigning an object (data point) such as a cell entirely to one cluster. “Soft clustering” refers to assigning a probability score to an object (data point) with regards to belonging to a cluster. In some embodiments, the automated clustering algorithm is a hard clustering algorithm. In some embodiments, the automated clustering algorithm is a soft clustering algorithm. Automated clustering methods such as clustering algorithms may further be grouped into unsupervised, semi-supervised and supervised clustering algorithms (Weber et al., Comparison of clustering methods for high-dimensional single-cell flow and mass cytometry data, Cytometry part A, 2016; Cai et al., A review on semi-supervised clustering, Information Sciences, 2023). „An “unsupervised” analysis, model, or algorithm is used for a dataset without predefined labels or output data, particularly to discover patterns or structures within the data; such as for grouping protein expression or gene expression data into clusters to identify proteins or genes with similar expression patterns. Unsupervised clustering algorithms use clustering methods to detect cell populations such as single cells or interacting cells, defined herein as groups (“clusters”) of cells or interacting cells with scattering properties and marker expression of a (in some cases predefined) degree of similarity. Clustering analysis may be performed on data from a single biological sample, on data from multiple samples on a per-sample basis, or on combined data from multiple samples. The assigned clusters (cell populations) can then be analyzed individually or compared across samples, for example by comparing cluster frequencies between samples in different biological conditions such as prior to and after medical treatment. Importantly, this procedure allows previously unknown cell populations such as a cluster comprising interacting cells to be identified in an unbiased, data-driven manner; this type of exploratory analysis is difficult or impossible with manual gating, especially when using high-dimensional data (Weber et al., Comparison of clustering methods for high-dimensional single-cell flow and mass cytometry data, Cytometry Part A, 2016). By contrast, supervised approaches rely on an external biological or clinical variable describing each sample. This could be a simple categorical variable such as disease status or tissue type, or a more complex clinical outcome such as survival time. Supervised approaches use this external variable as an input to train a model, which can then be used to predict the status of new samples (Weber et al., Comparison of clustering methods for high-dimensional single-cell flow and mass cytometry data, Cytometry part A, 2016). “Supervised” thus refers to an analysis, model or algorithm trained using a training dataset including both input data and corresponding output data (labels), particularly aiming to create a model that can accurately classify or predict new, unseen data. Semi-supervised approaches typically rely on both labeled and unlabeled data, for example incorporating given prior information into the clustering algorithm to guide the clustering process and improve the performance (Cai et al., A review on semi-supervised clustering, Information Sciences, 2023; AlZuhair et al., Soft Semi-Supervised Deep Learning-Based Clustering, Applied Sciences, 2023). As used herein, the term “neural network” refers to a neural network as understood by a skilled person. For example, a neural network is a network or circuit of neurons in an artificial neural network, composed of artificial neurons or nodes, connected by edges. Artificial neural networks (ANNs), also named neural networks (NNs), are based on a collection of connected units or nodes called artificial neurons, which loosely model the neurons in a biological brain. An artificial neural network preferably consists of a collection of simulated neurons. By way of example, each neuron is a node which is connected to other nodes via links that correspond loosely to biological axon-synapse-dendrite connections. Each link has a weight, which determines the strength of one node's influence on another. The neurons are typically organized into multiple layers, especially in deep learning. Neurons of one layer connect only to neurons of the immediately preceding and immediately following layers. The layer that receives external data is the input layer. The layer that produces the ultimate result is the output layer. In between them may be zero or more hidden layers. Single layer and unlayered networks may also be used. Example neural networks are known to a skilled person. As used herein, “training” a neural network refers to the term as commonly understood by a skilled person. By way of example, neural networks learn (or are trained) by processing examples, each of which contains a known "input" and "output," forming probability-weighted associations between the two, which are stored within the data structure of the net itself. The training of a neural network from a given example is usually conducted by determining the difference between the processed output of the network (often a prediction) and a target output. This difference is the error. The network then adjusts its weighted associations according to a learning rule and using this error value. Successive adjustments will cause the neural network to produce output which is increasingly similar to the target output. After a sufficient number of these adjustments the training can be terminated based upon certain criteria. This may also be referred to as supervised learning. Suitable training modes and software are known to a skilled person. The term “thresholding” or “thresholding method” refers to a method for digital image processing used to convert an image into two or more classes of pixels, which are typically termed “foreground” and “background”. This is achieved by applying a threshold value to the pixel intensity levels of the image. Thresholding methods can be classified into “manual thresholding” and “automated thresholding”. “Manual thresholding” refers to manual selection of a threshold by visual inspection of the image or histogram of the image by a user. An “automated thresholding” avoids or reduces manual user input (such as manual threshold selection by visual inspection). Automated thresholding methods can further be classified into global or local thresholding methods. “Global thresholding” refers to applying a single threshold to every pixel of the image or histogram of the image, such that that every pixel less than that value is considered one class, while every pixel greater than that value is considered the other class. Global thresholding methods are well known to a person skilled in the art and include without limitation Otsu thresholding, the minimum method, the triangle method, the mean method, the median and median absolute deviation (MAD) method. “Local thresholding” refers to applying different threshold values to different pixels of the image or histogram of the image. Local thresholding may be useful for images with varying lighting conditions or non-uniform backgrounds. Methods for local thresholding are well known to a person skilled in the art include without limitation the Niblack method, mean or median-based methods, wherein the threshold for a pixel is determined based on the mean or median intensity of its neighborhood and Gaussian adaptive thresholding, wherein a weighted sum (Gaussian distribution) of the neighborhood intensities is used to determine the threshold. In one embodiment the thresholding method is an automated thresholding method. In one embodiment the automated thresholding method is a global thresholding method. In one embodiment the thresholding method is selected from the group consisting of Gaussian Mixture Model thresholding, Huang thresholding, Intermodes thresholding, IsoData thresholding, kmeans thresholding, Li thresholding, manual gating, mean thresholding, Otsu thresholding, Renyi Entropy thresholding, Shanbhag thresholding and / or triangle thresholding. The term “Otsu thresholding” refers to an automated image processing method used for automatic threshold selection to distinguish between foreground and background in a grayscale image. It is a type of global thresholding method that minimizes the intra-class variance of pixel values in the resulting classes (foreground and background). Developed by Nobuyuki Otsu (Otsu et al., 1979), this technique utilizes the histogram data of an image to determine an optimal threshold that minimizes the sum of variances within the two resulting pixel groups. This results in a clear separation between the object and the background in the image, which is particularly important for image segmentation and analysis. Otsu thresholding can be considered as automated, global thresholding method. Particular aspects of the present invention may be computer-implemented. Accordingly in some embodiments the present method may be a computer-implemented method. The person skilled in the art is aware of which aspects and features of the present invention may be computer- implemented. In a further aspect the present invention relates to a computer-readable storage device, comprising a software or computer program product suitable to carry out the invention. In a further aspect the present invention relates to a computer-readable storage medium having stored thereon the software or computer program product according to the invention, or suitable to conduct the method or other aspects of the invention. The method, system or computer product of the invention may in some embodiments comprise one or more conventional computing devices having a processor, an input device such as a keyboard or mouse, memory such as a hard drive and volatile or nonvolatile memory, and computer code (software) for the functioning of the invention. The computers may also comprise a programmable printed circuit board, microcontroller, or other device for receiving and processing data signals such as those received from the local controllers, programmable manufacturing equipment, programmable material handling equipment, and robotic manipulators. The method, system or computer product may comprise one or more conventional computing devices that are pre-loaded with the required computer code or software, or it may comprise custom-designed hardware. The system may comprise multiple computing devices which perform the steps of the invention. In certain embodiments, a plurality of clients such as desktop, laptop, or tablet computers can be connected to a server such that, for example, multiple users can enter their orders for personalized products at the same time. The computer system may also be networked with other computers over a local area network (LAN) connection or via an Internet connection. The system may also comprise a backup system which retains a copy of the data obtained by the invention. The data connections of step d) may be conducted or configured via any suitable means for data transmission, such as over a local area network (LAN) connection or via an Internet connection, either wired or wireless. A client or user computer can have its own processor, input means such as a keyboard, mouse, or touchscreen, and memory, or it may be a dumb terminal which does not have its own independent processing capabilities, but relies on the computational resources of another computer, such as a server, to which it is connected or networked. Depending on the particular implementation of the invention, a client system can contain the necessary computer code to assume control of the system if such a need arises. In one embodiment, the client system is a tablet or laptop. The components of the computer system may be conventional, although the system may be custom-configured for each particular implementation. The computer system may run on any particular architecture, for example, personal / microcomputer, minicomputer, or mainframe systems. Exemplary operating systems include Apple Mac OS X and iOS, Microsoft Windows, and UNIX / Linux; SPARC, POWER and Itanium-based systems; and z / Architecture. The computer code to perform the invention may be written in any programming language or model-based development environment, such as but not limited to C / C++, C#, Objective-C, Java, Basic / VisualBasic, MATLAB, Simulink, StateFlow, Lab View, or assembler. The computer code may comprise subroutines which are written in a proprietary computer language which is specific to the manufacturer of a circuit board, controller, or other computer hardware component used in conjunction with the invention. The method, system or computer product, and database of the invention, can employ any kind of file format which is used in the field. For example, a digital representation of a protein sequence, protein structure, or other computer-implemented aspect of the invention, can be stored in a proprietary format, DXF format, XML format, or other format for use by the invention. Any suitable computer readable medium may be utilized. The computer-usable or computer-readable medium may be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD- ROM), an optical storage device, a transmission media such as those supporting the Internet or an intranet, cloud storage or a magnetic storage device. Herein the term “graphical user interface” or “GUI” relates to a form of user interface that allows a user to interact with electronic devices through a graphical representation, graphical icons and / or audio-visual or haptic instead of text-based user interfaces, typed command labels or text navigation. Herein a GUI may be a system of interactive visual components for computer software that displays objects, instructions or data that convey information and represent actions that can be taken by a user. Treatment / Therapeutic application Herein the terms “subject”, ”patient” or “host” may be used interchangeably. A subject, patient or host may be an organism, a cell culture of patient cells or cell lines, an animal, or a cell culture of animal cells or cell lines. Herein a subject or patient refers preferably to a species from which a sample is taken, and / or whose biological material makes up the majority of the biological material of a sample. A subject or patient can be selected from the group comprising vertebrae, animals, mammals, humans, preferably mammal or human. As used herein "treatment" or "treating" includes any beneficial or desirable effect on the symptoms or pathology of a disease or pathological condition, and may include even minimal reductions in one or more measurable markers of the disease or condition being treated. Treatment can involve optionally either the reduction or amelioration of symptoms of the disease or condition, or the delaying of the progression of the disease or condition. "Treatment" does not necessarily indicate complete eradication or cure of the disease or condition, or associated symptoms thereof. Generally, the term ‘cellular immunotherapy’ or “therapeutic immune cell” may refer to a therapeutic approach utilizing immune cells, such as immune cells modified to express a transgenic construct such as a chimeric antigen receptor (CAR) or a T-cell receptor (TCR), e.g., CAR-T cells, CAR-NK cells and TCR-T cells, to target and eliminate target cells of interest, such as diseased or abnormal cells. In some embodiments this process may involve isolating, modifying (if necessary), and expanding specific immune cells outside the body, typically in a laboratory setting, before reintroducing them into the patient. Such manipulated immune cells may in embodiments be engineered to recognize and attack specific antigens present on target cells of interest, such as cancer cells, infected cells, or other harmful targets. In general, cellular immunotherapy may aim, e.g., to enhance an immune system's ability to recognize and destroy target cells of interest, such as abnormal cells, thereby providing a promising treatment strategy for various diseases, including cancer, autoimmune disorders, and infectious diseases. The term “immunodrug” or “immune modulator” refers to medications that modulate the immune system to treat or prevent diseases. Depending on treatment requirements, these medications can either enhance the immune response (immunostimulatory drugs), mediate immune response, mediate cellular interactions, or suppress it (immunosuppressive drugs). The term “responsiveness”, “therapy responsiveness” or “therapeutic success” refers to the patient’s response to a medical treatment such as an immunotherapy, in particular regarding a therapeutic effect of the medical treatment being achieved in the subject, for example by assessing or prognosing remission of a disease (e.g., a cancerous disease) after treatment with a cellular immunotherapy or a bispecific antibody. In some embodiments of the present invention the term “non responder” refers to a subject showing no remission after treatment with a medical treatment such as a cellular immunotherapy. In further embodiments the term “non responder” refers to a subject showing only partial remission after medical treatment. In some embodiments the term “non responder” refers to a subject showing minimal residual disease after medical treatment. In some embodiments the term “responder” refers to a subject showing remission after medical treatment such as after treatment with a cellular immunotherapy such as a CAR-T cell therapy. In some embodiments the term “responder” refers to a subject showing full remission after medical treatment. Immunotherapy in the context of the present invention is to be understood to comprise any therapeutic agent employing the immune system for treatment of an unwanted (pathogenic) condition, such as an immune reaction prior to and / or after allogenic transplantation, an autoimmune disease or a cancerous disease. Immunotherapy encompasses, without limitation, cellular and antibody therapy and antibody-drug conjugates (ADC). Cellular therapies according to the present invention involve without limitation the administration of genetically modified immune cells such as T cells and NK-cells, preferably modified to express a CAR or other targeting moiety, such as any other naturally occurring or synthetic construct providing antigen- specific targeting, including chimeric antigens (CARs) or T cell receptors (TCRs), and expanding said genetically modified immune cells. ‘Chimeric antigen receptors’ (CARs) or ‘chimeric immune receptors’ are receptor proteins constructed to provide T cells the novel ability to target a specific antigen. The receptors are chimeric, as they combine T cell-activating functions and antigen-binding in a single receptor. A CAR-T cell cancer therapy employs T cells engineered with chimeric antigen receptors. In a “CAR T cell” or “CAR-T” immunotherapy the T cells are modified to recognize and destroy cancer cells more effectively. CAR T cells may be derived from a patient's own blood (autologous) or from a donor (allogeneic). Isolated T cells may be genetically engineered to express a specific CAR, that enables the recognition of antigen specifically expressed by cancer cells and present on their surface. When CAR-T cells are introduced into a subject (e.g., cancer patient), they function similar to an anti-cancer drug. When a CAR T cell gets into contact with its target antigen on a (cancer) cells surface, the CAR-T cell binds to it and becomes activated, proliferates and induces cytotoxicity in the cancer cell. Cytotoxicity comprises mechanisms, such as an increased release of cell signalling factors that can affect other cells, e.g., cytokines, interleukins and growth factors. For example in a non-limiting example CAR T-cell therapy comprises the engineering of a patient’s T cells in the laboratory so they will, when administered back to the patient, bind to cancer cells and kill them. Therefore, leukocytes, including T cells, are extracted from the blood of a patient by an apheresis machine. Subsequently, a gene for a special receptor called a chimeric antigen receptor (CAR) is inserted in vitro into the isolated T cells. Therefrom, a plurality of the CAR T cells are grown in vitro and administered back to the patient (autologous) or to another recipient (allogeneic) by infusion. The CAR T cells are able to bind to the respective target antigen on the cancer cells of the recipient (e.g., a cancer patient) such that said cancer cells are killed (see e.g., US-NIH; www.cancer.gov). The generation of CAR T cells may be performed in embodiments as described in current literature, such as in Hong et al., 202046 and Hiltensperger et al., 202347. The T-cell receptor, or TCR, is a molecule typically found on the surface of T cells, or T lymphocytes that is responsible for recognizing fragments of antigen as peptides bound to major histocompatibility complex (MHC) molecules. The TCR is composed of two different protein chains. In humans, in 95% of T cells the TCR consists of an alpha (α) chain and a beta (β) chain (encoded by TRA and TRB, respectively), whereas in 5% of T cells the TCR consists of gamma and delta (γ / δ) chains (encoded by TRG and TRD, respectively). Each chain is composed of two extracellular domains: Variable (V) region and a Constant (C) region, both of Immunoglobulin superfamily (IgSF) domain forming antiparallel β-sheets. The Constant region is proximal to the cell membrane, followed by a transmembrane region and a short cytoplasmic tail, while the Variable region binds to the peptide / MHC complex. The variable domain of both the TCR α-chain and β-chain each have three hypervariable or complementarity determining regions (CDRs). There is also an additional area of hypervariability on the β-chain (HV4) that does not normally contact antigen and, therefore, is not considered a CDR. The residues in these variable domains are located in two regions of the TCR, at the interface of the α- and β-chains and in the β-chain framework region that is thought to be in proximity to the CD3 signal-transduction complex. CDR3 is the main CDR responsible for recognizing processed antigen, although CDR1 of the alpha chain has also been shown to interact with the N-terminal part of the antigenic peptide, whereas CDR1 of the β-chain interacts with the C-terminal part of the peptide. Recombinant TCRs have been previously transfected into therapeutic T cells intended for the treatment of proliferative disease. For example, TCR-T cells are engineered by transducing preferably autologous alpha-beta or gamma-delta cells with a retroviral or lentiviral vector encoding TCR (typically an alpha chain non-covalently bound with a beta chain) that recognizes peptides of interest and CD3z genes. When the engineered T cells recognize peptides bound to the major histocompatibility complex (MHC) on the surface of antigen-presenting or tumor cells, they become activated and start expanding. Thefirst TCR-T cell therapy was used in clinical trial for metastatic melanoma, whose TCR recognizing an HLA-A2-restricted peptide from a melanocytic differentiation antigen, melanoma antigen recognized by T cells 1 (MART-1). A ‘bi-specific antibody’ is usually an artificial protein that has two distinct binding domains that can each bind to an antigen or epitope. A bispecific antibody can simultaneously bind to two different types of antigens or two different epitopes on the same antigen. Examples of bispecific antibodies include without limitation bi-specific T-cell engager, immune mobilising monoclonal T-cell receptors against cancer, and several bi-specific antibody drug, which have been approved by authorities, i.e. the FDA, such as blinatumomab, emacizumab, amivantamab, tebentafusp, faricimab, teclistamab, mosunetuzumab, epcoritamab, and glofitamab. The term antibody-drug conjugates (ADC) refers to a class of biopharmaceutical drugs comprising an antibody and a therapeutic agent, preferably a cytotoxic therapeutic agent, which are covalently linked to each other by a linker. The antibody of the ADC specifically binds to the target antigen on the surface of a cell, such as an antigen specifically found on the surface of a tumor cell, thereby facilitating cellular internalization of the ACD by the cell. Within the cell the linker covalently binding the therapeutic agent to the antibody is cleaved releasing the therapeutic agent inside the cell and the released therapeutic agent exerts its therapeutic effect, such as a cytotoxic effect, inside the cell. Several ADC have been approved for medical use, including without limitation Gemtuzumab ozogamicin, Brentuximab vedotin, Trastuzumab emtansine, Inotuzumab ozogamicin, Polatuzumab vedotin, Enfortumab vedotin, Trastuzumab deruxtecan, Sacituzumab govitecan, Belantamab mafodotin, Moxetumomab pasudotox, Loncastuximab tesirine, Tisotumab vedotin-tftv and Mirvetuximab soravtansine. General remarks All words and terms used herein shall have the same meaning commonly given to them by the person skilled in the art unless the context indicates a different meaning. All terms used in the singular shall include the plural of that term and vice versa. lt will be understood that particular embodiments described herein are shown by way of illustration and not as limitations of the invention. The principal features of this invention can be employed in various embodiments without departing from the scope of the invention. Those skilled in the art will recognize or be able to ascertain, using routine study, numerous equivalents to the specific procedures described herein. Such equivalents are considered to be within the scope of this invention and are covered by the claims. All publications and patent applications mentioned in the specification indicate the skill level of those skilled in the art to which this invention pertains. All publications and patent applications are herein incorporated by reference to the same extent as if each individual publication or patent application was specifically and individually indicated to be incorporated by reference. The use of the word "a" or "an" when used in conjunction with the term "comprising" in the claims and / or the specification may mean "one," but it is also consistent with the meaning of "one or more," "at least one," and "one or more than one." The term "or" in the claims is used to mean "and / or" unless explicitly indicated to refer to alternatives only or the alternatives are mutually exclusive. However, the disclosure supports a definition of only alternatives and "and / or." Throughout this application, where relevant, the term "about" indicates that a value includes the inherent variation of error for the device, the method employed to determine the value or the variation among the study subjects. FIGURES The invention is demonstrated by way of the example through the figures disclosed herein. The figures provided represent particular, non-limiting embodiments and are not intended to limit the scope of the invention. Brief description of the figures Fig.1: A cytometry-based framework for the accurate identification of physical cellular interactions. Fig.2: Ultra-high scale cellular interaction mapping across complex immune landscapes. Fig.3: Cellular interaction mapping reveals mechanisms and kinetics of immunotherapies. Fig.4: Interact-omics framework of the present invention reveals features underlying therapy response to Blinatumomab. Fig.5: Virus-induced alterations of cellular landscapes and interaction networks. Fig.6: Cellular interaction dynamics underlying immune response to LCMV infection. Fig.7: Technical aspects of cytometry-based cellular interaction mapping. Fig.8: Interact-omics framework of the present invention resolves complex cellular interactions induced by CytoStimTM. Fig.9. Interact-omics framework of the present invention resolves antigen-specific interactions in the ovalbumin (OVA)-OT-II model. Fig.10: Interact-omics framework of the present invention maps cellular interaction dynamics of CAR-T cells. Fig.11: Cellular interaction dynamics upon Blinatumomab treatment. Fig.12: Interact-omics framework of the present invention identifies independent and additive features of Blinatumomab response Fig.13: Cluster frequencies of single and interacting cell landscapes of LCMV- infected mice. Fig.14: Interacting cell clusters used for annotation of LCMV-infected mice. Fig.15: Technical aspects of cytometry-based cellular interaction mapping. Fig.16: Effect of sample processing methods on ex vivo cellular interactions. Fig.17: Characterization of functional interactions by assessment of phosphorylated CD247 Fig.18: Comparison of the Interact-omics approach to imaging flow cytometry. Fig.19: Interactions acquired a posteriori after in vivo experiment. Fig.20: Interacting cell landscape in juvenile idiopathic arthritis (JIA). Fig.21: Interacting cell landscape in the proximal small intestine. Detailed description the figures Fig.1: A cytometry-based framework for the accurate identification of physical cellular interactions. (A) Schematic overview of the experimental approach and exemplary ground truth image data. PBMCs were incubated with the T cell cross-linker CytoStimTM, followed by the manual classification of 1000 living cells into singlets or multiples across four technical replicates. (B) Importance of features obtained from a decision tree model to classify the data into singlets and multiplets. X-axis labels of features from imaging flow cytometry are written in italic. n = 4 (C) Heatmap of most important features, colored by mean z-score of features across replicates grouped by multiplicity into singlet, doublet, triplet, higher-order multiplet. (D) UMAP embedding of classified cells (n = 3,865) based on all features from the heatmap in (C) and the expression of CD3, CD19, CD33, HLA-DR, and CD45. (E) Relative frequency of different multiplicities across the annotated clusters. (F) UMAP embedding of D with FSC-ratio high cells highlighted in blue (based on Otsu thresholding). (G) Relative frequency of cells classified according to the FSC- ratio. (H) Precision and recall of events assigned as singlets and multiplets based on FSC-ratio alone (green) compared to FSC-ratio and clustering (pink). Ground truth is the manual annotation. P values determined using the least-squares means (Lenth et al., 2022) method; correction for multiple testing according to Benjamini-Hochberg. Relative frequency of singlets and interacting cells based on the ground truth annotation. n = 4 replicates. (I) Adjusted rand index of clustering obtained for (i) all features used in (E) vs. (ii) only conventional cytometry features for different resolutions in Louvain clustering (Scrucca et al., 2023). (J) FSC ratio histogram, colored by the ground truth annotation. The classification into singlets and multiplets by Otsu thresholding is shown. (K) Performance of different classification methods as measured by the F1 score. In all methods displayed, cells were classified by Otsu thresholding of the FSC ratio. The first method (left, dark blue) relies on Otsu thresholding of the FSC ratio only, all others (second bar to fourth bar from the left, light blue) involve Louvain clustering based on different feature spaces as indicated below the x-axis, followed by assertion of clusters to either singlets or multiplets based on the proportion of cells exceeding the FSC ratio threshold. The third bar from the left represents the Interact-omics workflow. Louvain clustering was performed for n = 100 iterations, and the results for each replicate (n = 4) are shown in the point plot. Bars indicate the mean F1 score. X*: only the most important scatter parameters were used (see panel C). (L) Left: UMAP embedding of classified cells (n = 3,865) based on conventional flow cytometry parameters, including cell type markers, scatter parameters and the FSC ratio. Right: UMAP embeddings with cells exceeding the Otsu threshold of the FSC ratio highlighted in blue (top) or cells colored by their ground truth annotation (bottom). Abbreviations: UMAP = uniform manifold approximation and projection, FSC = forward scatter, SSC = side scatter, W = width, H = height, A = area, My = myeloid. Interactions between cell types are encoded by an asterisk between the two cell type labels. P values legend: * P < 0.05, ** P < 0.01. Fig.2: Ultra-high scale cellular interaction mapping across complex immune landscapes. (A) UMAP display of a 25-plex cytometry dataset of PBMCs cultured in the presence or absence of the cross-linking agent CytoStimTM, n = 4. Recorded cells were processed with PICtR, out of 226,301 cells, 50,000 sketched (Hao et al., 2023) cells are displayed. (B) UMAP of interacting cells (n = 9,988) (C) Heatmap colored by marker enrichment modeling (MEM) score (Diggins et al., 2017) of cell type defining markers across the clusters of cellular interactions. (D) Circos plots displaying the relative enrichment between T and antigen-presenting cells split by treatment conditions. (E) Box plots depicting log2 fold changes (FC) of normalized interactions between the CytoStimTMtreated and untreated conditions. Doublet counts were normalized by the harmonic mean of the singlet counts of the contributing cells to correct for background of randominteractions (for details see methods section “Normalization of physically interacting cells”). Pvalues were determined with a two-sided Wilcoxon rank sum test and adjusted for multiple testing according to Benjamini-Hochberg. n=4. (F) Schematic overview of the experimental setup of co- cultures of chicken ovalbumin (OVA)- specific OT-II CD4 T cells with murine splenocytes. (G) UMAP of the overall cellular landscape; n = 125,554 events. (H) UMAP of the interacting cell landscape; n = 6399. (I) Point density UMAP of (H) split into the treatment conditions. (J) Log2 fold changes of frequencies of interacting cells in the presence or absence of OVA. OVA_CD4+T*CD8+T interactions are not depicted, as they appeared exclusively upon OVA treatment. The p-values were calculated using least squared means (Lenth et al., 2022) and were Bonferroni-corrected. Error bars indicate mean and standard deviaton. n = 4. (K) Circos plot displaying the relative enrichment between T and antigen-presenting cells. Colors of the contributing singlets (highlighted on the circumference) are analogous to panel A. P value legend: ns not significant, * P < 0.05, **** P < 0.0001. Abbreviations: UMAP = uniform manifold approximation and projection, NK = natural killer, EoBaso = eosinophils-basophils, class = classical, non-class = non-classical, (p)DC = (plasmacytoid) dendritic cells, MEM = marker enrichment modeling, FC = fold change. Asterisks in cell type labels indicate interactions between the respective cell types. Fig.3: Cellular interaction mapping reveals mechanisms and kinetics of immunotherapies. (A) Schematic overview of the experimental setup of cultures from murine splenocytes and anti-CD19 CAR-T cells. (B) UMAP of the overall cellular landscape. Recorded cells were processed with PICtR; out of 849,845 cells, 96,988 sketched cells are displayed. (C) UMAP of the interacting cell landscape, n = 9,974. (D) Point density UMAP of (C) in the absence of CAR-T cells (top) and 1h after adding CAR-T cells (bottom) (E) Paired analysis of interactions between B cells and CAR-T cells or B cells and endogenous T cells normalized to the frequency of singlets available for the interaction. ** P < 0.01, *** P < 0.001, paired two-sided Welch’s t-test. (F) Schematic overview of the experimental setup for investigating cellular interactions upon treatment with Blinatumomab. (G) UMAP of the overall cellular landscape. Recorded cells were processed with PICtR; out of985,735 cells, 49,210 sketched cells are displayed. (H) UMAP of the interacting cell landscape, n= 34,362. I. Point density UMAP of H, in the absence of Blinatumomab (left) and 1h post Blinatumomab treatment (right). (J) Comparison of the B cell frequencies and frequencies of cellular interactions involving B cells over time. (K) Time-resolved Log2 fold change of certain cellular interaction frequencies normalized to the available singlets in the control. CAR = chimeric antigen receptor, NK = natural killer, cDC = conventional dendritic cells, pDC = plasmacytoid dendritic cells, Treg = regulatory B cells, gdT = gamma-delta T cells, prog = progenitors. Fig.4: Interact-omics framework of the present invention reveals features underlying therapy response to Blinatumomab. (A) Schematic overview of the experimental setup of ex vivo treated B-ALL bone marrow aspirates with Blinatumomab. (B) UMAP of the overall cellular landscape. Recorded cells were processed with PICtR; out of 4,292,770 cells, 70,000 sketched cells are displayed. Patient-specific leukemic clusters were merged into a common meta-cluster. (C) UMAP of interacting cells (n = 29,232). Point density UMAP of interacting cells split into the treatment conditions (middle). Bar graph illustrating the log2 mean fold change (FC) of cellular interactions (Blinatumomab treated vs. control) (right). P values were determined with a two- sided Welch’s t-test. Bonferroni-adjusted P values are displayed. (D) Volcano plot representing enrichment and depletion for good responders vs. non-responders for both singlets and cellular interactions. (E) Top panel: Box plot showing the comparison of the fold change of B*T(*My) interactions after ex vivo Blinatumomab treatment in good responders (GR) and non responders (NR). Bottom panel: Box plot showing the frequency of singlet CD8+ T cells in GR and NR. Interaction frequencies were adjusted for singlet frequencies of the contributing cells (harmonic mean, see Methods). P values were determined with a t-test. (F) Scatter plots displaying the fold change of B*T(*My) interactions upon Blinatumomab treatment against the singlet T cell / B cellratio at baseline. (G) Scatter plot displaying the fold change of B*T(*My) interactions against thefrequency of T*My interactions at baseline. (H) Heatmap of Pearson correlation coefficients between various features, including frequencies of singlets and cellular interactions as well as fold change induction of cellular interactions after Blinatumomab treatment. P value legend: ns not significant, * P < 0.05, *** P < 0.001. Abbreviations: UMAP = uniform manifold approximation and projection, GR = good responder, NR = non-responder, FC = fold change. Asterisks in cell type labels indicate interactions between the respective cell types. Fig.5: Virus-induced alterations of cellular landscapes and interaction networks. (A) Schematic overview of the experimental design. (B) UMAP display of the cellular landscape (left). Recorded cells were processed with PICtR; out of 34,369,995 cells, 262,638 sketched cells are displayed. Alluvial plots depicting the change of single-cell frequencies over time and across organs (right). (C) UMAP displaying the interacting cell landscape (n = 414,560) (left). Alluvial plots depicting the change of interacting cell frequencies over time and across organs. (D) Principal component analysis (PCA) of single-cell and interacting cell frequencies across organs and time points, encoded by color and shape, respectively. (E) Scaled Euclidean distances to the naive state in PCA space, representing global similarities or differences in single-cell and interaction landscapes. P values were calculated with Wilcoxon Rank sum test and adjusted according to Benjamini Hochberg. Error bars indicate the mean and standard deviation. Abbreviations: UMAP = uniform manifold approximation and projection, D3 = day 3, D7 = day 7, BM = bone marrow, LN = lymph node, Ag = LCMV antigen-specific, Plb = plasmablast, HSPC = hematopoietic stem and progenitor cells, Eosino = eosinophil / basophil, PC = plasma cells, PC1 / PC2: principal components 1 / 2, Granulo = granulocytes, Macro-like = macrophage-like, IgD = immunoglobulin D, Prog = progenitor cells, cDC = classical dendritic cells, pDC = plasmacytoid dendritic cells, My = myeloid, NK = natural killer, Mono = monocytes. P values legend: ns not significant, **** P < 0.0001. Fig.6: Cellular interaction dynamics underlying immune response to LCMV infection. (A) Line plots depicting the frequency of cell types across time points and organs; obtained from k-means clustering (k=7). Clusters with no change in dynamics are not shown. Horizontal bar plots at the top indicate the percentage of interactions contributing to each organ or comprising LCMV- specific cells for each cluster. (B) Alluvial plots showing the fraction of LCMV-specific CD4+ and CD8+ T cells (C) Alluvial plots showing the fraction of cellular interactions comprising LCMV- specific CD4+ and CD8+ T cells. (D) Box plots displaying the log2OR enrichment or depletion of LCMV-specific T cell interaction against non-antigen-specific T cell interactions relative to corresponding singlet population on day seven across the organs. P-values were calculated using Fisher’s exact test (two sided). (E) Line plots showing the scaled fraction of HSPCs and of NK*myeloid cellular interactions in the bone marrow. Error bars depict the standard error of the mean. (F) Box plots depicting the log2 fold change of monocytes normalized to total event number. (G) Box plots showing normalized Monocyte*B cell interactions. (H) Box plot showing the frequency of single plasma cells (I) Box plot showing the frequency of cellular interactions between plasmablasts and LCMV-specific CD4+ T cells (adjusted to the respective single cellfrequencies). The p-values were calculated using least squared means (Lenth et al., 2022) andcorrected according to Benjamini-Hochberg. Asterisk in cell type labels indicates interactions between the respective cell types. Abbreviation: UMAP = uniform manifold approximation and projection, D3 = day 3, D7 = day 7, Ag = LCMV antigen-specific, BM = bone marrow, LN = lymph node, Plb = plasmablast, HSPC = hematopoietic stem and progenitor cells, Eosino = eosinophil / basophil, PC = plasma cells, Granulo = granulocytes, Macro = macrophages, IgD = immunoglobulin D, Prog = progenitor cells, cDC = classical dendritic cells, pDC = plasmacytoid dendritic cells, My = myeloid, NK = natural killer, Mono = monocytes. P values legend: ns not significant, * P < 0.05, ** P < 0.01, **** P < 0.0001. Fig.7: Technical aspects of cytometry-based cellular interaction mapping. (A) UMAP of the overall cellular landscape corresponding to Figure 1D colored by the numeric cluster labels. (B) Histogram of the FSC ratio. Cells exceeding Otsu’s threshold are highlighted. (C) Feature plots of UMAP embeddings from Figure 1D color-coded by the expression levels of selected markers. (D) Heatmap depicting normalized mean feature expression (rows) within merged clusters derived from Louvain clustering (columns) across replicates. (E) Heatmap depicting the adjusted rand index values across different cluster resolutions, comparing (i) conventional flow cytometry parameters to (ii) a combination of imaging and conventional flow cytometry markers. Coloring indicates the degree of similarity between the two feature sets. (F) Impact of flow rate and cellular concentration on cellular interactions is relatively mild in human PBMCs. Left: Short-term cultures of PBMCs with or without CytoStim measured at low or high flow rate (n = 4). Two-way ANOVA (CytoStim: F(1,13) = 189.138, P = 4.01e-09, flow rate: F(1,13) = 6.598, P = 0.023), followed byTukey’s Honest Significant Differences test for the flowrate. Right: Short-term cultures of PBMCswith or without CytoStim at different cell densities but constant flow rate (n = 4). Two-way ANOVA (CytoStim F(1,20) = 280.951, P = 3.05e-13, cell concentration: F(2,20) = 2.053, P = 0.155). (G) Impact of flow rate and cellular concentration on cellular interactions in murine spleens is more pronounced. Left: Baseline interactions in spleens measured with different flow rates (n = 4). Right: Baseline interactions in spleens at different cell densities but constant flow rate (n = 4). One-way ANOVA (flow rate: F(2,9) = 115.749, P = 3.79e-07, cell concentration: F(2,9) = 61.397, P = 5.68e-06), followed by Tukey’s Honest Significant Differences tests. (H) Boxplots of T*B cell interactions upon Blinatumomab treatment that were either fixed or fixed after a freeze-thaw cycle (n = 4). P values were determined with a two-sided Welch’s t-test. P value legend: ns not significant, * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Fig.8: Interact-omics framework of the present invention resolves complex cellular interactions induced by CytoStim™. (A) UMAP embedding of Fig.2A, depicting human PBMCs treated with or without CytoStim, merged UMAP of n = 4 replicates per condition. (B) Histogram of the FSC ratio. Cells exceeding Otsu’s threshold are highlighted. (C) Stacked bar plot displaying the fraction ofPICs in each cluster obtained from Louvain clustering (bottom). (D) Heatmap showing scores ofmarker enrichment modeling (MEM) on clusters shown in Figure 2A. Interacting cells have an enrichment of the FSC ratio (bar plots on top) and co-express various cell type specific markers, like CD3, CD33 and CD19, and HLA-DR. (E) Split UMAP display based on CytoStim™ treatment. Interacting cells show specific enrichment upon co-incubation with CytoStim™. (F) Feature plots, showing UMAP embeddings from Figure 2B color-coded by expression levels of selected markers. Fig.9. Interact-omics framework of the present invention resolves antigen-specific interactions in the ovalbumin (OVA)-OT-II model. (A) UMAP of the overall cellular landscape (numeric cluster labels instead of the partial annotation in Figure 2G); the numeric cluster labels are used in Figure 9D. (B) Histogram of the FSC ratio. Cells exceeding Otsu’s threshold are highlighted. (C) UMAP of the overall cellular landscape with cells having an FSC ratio above Otsu’s threshold highlighted. (D) Proportions for singlets and interacting cells in each cluster. Clusters 6 and 8 were selected as interacting cell clusters (85th percentile). (E) Selected feature plots for the overall cellular landscape. (F) Marker Enrichment Modelling (MEM) heatmap for the overall cellular landscape. (G) Singlet point density UMAPs in the presence or absence of ovalbumin. (H) UMAP of the interacting landscape from Figure 2H with selected features highlighted. (I) Marker Enrichment Modelling (MEM) heatmap for the interacting cell landscape. Fig.10: Interact-omics framework of the present invention maps cellular interaction dynamics of CAR-T cells. (A) UMAP of the overall cellular landscape, corresponding to the UMAP in Figure 3B; labels correspond to panel (D). Decimal points indicate subclustered populations. (B)Histogram of the FSC ratio. Cells exceeding Otsu’s threshold are highlighted in orange. (C)UMAP of the overall cellular landscape with cells above Otsu’s threshold highlighted. (D) Proportions for singlets and doublets in each cluster. Clusters 8.1, 11, 13, 18, and 19 were selected as interacting cell clusters (85th percentile). (E) Selected feature plots for the overall cellular landscape. (F) Marker Enrichment Modelling (MEM) heatmap for the overall cellular landscape. (G) Overall point density UMAPs for the control condition and CAR-T cell-treated samples at each timepoint. (H) Interacting cell landscape corresponding to the UMAP of Figure 3C, highlighting selected features. (I) Marker Enrichment Modelling (MEM) heatmap for the interacting cell landscape. (J) Paired analysis of interactions between B cells and CD4+ CAR-T cells or B cells and endogenous CD4+ T cells normalized to the frequency of singlets available for the interaction at each timepoint. (K) Paired analysis of interactions between B cells and CD8+ CAR-T cells or B cells and endogenous CD8+ T cells normalized to the frequency of singlets available for the interaction at each timepoint. ns not significant, * P < 0.05, ** P < 0.01, *** P < 0.001, paired two-sided Welch’s t- test. (L) Interacting cell point density UMAPs of the control condition and after adding CAR-T cells for the indicated time points. Fig.11: Cellular interaction dynamics upon Blinatumomab treatment. (A) UMAP of the overall cellular landscape, corresponding to the UMAP in Figure 3G; labels correspond to panel (D). Decimal points indicate subclustered populations. (B) Histogram of the FSC ratio. Cells exceeding Otsu’s threshold are highlighted. (C) UMAP of the overall cellular landscape with cells exceeding Otsu’s threshold highlighted. (D) Proportions for singlets and interacting cells for eachcluster. Clusters 9, 15, 16, 19, and 20 were selected as interacting cell clusters (85th percentile).(E) Selected feature plots for the overall cellular landscape. (F) Marker Enrichment Modelling (MEM) heatmap for the overall cellular landscape. (G) Overall point density UMAPs for the control condition and Blinatumomab-treated samples at each timepoint. (H) Point density UMAPs of the cellular interaction landscape for the control condition and Blinatumomab-treated samples at each timepoint, corresponding to Figure 3H. (I) UMAP corresponding to Figure 3H, displaying selected features of the interacting cell landscape. (J) Marker Enrichment Modelling (MEM) heatmap for the interacting cell landscape. (K) Composition of all B cell events across time, including single B cells and cellular interactions that involve B cells. Fig.12: Interact-omics framework of the present invention identifies independent and additive features of Blinatumomab response (A) GGally plot displaying pairwise scatter plots for the top features identified in the univariate analysis. Histograms and distributions of all features are shown on the diagonal. The Pearson correlation coefficients (Corr) are shown for all samples together and separately for the response groups (NR: non responder or GR: good responder), respectively. Box plots in the column on the right show comparisons of good responders vs. non- responders for all features. (B) The Receiver Operating Characteristic (ROC) curves showing the accuracy of three logistic regression models with either the T cell / B cell ratio (light blue line), the fold change of B*T(*My) interactions between good responders and non responders upon Blinatumomab treatment and a combined model with both features. The Area Under the Curve (AUC) metrics are indicated above the plot. P value legend: * P < 0.05. Abbreviations: GR = good responder, NR = non responder, FC = fold change. Red asterisks in cell type labels indicate interactions between the respective cell types. Fig.13: Cluster frequencies of single and interacting cell landscapes of LCMV- infected mice. (A) Heatmap of single cell cluster frequencies across organs, time points and replicates. (B) Cellular interaction cluster frequencies across the organs, time points and replicates. Samples and clusters are ordered based on hierarchical clustering (for reasons of readability, the dendrograms not shown). Abbreviations: D3 = day 3, D7 = day 7, BM = bone marrow, LN = lymph node, Ag = LCMV antigen-specific, Plb = plasmablast, HSPC = hematopoietic stem and progenitor cells, Eosino = eosinophil / basophil, PC = plasma cells, Granulo = granulocytes, Macro-like = macrophage-like, IgD = immunoglobulin D, Prog = progenitor cells, cDC = classical dendritic cells, pDC = plasmacytoid dendritic cells, My = myeloid, NK = natural killer, Mono = monocytes. Fig.14: Interacting cell clusters used for annotation of LCMV-infected mice. (A) Cellular interactions sampled for CD45.1+ and CD90.1+ cells based on Otsu’s thresholding method, clustered, annotated and visualized in a UMAP. (B) UMAP as in (A) colored by three exemplary makers (C) UMAP as in (A) split by time and organ. (D) Marker Enrichment Modelling (MEM) heatmap for LCMV-specific interactions. (E) Cellular interactions sampled for CD19+ cells based on Otsu’s thresholding method, clustered, annotated and visualized in a UMAP. (F) UMAP as in (E) colored by three exemplary makers (G) UMAP as in (E) split by time and organ. (H) Marker Enrichment Modelling (MEM) heatmap for cellular interactions involving B cells. (I) Cellular interactions sampled for CD3+ cells based on Otsu’s thresholding method, clustered, annotated and visualized in a UMAP. (J) UMAP as in (I) colored by three exemplary makers (K) UMAP as in (I) split by time and organ. (L) Marker Enrichment Modelling (MEM) heatmap for T and NK interactions. Abbreviations: D3 = day 3, D7 = day 7, BM = bone marrow, LN = lymph node, Ag = LCMV antigen-specific, Plb = plasmablast, HSPC = hematopoietic stem and progenitor cells, Eosino = eosinophil / basophil, PC = plasma cells, Granulo = granulocytes, Macro = macrophages, IgD = immunoglobulin D, Prog = progenitor cells, cDC = classical dendritic cells, pDC = plasmacytoid dendritic cells, My = myeloid, NK = natural killer, Mono = monocytes. Fig.15: Technical aspects of cytometry-based cellular interaction mapping. (A). Performance of different classification methods based on the FSC ratio as measured by the F1 score. Manual image annotation served as the ground truth; see Methods for details. n = 4 replicates are shown in the scatter plot; bars indicate the mean F1 score. (B). Dot plot displaying the forward scatter area (FSC-A) and forward scatter height (FSC-H) properties of ground truth singlets; the Otsu threshold of the FSC ratio is shown as a diagonal line. The bar plots show all ground truth singlets split into correctly classified and misclassified events according to the FSC ratio threshold and are colored by marker expression. n = 4 replicates; error bars indicate the standard deviation. (C). Gating strategy to select Lymphocytes and live cells using scatter properties and a live-dead (LD) marker. This gating strategy was employed throughout the manuscript. (D-G). Louvain clustering performed on the top important features from the feature importance analysis (see Figure 1C) and cell type markers. (D). Annotated UMAP representation. (E). UMAP embedding from D with cells exceeding the Otsu threshold of the FSC ratio highlighted in blue (top) or cells colored by their ground truth annotation (bottom). (F). Relative frequency of singlets and interacting cells in each population classified according to the FSC-ratio. (G). Relative frequency of cells in each population based on the ground truth annotation. (H-K). Louvain clustering performed on cell type markers only. (H). Annotated UMAP representation. (I). UMAP embedding from H with cells exceeding the Otsu threshold of the FSC ratio highlighted in blue (top) or cells colored by their ground truth annotation (bottom). (J). Relative frequency of singlets and interacting cells in each population classified according to the FSC-ratio. (K). Relative frequency of cells in each population based on the ground truth annotation. (L). Feature plots showcasing cell type marker expression in the UMAP embedding from Figure 1F. (M). Heatmap depicting normalized mean feature expression (rows) within merged clusters derived from Louvain clustering (columns, on conventional flow parameters) across replicates. Populations are the same as in Figure 1F. (N). Histogram colored by the ground truth annotation and split by the identified singlet and multiplet clusters in Figure 1F. The Otsu threshold is shown. (O). Performance of different clustering methods evaluated regarding their ability to resolve singlet and interacting populations. All algorithms were used for n = 100 iterations on conventional flow parameters including forward scatter parameters, side scatter parameters, cell type markers and the FSC ratio, see Methods for details. n = 4 replicates are shown in the point plot; bars indicate the mean F1 score. Abbreviations: UMAP: uniform manifold approximation and projection, CD33: myeloid marker, CD19: B cell marker, CD3: T cell marker. Fig.16: Effect of sample processing methods on ex vivo cellular interactions. (A). Schematic depiction of the experimental approach. (B). Overall cellular landscape across all experimental conditions. n = 3,292,837. Labels 1 and 2 refer to CD45 APC-Fire810 and CD45 PE-Fire640, respectively. (C). Feature plots for the UMAP display in B, colored by the two differently labeled antibodies against CD45. (D). Interacting landscape across all experimental conditions. n = 23,620. (E). Dot plot for the CD45 signals in interacting populations, showcasing single-positive and double-positive interactions. (F). Feature plots for the UMAP from panel D, colored by the CD45 signal intensities. (G). Quantification of single-positive (white) and double positive, newly acquired (black) interactions in CytoStimTMtreated and untreated samples, across different experimental conditions. Left: Varying incubation times at 4°C after mixing, mimicking long sample processing times. Middle: Different cellular concentrations ranging from 25,000 to 250,000 cells in 50 μL during staining and acquisition. Right: Fixation with 2 % paraformaldehyde after staining compared to no fixation. The number of replicates is shown in each plot and ranges from n = 2 to 3. Fig.17: Characterization of functional interactions by assessment of phosphorylated CD247. (A). Schematic overview of the experimental approach. PBMCs from 4 healthy donors were incubated in the presence or absence of Blinatumomab (Blina) or Cytostim (CS). (B). UMAP of the overall cellular landscape. Recorded cells were processed with PICtR, out of 1,204,382 cells, 70,954 sketched cells are displayed. (C). UMAP of interacting cells (n = 52,239) (D). Point density UMAP of interacting cells split into the conditions. (E). Histograms of scaled fluorescence intensity of pCD247 for each interacting cell population from C. Left panel: T*B interactions. Middle panel: T*My interactions. Right panel: T*B*My. (F). Mean fluorescence intensity of pCD247 per donor and condition. Left panel: T*B interactions. Middle panel: T*My interactions. Right panel: T*B*My. P values were determined with a paired t-test. Abbreviations: UMAP = uniform manifold approximation and projection, Ctrl. = control, Blina = Blinatumomab, CS = Cytostim. Red asterisks in cell type labels indicate interactions between the respective cell types. Fig.18: Comparison of the Interact-omics approach to imaging flow cytometry. (A). UMAP representation of the overall cellular landscape derived from the fluorescent intensity values, n = 306538. Intensity values are based on the sum of the pixel intensities in the mask as selected by ImageStream®X, background subtracted. The experiment corresponds to day 7 in Figure 19. (B). UMAP representation of the interacting landscape, n = 8683. The heterogeneous cluster is mostly comprised of likely B*CD4*CD8 multiplets. The unknown cluster expresses CD19 and CD3 but no other T cell markers, hindering confident annotation. (C). Pseudo-colored example images for cellular interactions in the brightfield and fluorescence channels. (D). Left: UMAP displays from A and B colored by the number of cells identified through image segmentation. Right: Bar plots comparing the populations identified through Interact-omics (x-axis labels) and image segmentation (color code). (E). Left: UMAP displays from A and B colored by populations as identified through conventional gating. Right: Bar plots comparing the populations identified through Interact-omics (x-axis labels) and conventional gating (color code). NA indicates that the event does not fall into any conventional gate. (F). Fold changes of the frequencies + / - LCMV infection. Holm-corrected estimated marginal means comparison. Left: Populations identified by Interact-omics. Right: Populations identified through conventional gating. (G). Gating strategy for conventional gating. Abbreviations: Ag = antigen-specific, UMAP = uniform manifold approximation and projection, BF = brightfield. Fig.19: Interactions acquired a posteriori after in vivo experiment. (A). Schematic overview of the experimental approach. LCMV-specific CD4+ T cells were transferred into CD45.2 host mice 5 days before infection with LCMV (group A) or the respective control (group B). Additionally, CD45.1 host mice were infected with LCMV (group C) or left untreated (group D). n = 3 for groups A, C, D and n = 4 for group B. (B). Single-cell landscape of all experimental groups. Out of n = 23,490,812 processed cells, n = 245,316 are shown in the UMAP display. (C). Feature plots for panel B, colored by the expression of the congenital markers CD45.1 and CD45.2. (D). Interacting landscape across all experimental groups. Out of n = 731621 identifiable interactions, n = 93065 are shown. (E). Dot plots showcasing the expression of the congenital markers CD45.1 and CD45.2 in interacting populations from the unmixed controls for the untreated and infected conditions, and the mixed spleens from infected mice (infected+infected, group A + group C) or untreated mice (control+control, group B + group D). (F). Bar plots depicting the log2 fold changes (FC) between the LCMV infected and untreated conditions for each interacting population. Solid bars indicate the log2FC between group A (infected) und group B (control). Semi-transparent bars show the log2FC for single-positive interactions in mixed samples (A+C for the infected condition, and B+D for the control). Transparent bars depict log2FC between the respective double-positive interactions, which were definitely acquired ex vivo. n = 3 mixes. (G). Linear relationships between the log2FC between infected and control conditions for unmixed controls, single-positive interactions after mixing and double-positive interactions after mixing. n = 3 mixes. Fig.20: Interacting cell landscape in juvenile idiopathic arthritis (JIA). (A). Publicly available spectral flow cytometry data of PBMCs and SFMCs of JIA patients54. hree comparisons (indicated by the arrows) were made for the interacting cell landscape. (B). UMAP of the overall cellular landscape. Recorded cells were processed with PICtR, out of 7,843,646 cells, 80,000 sketched cells are displayed. (C). UMAP of interacting cells (n = 12,908) (D). Point density UMAP (left panel) and differential abundance (right panel) of interacting cells comparing PBMCs from healthy donors vs. JIA patients. (E). Quantitative comparisons of interacting cell frequencies between PBMCs from healthy donors (n=18) and JIA patients (n=36). Top: Non-adjusted frequencies. Bottom: Interaction frequencies adjusted by the harmonic mean of the singlet frequencies of the contributing cells (see Methods). P values were determined with a two-sided t- test and adjusted for multiple testing using Benjamini-Hochberg correction. (F). Qualitative differences in CD4T*cl.mono interactions. P values were determined with a two-sided Wilcoxon rank sum test and adjusted for multiple testing using Benjamini-Hochberg correction. (G). Point density UMAP (left panel) and differential abundance (right panel) of interacting cells comparing PBMCs of JIA patients with inactive disease (n=11) vs. active (n =25). (H). Quantitative comparisons of interacting cell frequencies between PBMCs from JIA with inactive and active disease. Top: Non-adjusted frequencies. Bottom: Interaction frequencies adjusted by the harmonic mean of the singlet frequencies of the contributing cells (see Methods). P values were determined with a two-sided t-test and adjusted for multiple testing using Benjamini-Hochberg correction (I). Qualitative differences in T*B interactions. P values were determined with a two- sided Wilcoxon rank sum test and adjusted for multiple testing using Benjamini-Hochberg correction. (J). Point density UMAP (left panel) and differential abundance (right panel) of interacting cells comparing PBMCs of JIA patients with active disease vs. SFMC of active disease. (K). Quantitative comparisons of interacting cell frequencies between PBMCs of JIA patients with active disease (n=25) vs. SFMC of active disease (n=8). Top: Non-adjusted frequencies. Bottom: Interaction frequencies adjusted by the harmonic mean of the singlet frequencies of the contributing cells (see Methods). P values were determined with a two-sided t- test and adjusted for multiple testing using Benjamini-Hochberg correction. (L). Qualitative differences in CD4T*mono interactions. P values were determined with a two-sided Wilcoxon rank sum test and adjusted for multiple testing using Benjamini-Hochberg correction. Abbreviations: UMAP = uniform manifold approximation and projection, PBMC = peripheral blood mononuclear cells, SFMC = synovial fluid mononuclear cells. Red asterisks in cell type labels indicate interactions between the respective cell types. Fig.21: Interacting cell landscape in the proximal small intestine. (A) Schematic overview of the experimental approach (adapted from Funk, M. C. et al. Aged intestinal stem cells propagate cell- intrinsic sources of inflammaging in mice. Dev. Cell 58, 2914-2929.e7, 2023). n = 3 young mice and n = 3 aged mice were analyzed. Created in BioRender. (B) Overall cellular landscape of epithelial cells and the immune microenvironment. Recorded cells were processed with PICtR; out of 4,167,516 cells, 399,608 sketched cells are displayed. (C) UMAP of the interacting cell landscape, n = 32,554. (D) Comparison of the T cell population frequency in young (n = 3) and aged (n = 3) mice. P values were calculated using least squared means (two-sided) and were Bonferroni-corrected. Error bars indicate the mean and standard deviation. € Comparison of interacting populations that involve T cells in young (n = 3) and aged (n = 3) mice. P values were calculated using least squared means (two-sided) and were Bonferroni-corrected. Error bars indicate the mean and standard deviation. EXAMPLES The invention is demonstrated through the examples disclosed herein. The examples provided represent particular embodiments and are not intended to limit the scope of the invention. The examples are to be considered as providing a non-limiting illustration and technical support for carrying out the invention. Methods Animals For experiments with antigen-specific T cells, cells were isolated from B6.Cg- Tg(TcraTcrb)425Cbn / J (OT-II) or LCMV-TCRtg P1454 and Smarta55 mice expressing the congenic markers CD45.1 or CD90.1. Human samples PBMC samples were obtained from healthy blood donors as buffy coats Mononuclear cells were isolated by Ficoll (GE Healthcare) density gradient centrifugation and stored in FCS 10% DMSO in liquid nitrogen until usage. For the Blinatumomab response analysis, bone marrow samples from 42 patients with a B-ALL relapse were assessed. Samples were directly collected before the start of the Blinatumomab course and processed in routine diagnostics by Ficoll density gradient centrifugation and minimal residual disease quantification. Leftover cells were stored in FCS 10% DMSO in liquid nitrogen for research purposes. The analysis was conducted according to the Declaration of Helsinki and in accordance with local ethical guidelines; written informed consent of patients was obtained. Good response to Blinatumomab (n = 18) is defined as minimal residual (MRD) negativity directly after a Blinatumomab course (28 days) and all subsequent time points. Non-response to Blinatumomab (n = 4) is defined as leukemic cell persistence (at morphological or high MRD level) without any reduction after a Blinatumomab course. The remaining 20 patient samples could not be unequivocally assigned to these response states (Good response vs. Non- response) or had no residual disease at the start of the Blinatumomab course. Isolation of murine immune cells Mice were euthanized through cervical dislocation. For isolation of antigen-specific T cells, the spleen and various lymph nodes (including inguinal, axial, submandibular, and mesenteric) were carefully extracted. Tissues were homogenized using a 40μm filter (Falcon) and a syringe plunger in cold RPMI (Sigma Aldrich) with 2% FCS (Gibco by Lifetechnologies). Subsequently, single-cell suspensions from spleens were treated with erythrocyte lysis solution (ACK buffer, containing 0.15 M NH4Cl, 1 mM KHCO3, and 0.1 mM Na2EDTA in water from Lonza) for a duration of 5 minutes. For some readouts, these suspensions were combined with the lymph node samples or maintained separated. CD4+ and CD8+ T cells were purified using either the Dynabeads Untouched Mouse CD4 Cells Kit (Invitrogen) or the murine CD4+ T cell isolation kit and the murine CD8+ T cell isolation kit (Miltenyi) according to the manufacturer’s instructions. Purified fractions were stained for further purification using FACS (see section Flow cytometry, cell sorting and image cytometry). For in vivo experiments, femurs, spleen and various lymph nodes were dissected and kept separate on ice. Lymph nodes and spleens were individually processed as described above. Femurs were flushed using FACS buffer and homogenized using a 40μm filter (Falcon) and a syringe plunger. Ex vivo murine co-cultures Cultures containing OT-II CD4+ T cells were incubated at 37°C with 5% CO2 in U- bottom plates in 200μL Dulbecco’s Modified Eagle’s Medium GlutaMAX (DMEM GlutaMAX, Gibco), supplemented with 10% heat-inactivated Fetal Calf Serum (FCS, Gibco), sodium pyruvate (1.5mM, Gibco), L-glutamine (2mM, Gibco), L-arginine (1x, Sigma), L-asparagine (1x, Sigma), penicillin / streptomycin (100 U / mL, Sigma), folic acid (14μM, Sigma), minimum essential medium, non-essential amino acids (1x, ThermoFisher), MEM vitamin solution (1x, ThermoFisher) and β- mercaptoethanol (57.2μM, Sigma).5 × 104OT-II cells were incubated with 1 × 105splenocytes containing various antigen presenting cell populations (B cells and myeloid cells) in presence or absence of ovalbumin peptide (323-339, Invivogen). For murine CAR-T in vitro assays, GFP- expressing CD19 specific CAR-T cells were generated as previously described (Gottschlich et al., 2023), thawed and washed with PBS. Next, cells were transferred to 10% FCS RPMI1640 containing 0.05 μg / ml IL15 (Peprotech) and 0.1% ß-mercaptoethanol. To recover from freezingprocedures, cells were incubated under the same conditions as described above prior to the co-culture assay. Frozen murine splenocytes were thawed and incubated together with CAR-T cells at a ratio of 1:2 target:effector ratio (CAR-T cells:splenocytes) for 0.5 to 3h. Subsequently, cells were harvested, washed with FACS-buffer, stained with surface markers and analyzed. Ex vivo human co-cultures Cryopreserved PBMCs were thawed in a water bath at 37°C, transferred to 10% FCS RPMI-1640 and washed twice. After each washing step, cells were centrifuged at 350g for 5 min.2 × 105cells were plated in 10% FCS RPMI-1640 and cultured short term for up to 5h in 200 μl RPMI 10% FCS. CytoStim™ (Miltenyi) was used in concentrations recommended by the manufacturer at 37°C for 2 hours before harvest. For experiments using a Blinatumomab analog (Invivogen), a concentration of 50 ng / ml was used. The incubation period ranged from 0.25 to 5 hours at 37°C and 5% CO2 in 96-well U-bottom plates. For experiments assessing the stability of Blinatumomab- induced interactions upon cryopreservation, cells were either incubated for 2h in presence of the compound and stained with surface antibodies and fixed with 4% PFA (ThermoFisher) or frozen in Bambanker freezing medium (Nippon Genetics), thawed after 18h and treated in the same way as the non-frozen cells. For in vitro performance experiments, human PBMCs were treated with CytoStimTMas described above; control groups were left untreated. Subsequently, cells were split into two groups each and stained with CD45-APC-Fire810 or CD45-PE-Fire640, respectively. After mixing the labeled groups, cells were incubated for 0-4h at 4 °C (200000 cells / well in 50 μL during staining / acquisition) or processed at seeding densities of 25,000 to 250,000 cells per well in 96- well plates (50 μL during staining / acquisition). Subsequently, cells were harvested, washed with FACS-buffer, stained with surface markers, fixed with 2 % PFA (except the non-fixed control) and analyzed. For measuring phosphorylated CD247, human PBMCs were seeded at 100,000 cells / well in 200 μL and treated for 1h with Blinatumomab (160 ng / mL) or CytoStimTMas described above. Following the stimulation period, cells were 1205 fixed immediately by adding CytoFix buffer (15 min, 4°C). Cells were washed and resuspended in 200 μL 2.5x Perm / Wash buffer, incubated for 30 min at 37°C, and stained overnight at 4°C before analysis. In vivo mouse experiments 5 days prior to infection, 1x104 LCMV-specific CD4+ T cells (SMARTA; expressing congenic marker CD90.1) and CD8+ T cells (P14; expressing congenic marker CD45.1) were administered intravenously into C57BL6 / J in 300 µl of balanced salt solution (BSS), resulting in an approximate seeding of 1x103cells per mouse (Hataye et al., 2006). The viral infection was induced intraperitoneally using 200 PFU of the LCMV Armstrong strain (Zajac et al., 1998). Mice were euthanized on day 4 and day 7 post-infection, and various tissues including the spleen, mesenteric lymph nodes, and bones were dissected and processed for spectral flow cytometry analysis. For the in vivo performance experiment, LCMV-specific CD4 T cells were transferred into C57BL6 (CD45.2) hosts 5 days prior to infection as described above. CD45.1 (B6.SJL- PtprcaPepcb / BoyCrl) and CD45.2 hosts were infected intraperitoneally as described above, and spleens were harvested on day 7 post-infection. Spleens were split into 4 equal pieces and mixed across CD45.1 / CD45.2 hosts for joint tissue homogenization (see Figure 18A). Mixed samples were processed for spectral flow cytometry analysis. Flow cytometry, cell sorting and image-enabled flow cytometry Unless otherwise stated, cell suspensions were resuspended in 2% FSC 0,5 mM EDTA PBS (FACS buffer) for performing flow cytometric stainings. For ex vivo readouts with bi-specific engagers and antigen specific T cells, cells were harvested, centrifuged 5 min at 350 g and stained with surface marker panel master mixes using ACS buffer and addition of Brilliant Stain buffer (BD) according to the manufacturer’s recommendation. Cells were stained for 30 min on ice in 96-well V-bottom plates, followed by washing with FACS buffer, centrifugation for 5 min at 350g and resuspension in 200 μl FACS buffer. For more time-consuming in vivo experiments, cells were labeled with fixable dead cell exclusion dyes followed by fixation of obtained single-cell suspensions with cold 2% PFA PBS for 15 min at room temperature. Cells were washed, centrifuged for 5 min at 350 g and then stained for 12h at 4°C. After washing and centrifugation for 5 min at 350g, cells were filtered through a 35-μm cell strainer and kept on ice until flow cytometric analysis. For flow cytometric analysis, a Cytek Aurora (Cytek Biosciences) or LSR Fortessa (BD) equipped with 5 lasers was used. For sorting of naive T cells in ex vivo setups, FACSAria Fusion or FACSAria II sorters equipped with 70 μm nozzles were used. For imaging cytometry, image- enabled cell sorting (ICS) using the BD CellViewTM Imaging Technology was used (Schraivogel et al, 2022). For ICS, PBMCs were incubated for 2h with CytoStim™, stained with surface markers followed by fixation with 2% PFA PBS as described above and operated using a 100 μm sort nozzle, with the piezoelectric transducer driven at 34 kHz and automated stream setup by BD FACSChorusTMSoftware, and a system pressure of 20 psi. For the ImageStream®X experiment, data was acquired using the Cytek® INSPIRE™ software. Image-enabled flow cytometry analysis For image-enabled flow cytometric analysis, radiofrequency images obtained from an image- enabled cell sorter underwent processing as described in Schraivogel et al (2022). The raw image .tiff files were imported into ImageJ and processed with the BD CellViewTM plugin. The corresponding .fcs files were loaded into FlowJoTM (BD), and cells were gated as living CD45+ cells. Using the flowCore (Hahne et al., 2009), CytoML (Finak et al., 2018), and flowWorkspace (Greg et al., 2017) packages, the generated FlowJo workspace was loaded into R (v.4.3.0) for further processing. Subsequently, images were converted to .jpg format, and channels containing the light-loss parameter (brightfield), and FSC and SSC parameters were kept for downstream analysis. Four replicates, each comprising 1000 images, were manually categorized as singlet, doublet, triplet or higher-plex multiplets, and categories were used to train a decision tree based classification model using Rpart (R Foundation for Statistical Computing, 2021, Therneau et al. 2024) and caret (Kuhn et al., 2008). As features for the model, the image-based features, the conventional flow cytometry parameters, and the forward scatter channel (FSC) ratio, defined by the quotient of FSC-A and FSC-H, were used. Feature importances in the model were determined to identify relative contributions of each variable in making accurate predictions. Otsu (Otsu et al.1979) thresholding, which minimizes intra-group variance, was applied to a histogram of the FSC ratio divided into at least 1000 bins, effectively separating the data into two categories based on whether their FSC ratio is above or below the threshold. Louvain clustering was performed for n = 100 iterations (resolution = 1) on all or a subset of the following features: image-based parameters, conventional flow parameters and the FSC ratio. Consensus clustering solutions were calculated using soft least squares Euclidean consensus partitions as implemented in the clue (Hornik et al.2005) package. Data from the imaged-enabled flow cytometer were visualized in UMAP (McInnes et al.2018) embeddings using image-based, conventional flow parameters and the FSC-ratio as input features. UMAP embeddings were computed across 15 nearest neighbors and a minimum Euclidean distance of 0.1. Louvain clustering was performed on the same input data with different cluster resolutions and clusters were annotated based on cell type- exclusive markers and their combinations. Additionally, Louvain clustering was performed for n = 100 iterations using only conventional flow cytometry parameters and the FSC-ratio with different resolution parameters. Variation in cluster labels was assessed using the adjusted rand index (Scrucca et al., 2023). For ImageStream-based analyses, ImageStream®X fluorescence intensity values (based on the sum of the pixel intensities in the mask as selected by ImageStream®X, background subtracted) were compensated and transformed using FlowJo (v10.10) and IDEAS (v6.2). Data was processed using PICtR (see below). Interacting populations were solely annotated based on mutually exclusive marker expression, since forward scatter properties are not acquired by ImageStream®X. For conventional gating, gates were selected in FlowJo according to the strategy shown in Figure 18G. For cell segmentation from brightfield images, the cyto2 model from the Python package CellPose (Stringer et al.2021) was used with a cell pixel diameter of 20. To remove cellular debris, events that met any of the following criteria were excluded: major axis length < 15 pixels or > 40 pixels, circularity < 0.7, area < 100 pixels or > 1000 pixels. Area and major axis length were computed using the Python package scikit-image (van der Walt et al. 2014). Circularity was calculated using the formula 4∙ ^ ∙ ^^^^^^^^^^^^^² with perimeter values also obtained from scikit-image. This filtering process excluded approximately 4% of the detected objects. Identification and analysis of physically interacting cells with Physically Interacting Cells toolkit in R (PICtR) 1) Performance Performance test was accomplished on the imaged-enabled flow cytometry data with n = 3865 manually classified events across n = 4 replicates. Several thresholding methods based on the FSC ratio were used to define a cutoff of events with a high or low FSC ratio (see Table 1 for details). Otsu, IsoData, Intermodes, RenyiEntropy, Li, Shanbhag, Huang, and Mean algorithms were used as implemented in the R package “autothresholdr” (Landini et al.2017), and the Triangle algorithm was ported from the ImageJ implementation in Java. kmeans clustering was used with k = 2 for thresholding and Gaussian mixture models were computed as implemented in the R package “mclust” (Scrucca et al.2023). Performance of the methods was evaluated based on the annotation of the image enabled flow cytometry data and reported as F1 scores, where 1 indicates a perfectly accurate reproduction of the manual ground truth classification. Next, different clustering algorithms (see Table 2) were evaluated regarding their ability to discriminate single and interacting cells considering conventional flow cytometry features (forward scatter, side scatter, cell type markers CD45, CD3, CD19, HLA-DR and CD33, and the FSC ratio). Candidates were selected based on their popularity in the single-cell and flow cytometry fields or based on their performance on high-dimensional single-cell flow and mass cytometry data as evaluated by Weber and Robinson (Weber et al.2016). Louvain and Leiden clustering (implemented through igraph) were used on a shared nearest neighborhood graph with k = 5 nearest neighbors, HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) was employed on a UMAP embedding with k = 15 nearest neighbors and Phenograph was used with Louvain or Leiden clustering. FlowSOM (Spectre implementation), FlowMeans, Rclusterpp and Immunoclust were used directly on the features. Each method was run for n = 100 iterations and the performance was reported as F1 scores based on the ground-truth classification.

[0002] Tab.1: Thresholding methods evaluated for the FSC ratio Tab.2: Methods evaluated for distinguishing singlet and interacting communities 5 10 2) Flow cytometric data pre-processing For full spectrum flow cytometry data, raw FCS files were spectrally unmixed using the inbuilt unmixing function of the SpectroFlo (Cytek Biosciences) software. FCS files were imported into FlowJo (BD) to assess unmixing by visualizing N x N plots. Axes were adjusted wherever needed and parameters for logicle (Parks et al., 2006) or generalized bi- exponential transformation of data were defined for every surface marker individually. PeacoQC (Emmaneel et al., 2022) was used as an automatic quality control mechanism for cytometry data where needed. The population of interest (live cells, see Fig.15B) were exported using channel values defined by the inbuilt export function of FlowJo. 3) Physically Interacting Cells toolkit in R (PICtR) The workflow starts by importing compensated and transformed FACS CSV files into the R programming environment (version 4.3.0). To manage the extensive data without encountering memory issues, BPCells was used for bit-packing compression on a high-performance computing cluster (Parks, 2023). For each measured event, the FSC-ratio, defined by the ratio of FSC-A and FSC-H, was calculated and scaled to transform the data into a similar range as recorded marker expression values. For the downstream analysis, the measured marker expression values, forward scatter, side scatter, and the determined FSC-ratio parameters were used as features. Next, the data was sampled using an atomic sketching approach as implemented in Seurat v5 (Hao et al., 2023). This approach is particularly effective in preserving rare events, including cellular interactions. The underlying concept of atomic sketching involves treating data points as 'atoms' in a dictionary from which a representation is learned through a weighted linear combination. The selection of atoms is based on a leverage score, which reflects the importance of each event in terms of its contribution to the overall data structure and variance. Events with a high leverage score will have a higher contribution to the gene-covariance-matrix, a higher importance for understanding the complexity of the data, and are more likely to be sampled. As a result, non-uniform sampling of the data is obtained. Based on the leverage score, up to 100,000 cells per data set were sampled. Sampled data were further processed with the Seurat workflow. In detail, flow data can be subjected to dimensionality reduction. Hrein, the datasets were transformed into n-1 principal components (PCs), with n being the maximum number of dimensions in the data, however the number of components can be adjusted. The resulting PC space is used to construct shared nearest neighbor graph across the 20 nearest neighbors and to determine the UMAP (Ashhurst et al., 2022) embeddings using 30 neighbors and a minimum cosine distance of 0.3 for the manifold approximation. Furthermore, the shared nearest neighbor graph is used as input for Louvain clustering. Other clustering methods are provided as alternatives. For cells not included in the initial sketching process, cluster labels were determined using Linear Discriminant Analysis as implemented in the R package MASS (Venables et al., 2002). Clusters that contain interacting cells were selected based on the FSC-ratio. A discriminating threshold is obtained using the Otsu method (Otsu et al., 1979), which minimizes inter-group variance, but alternative thresholding methods are also possible. Otsu thresholding is applied to a histogram of the FSC-ratio divided into at least 2000 bins, effectively separating the data into two categories based on whether their FSC-ratio is above or below the threshold. Next, the fraction of cells above and below the FSC-ratio threshold is determined per cluster. Based on the distribution of these fractions, a percentile cutoff is determined. In this study, clusters falling into the 80-85th percentile were selected as physically interacting cell (PIC) clusters, and all cells within the PIC clusters were considered to be interacting cells. However, depending on the dataset, other thresholds may be more accurate. Finally, based on the predicted cluster labels from Linear Discriminant Analysis a comprehensive resolution of all interacting cells was achieved. The identified interacting cells were used to perform principal component analysis, shared nearest- neighbor graph, clustering, and UMAP analysis to obtain a refined cluster resolution of interacting cells. 4) Annotation of physically interacting cells Clusters of single and interacting cells were annotated based on cell type exclusive markers and expert knowledge. Clusters identified as interacting cells populations containing doublets or multiplets were annotated based on the combination of surface markers, and FSC-ratio determined by Otsu thresholding. For example, co-expression of the B cell marker CD19 and the T cell markers CD3 and CD4 within an interacting cluster indicates an interaction of a B and T cell. The key criterion for identification of interacting cell clusters was a high FSC-ratio and the presence of more than one cell type-specific marker. Interacting cell clusters not fulfilling this criterion might represent homotypic cellular interactions (e.g., interactions between two B cells), which we excluded from downstream analysis. Based on the sampled object, a new UMAP embedding and clustering was defined to enhance resolution. 5) Normalization of physically interacting cells Frequencies of interactions are reported as the frequency among all live, high-quality events or the frequency among all interacting cells. Alternatively, interaction frequencies are normalized by taking the frequency of the respective interaction partners into account: Let ^A denotes the fraction or rate of cell type A, and analogously let ^B denotes the fraction or rate of cell type B. Furthermore, let ^ABdenotes the fraction or rate of interacting cells of type A and B. In order to assess the number of such interacting cells, we introduce the enrichment term where denotes the observed and ^AB denotes the expected rate of interacting cells. The expected rate is given by the harmonic mean ^(^A, ^B) of the two singlet rates: We thus get for the enrichment ^AB: (equation 1) Of note, the harmonic mean of a list of numbers tends strongly toward the least element of thelist. In our case with two entries, in case ≫ ^B, we get: (equation 2) The frequency of expected interactions between two cell types with strongly different abundance is thus given by the less abundant cell type. Still, even for the more abundant cell type A, ^ABincreases with increasing ^A: (equation 3) Blinatumomab response analysis Bone marrow aspirates obtained from 42 relapsed B-ALL patients were thawed in a water bath at 37°C, transferred to 10% FCS RPMI-1640 and washed twice. After thawing, each sample was split into two. One half of the sample was cultured in 200 µl RPMI 1640 (10% FCS) supplemented with 50ng / ml Blinatumomab analog (Invivogen) for 1 hour at 37°C and 5% CO2 in a 96 well U- bottom plate. The other half of the sample was cultured in RPMI 1640 (10% FCS) without Blinatumomab supplementation for 1h at the same conditions. After the incubation, cells were harvested, washed with FACS buffer, stained with the surface marker panel and analyzed. Raw FCS files were preprocessed as described above. CSV files were processed with the PICtR workflow as stated above using a 70th percentile cutoff for the identification of interacting cells. To analyze whether certain features are associated with the two response groups (good responder (GR) and non-responder (NR)) the mean value for each feature in the dataset (singlet frequency, interacting cell frequencies and fold changes of interacting cells after Blinatumomab treatment), was calculated for both groups. Then, the fold change for each feature was computed as the ratio of the mean value in the GR group to that in the NR group. Additionally, a two-sided t- test was performed for each feature to test for significance between groups. Prior to performing the correlation analysis, a feature selection was conducted to refine the dataset for more targeted analysis. This selection was based on the results of a univariate analysis, by which features were selected based on abs(t-value) threshold greater than 1.5. For these selected features, a correlation matrix was computed using the cor () function from the stats package. Afterwards, the distance matrices were created and a hierarchical clustering was performed on the rows and columns of the correlation matrix separately. Data visualization: In boxplots the middle line corresponds to the median; the lower and upper end of the box correspond to first and third quartiles, respectively; the whisker spanning from the box to the largest or lowest value no further than 1.5x the interquartile range. Data points beyond the end of the whiskers are plotted individually. If not otherwise stated, error bars indicate the mean and standard deviation. Results A universal framework for cytometry-based quantification of cellular interactions To develop a universal and flexible cytometry-based framework for mapping physical interactions among immune cells, we focused on identifying strategies to accurately discriminate between single cells and physically interacting cells (PICs) in cytometry data. For this purpose, we induced a defined set of cellular interactions among human peripheral blood mononuclear cells (PBMCs) using a bispecific antibody-based reagent (CytoStimTM) that binds both T cell receptors and MHC molecules, thereby physically engaging T cells with antigen presenting cells (Figure 1A, see Methods). Subsequently, we utilized an imaging flow cytometer prototype (Schraivogel et al., 2022) to generate ground-truth data on cellular interactions and concurrently measured cytometry parameters. These comprised five surface markers broadly indicative for distinct immune cell populations, along with a range of image-based and cytometric parameters, such as light scatter profiles. Following data acquisition, we manually classified 1000 randomly selected cellular events, preferably based on imaging information, across 4 replicates into singlets, doublets, triplets or higher-plex cell-cell interactions. To extract cytometric features capable of discriminating between single cells and physically interacting cells, we performed a feature importance analysis considering the manually classified images as ground-truth (Figures 1A, B). This analysis revealed the ratio between signal intensities of forward scatter-area and -height(termed FSC- ratio), alongside other scatter properties, as highly indicative for singlet to multipletdiscrimination (Figures 1B, C), also in line with a common gating based strategy to exclude multiplets from cytometric analyses. Also by relying solely on the FSC ratio to distinguish singlets from multiplets, an F1 score between 0.50 to 0.84 was achieved, depending on the thresholding method used (Figure 15A). Notably, we identified Otsu-based thresholding of the FSC ratio as a robust, reproducible and data-driven approach for scatter-based multiplet identification, while alternative thresholding methods produced similar results (Figure 1J, K, Figure 15A). Clustering- based approaches for simultaneous multiplet discrimination and annotation have also been established for improved classification and identification of interacting cells. Using the most important identified features in the feature importance analysis for singlet to multipletdiscrimination, including image-based parameters (see Figure 1C), together with the surfacemarker expression from the acquired data for Louvain clustering revealed individual clusters single-positive for distinct lineage-defining markers, largely comprising ground-truth single cells of distinct PBMC cell types, as well as separate clusters characterized by the co-expression of mutually exclusive lineage-defining markers and also a high FCS ratio, largely comprising physically interacting cells (Figures 1D, E and Figures 7A, B). Classifying clusters based on FSC ratio into singlets versus multiplets outperformed the approach using scatter properties only (Figure 1K), and enabled annotation of interacting cell partners based on the co-expression of mutually exclusive lineage-defining markers. In line with our previous feature importance analysis, the clusters comprising physically interacting cells were specifically characterized by a high FSC-ratio; separation could be achieved by Otsu’s thresholding method (Figures 1F, G and Figure 7B). Jointly, these findings demonstrate efficient cytometric discrimination between single and physically interacting cells based on multiparametric clustering and FSC-ratio, with the combined information on cluster identity and FSC-ratio providing higher precision and recall compared to FSC-ratio alone (Figure 1H). Importantly, taking both image-based and conventional cytometry-based information into consideration provided highly overlapping information with using conventional flow cytometer parameters (alongside the FSC-ratio) only, suggesting that the proposed strategy can be applied to conventional cytometry approaches to discriminate single cells from physically interacting cells with high accuracy (Figure 1I, Figure 7C, D, E). To explore whether such an approach could also be applied to conventional cytometry without image-based information, we performed Louvain clustering on cell type markers only, followed by FSC ratio-based classification into singlet and multiplet clusters (Figure 15H-K). While this approach outperformed FSC ratio only classification, it remained inferior to using all important features (Figure 1K). In contrast, incorporating both cell type markers and scatter properties – including the FSC ratio – into the clustering, followed by FSC ratio-based classification into singlet and multiplet clusters, yielded results comparable to those achieved when all important features including image-based features were utilized (Figure 1G,H,K,L, Figure 15L-N). This result was reproducible across various cluster-resolutions (Figure 1I). A comparison between different clustering methods showed Louvain clustering, alongside others, as a most accurate approach (Figure 15O) but other clustering methods are also suitable. Based on these findings, we established the analysis framework according to the present invention (also termed “Interact-omics” framework, preferably flow cytometry-based), which also comprises a computational workflow for the quantification of cellular compositions and physical interactions of cells using flow cytometry (PICtR, “Physically Interacting Cell toolkit for R”, see Methods). Briefly, recorded flow cytometry datasets are preprocessed using standard pipelines without multiplet exclusion and are non-uniformly sampled to preserve rare cell types and cellular multiplets (sketching (Hao et al., 2023)), followed by clustering based on surface marker expression, scatter properties and FSC-ratio (see Methods). PIC-containing clusters, characterized by a high FSC-ratio and combinations of mutually exclusive cell type-specific markers, are selected and used for further downstream analysis, in-depth annotation and quantification. Otsu thresholding of the FSC ratio and Louvain clustering are usually applied as default settings, alternative approaches can also be selected (see Figures 15A, O). As non-limiting examples, cellular interaction frequencies using any of the following three distinct normalization approaches are further described: (1) the relative frequencies of cellular interactions among all live, high-quality events, which indicates how prevalent certain interactions are in relation to all cells and other interactions; (2) the relative frequencies of a given type of interaction among all interactions, providing insight into how the relative composition of cellular interactions changes across conditions; (3) in scenarios with unbalanced or rapidly changing frequencies of interacting partners, the harmonic mean can be used to calculate the expected interaction frequency based on singlet frequencies, which can then be compared to the observed interactions to assess relative enrichment (see Methods). Since these normalization methods address different biological questions, they are applied separately or in combination as appropriate. Compared to single-cell genomics-based workflows, the ultra-high cellular throughput, rapid processing time and low costs associated with the presented cytometry-based approach enable the seamless analysis of millions of cellular events within short time periods. The Interact-omics framework is designed to dissect heterotypic PICs and relies on carefully chosen case-control settings to determine an enrichment of true PICs above baseline interactions. Moreover, the experimental and cytometric settings, such as sample preparation, organ type, flow rate and cell densities should be kept constant as they may affect the formation and stability of cellular interactions (see Figure 7F, G). Ultra-high scale cellular interaction mapping across complex immune landscapes In order to simultaneously map cellular composition and cellular interactions in complex immune landscapes at high resolution, we established ultra-high parametric, data-informed flow cytometry assays for mouse and human. To optimize cell type resolution across all common blood and immune cell populations, we leveraged single- cell proteo-genomic datasets (Triana et al., 2021; Gayoso et al., 2021) to identify optimally discriminating cell type- and cell state-specific markers. Moreover, to enable the simultaneous detection of disparate cell type-specific markers in multiplets, we assigned cell type-specific markers to fluorophores with low spectral overlap to reduce spreading errors. While our approach can be applied to standard flow cytometry-based assays, we utilized full-spectrum flow cytometry (Bonilla et al., 2020) due to its superior capacity to disentangle high-plex marker panels. Applying the Interact-omics framework with such an optimized 24-plex panel to human PBMCs revealed an accurate representation of the CytoStimTM-induced changes in cellular composition and cellular interactions at cell type and cell state resolution (Figures 2A-E, Figures 8A-F). As expected, interactions between various T cell subsets and antigen presenting cell populations significantly increased upon CytoStimTMtreatment, whereas other cellular interactions remained unaffected or decreased (Figures 2D, E). Notably, the results were highly reproducible across replicates and interactions among rare populations could be accurately quantified, including multiple T cell subset and dendritic cell interactions. Next, to investigate whether the Interact-omics framework according to the present invention is capable of resolving antigen-dependent immune cell interactions, we isolated CD4 T cells carrying a transgenic T cell receptor specific for chicken ovalbumin (OVA) from OT-II mice and co-cultured them in the presence or absence of its cognate antigen with a complex cellular mixture of murine splenocytes (Figure 2F). As expected, cellular interactions between OVA- specific CD4 T cells and a range of antigen presenting cells were specifically induced in the presence of the respective antigen, whereas cellular interactions of bystander cells remained unaffected or changed only mildly (Figures 2G-J, Figures 9A-I). Together, these resultsdemonstrate the utility of our approach in resolving antigen-dependent and -independent cellularinteractions across complex immune landscapes with a broad range of potential applications. Performance of ex vivo cellular interaction mapping We conducted a series of ex vivo performance experiments. First, CytoStimTM-treated PBMCs were split, labeled with two distinct fluorescently conjugated CD45 antibodies, reunited, and processed under varying cell concentrations, processing times, and fixation methods (Figure 16A- G). Cellular interactions were then mapped using the Interact-omics approach. Interactions double- positive for both labels were newly acquired during the second incubation, while single-positive interactions could have occurred initially or throughout the culture. Our findings revealed a robust increase in single-positive interactions following CytoStimTMtreatment, whereas double-positive (newly acquired) interactions were negligible compared to thesurge in CytoStimTM induced single-positive interactions (approximately 5-10 times higher, Figure16G). Extending incubation periods post-CytoStimTMdid not elevate newly acquired interactions. These results show that ex vivo induced cellular interactions are primarily stable. The relativeeffect of CytoStimTM-induced single-positive interactions versus newly acquired double positiveinteractions remained consistent (Figure 16G). Fixation had only a negligible effect (Figure 16G). Collectively, these data provide quantitative insights into that experimental settings has no considerable impact on interactions while demonstrating that ex vivo modulations of cellular interactions can be effectively quantified in the explored settings. Dissecting the mechanism and kinetics of immunotherapies The molecular mode-of-action of most cancer immunotherapies is based on the redirection of cancer-immune cell interactions. For example, bispecific antibodies engage cancer cells with immune cells, whereas chimeric antigen receptor (CAR) T cells are engineered T cells that specifically target epitopes present on cancer cells. To investigate whether the Interact-omics framework according to the present invention is capable of resolving CAR-T cell-mediated cellular interactions, we utilized engineered GFP-tagged murine CAR-T cells targeting CD19-expressing cells in co-cultures with murine splenocytes (Figure 3A). Our analyses revealed that both CD4+ and CD8+ CAR-T cell subsets rapidly engaged in specific interactions with CD19-expressing target B cells (Figures 3B-D, Figures 10A-I). As a consequence, CAR-T cell interactions with B cells were highly enriched when compared to interactions between B cells and endogenous T cells (Figure 3E), reaching a maximum at one hour post treatment, followed by a gradual decline (Figures 10J-L). Bispecific antibodies engage T cells with tumor cells. Blinatumomab, which engages CD3-positive T cells with CD19-positive (malignant) B cell is a clinically approved immunotherapy. In order to investigate whether the Interact-omics framework is capable of resolving Blinatumomab-induced cellular interactions, we treated human PBMCs with Blinatumomab ex vivo (Figure 3F). Blinatumomab induced a strong increase in cellular interactions among a range of B and T cellpopulations, peaking one hour post treatment followed by a gradual decline of interactions overtime (Figures 3G-K, Figures 11A-J). As expected, the transient increase in cellular interactions of B cells was mirrored by a transient decrease of free single B cells and a time-delayed decrease in overall B cell-containing events, suggesting a rapid engagement of B and T cells, likely followed by a mild cytotoxic effect induced by Blinatumomab (Figure 3J, Figure 11K). In contrast to Blinatumomab- induced B-T cell interactions, interactions among other cell types remained unaffected, demonstrating the specificity of the mapped interactions (Figure 3K). Notably, following chemical fixation the quantification of cellular interactions induced by Blinatumomab remained unaffected by freeze-thawing associated cryopreservation, enabling a broad range of applications with primary patient material (Figure 7H). Together, these analyses demonstrate the broad utility of the Interact-omics framework to characterize cellular interactions induced by immunotherapies. Interact-omics reveals features underlying therapy response to Blinatumomab Blinatumomab has been approved for the treatment of B-cell acute lymphoblastic leukemia (B- ALL), the most common type of cancer in children, at relapsed or refractory stages (DQ Pediatric Treatment Editorial Board. Childhood acute lymphoblastic leukemia treatment (PDQ®): health professional version. in PDQ cancer information summaries (National Cancer Institute (US), 2002). Although Blinatumomab is progressing towards becoming the standard of care for relapsed and refractory pediatric ALL, the response rates remain heterogeneous (Queudeville et al., 2021; Queudeville and Ebinger et al., 2021). While few clinical and molecular parameters have been associated with outcome to Blinatumomab therapy, the underlying mechanisms remain poorly understood and a robust test predicting therapy response is lacking (Yin et al., 2018; Wei et al., 2021, Zhao et al., 2021; Klinger et al., 2012; Duell et al., 2017; Bocklein et al., 2019). To evaluate whether the Interact-omics framework according to the present invention can be used to extract parameters associated with therapy response, we acquired bone marrow aspirates from (Zhao et al., 2012) pediatric patients with relapsed B-ALL before Blinatumomab treatment. Subsequently, we applied the Interact-omics framework using an adjusted panel on the samples (i) one hour post ex vivo Blinatumomab treatment and (ii) in the absence of Blinatumomab (Figures 4A-C). We extracted a range of parameters from the data, including cellular frequencies of singlet populations in the absence of treatment and the induction of cellular interactions upon ex vivo Blinatumomab treatment (Figure 4B, C). In order to explore mechanisms underlying therapy response among patients with residual disease, we compared patients that could unequivocally be categorized into good responders (n = 18) and non- responders (n = 4) (Figure 4D). In line with previous studies (Wei et al., 2021; Zhao et al., 2021), high frequencies of various T cell subsets, particularly central and effector memory subsets, were associated with good response to Blinatumomab (Figures 4D, E). However, also the frequencies of cellular interactions prior to treatment or upon Blinatumomab were associated with therapy response (Figure 4D). For instance, Blinatumomab induced interactions of B and T cells, and B, T and myeloid cells more efficiently in good responders compared to non-responders (fold change good vs non-responders B*T(*My); Figures 4D, E, G). Interestingly, Blinatumomab failed to induce effective B*T cell interactions in patient samples with unbalanced T cell to B cell ratios (Figure 4F). Similarly, in patient samples with high T*myeloid interactions at baseline, Blinatumomab treatment failed to effectively induce B*T cell interactions, suggesting that T*myeloid interactions may inhibit or compete with B*T cell interactions. Accordingly, high T*myeloid interactions at baseline were associated with therapy failure (Figure 4D). Importantly, a correlation analysis of selected parameters associated with therapy response revealed that singlet frequencies of T cell subsets were highly correlated among each other, whereas cellular interactions provided independent and additive information on therapy response (Figure 4H, Figure 12). Jointly, these analyses provide new insights into the cellular mechanisms mediating response to Blinatumomab and may lay the foundation for personalized therapy response prediction. Intracellular signaling as consequence of cellular interactions To evaluate whether Interact-omics can be used to study intracellular signaling in response to cellular interactions, the inventors established a high-plex cytometry panel that includes an antibody detecting phosphorylation (pY142) of intracellular CD3 zeta (CD247), a transmembrane signaling adaptor protein phosphorylated upon T cell receptor signaling and T cell activation. Using this panel, we investigated intracellular TCR signaling in cellular interactions induced in human PBMCs upon CytoStim (crosslinks antigen-presenting cells with T cells) and Blinatumomab (crosslinks B cells with T cells) treatment (Figure 17A). Consistent with our previous results and the molecular mechanisms of the inducers used, we observed few background interactions at homeostasis but noted specific induction of B - T cell interactions upon Blinatumomab treatment and broader myeloid and B cell interactions with T cells upon CytoStim treatment (Figures 17B-D). As expected, CytoStim-induced interactions caused strong phosphorylation of the intracellular CD3 zeta domain in both T*B and T*Myeloid interactions, as well as in T*B*Myeloid triplets, demonstrating functional T cell receptor engagement in the interacting T cells (Figures 17E-F). In line with its more specific cross-linking activities, Blinatumomab caused a specific increase in phosphorylation of the intracellular CD3 zeta domain in T cells involved in interactions with B cells, but to a much lower degree in interactions not involving B cells. Collectively, these findings demonstrate that our approach can be used particularly efficiently to study intracellular signaling in response to cellular interactions. Virus-induced immune cell dynamics revealed by organism-wide cellular interaction mapping Infectious agents and pathogens induce complex cascades of organ-specific immune reactions in vivo, comprising cell-cell interactions, cell expansion and cellular trafficking, jointly establishing first line defense, long-lasting adaptive immunity and hematopoietic recovery after pathogeninsult. However, our comprehension of such pathogen-induced cellular immune dynamicsremains limited due to current technological restrictions. In particular, there is a lack of quantitative insights into organotypic differences in the composition, order and kinetics of cellular interactions induced upon pathogen exposure in vivo. The lymphocytic choriomeningitis virus (LCMV) serves as a well-established murine model pathogen to study key questions in immunology, including the induction of innate and adaptive immunity, pathologic consequences of virus infections, immune evasion mechanisms and virus-induced suppression of hematopoiesis (Zinkernagel et al., 2002; Oldstone et al., 2002). To systematically unravel LCMV-induced alterations in the immune cell and cellular interaction networks across distinct organ systems, we applied the Interact-omics workflow according to the present invention to mesenteric lymph nodes (LNs), spleens and bone marrow (BM) of mice at day 0 (naive), 3 and 7 post intraperitoneal LCMV infection in quadruplicates each (Figure 5A). In order to discriminate cellular interactions mediated by antigen-dependent and -independent mechanisms, we transferred congenic, LCMV- specific CD4 and CD8 T cells recognizing epitopes of the LCMV glycoprotein into mice 5 days prior to infection (Figure 5A), and included congenic markers (SMARTA:CD90.1; P14:CD45.1) in our cytometry panel (see methods). In total, we quantified more than 34 million single cells from 21 cell types, and around 415,000 cellular interactions from 52 cell type pairs, across 36 samples (Figures 5B, C, D, E, Figure 13, Figure 14). LCMV infection caused a wide range of alterations in cellular composition and cellular interactions. Principal component analysis (PCA) of cellular abundances and interactions revealed organ- and time-specific changes that were highly reproducible across replicates, demonstrating the robustness of our approach (Figures 5D, E). Clustering cellular interactions according to their virus-induced alterations over time revealed groups with distinct patterns of interaction dynamics (Figure 6A). For example, cellular interactions in cluster 4 were rapidly induced at day 3 and partially normalized towards day 7 postinfection. Cellular interactions in this cluster comprised mainly cell types of the innate arm of theimmune system (e.g. NK cells, monocytes, macrophages), in line with their rapid response and key role in first line defense, as well as few non-antigen-specific adaptive immune cells. In contrast, clusters 3 and 5 contained a variety of interactions comprising LCMV-specific T cells, which displayed a delayed but pronounced induction of cellular abundances and interactions at day 7, in line with their well-documented response pattern (Figures 6A, B). Notably, LCMV- specific T cell interactions were more pronounced in spleen when compared to mesenteric LNs (Figure 6C), likely reflecting a more rapid uptake of LCMV into the spleen after intraperitoneal administration as previously described (Olson et al., 2012). Interestingly, LCMV-specific T cells were also detected in the BM (Figure 6B), in line with the notion that BM may serve as primary immune organ (Feuerer et al., 2003). However, LCMV-specific BM T cells were less likely to engage in cellular interactions compared to their non-LCMV-specific T cell counterparts, as indicated by a negative odds ratio, taking their singlet frequencies into account (Figure 6D). In contrast, LCMV-specific T cells in LN and spleen were more likely to engage in cellular interactions when compared to their non- LCMV specific counterparts, in line with the key role of LNs and spleen in the orchestration of adaptive immune responses (Figure 6D). In LCMV infections, BM myelosuppression is associated with a transient activation of NK cells, peaking at day 3 post infection, followed by a rapid recovery (Thomsen et al., 1986; Binde et al., 1997). In line with this, we observed a massive increase in NK cell interactions with cells of the myeloid lineage, including myeloid progenitors, peaking at day 3 post infection in the BM (Figure 6E). Subsequently, NK cell-myeloid interactions decreased, followed by an expansion of HSPCs and BM monocytes at day 7 (Figures 6E), suggesting a switch from myelosuppression to active emergency hematopoiesis, in line with previously reported kinetics of LCMV-induced myelopoiesis (Schürch et al., 2014). Notably, upon infection, a rapid infiltration of monocytes into LNs and spleens was observed (Figure 6F). Recruited monocytes readily engaged with LN and spleen B cells at day 3, partially normalizing at day 7 (Figure 6G). Importantly, such extensive monocyte - B cell interactions have recently been described to serve as an LCMV- specific immune evasion mechanism, hindering early B cell responses in a chronic model of LCMV infection (Sammicheli et al., 2016). In line with this, increased interactions between plasmablasts and LCMV-specific CD4 T cells in LN and spleen, as well as an expansion of plasma cells coincided with the disappearance of suppressive monocyte*B cell interactions at day 7 (Figures 6H, I). Together, these results demonstrate the utility of the Interact-omics approach of the present invention for dissecting complex immune interaction networks in vivo. Our data accurately recapitulate previous findings and provide a first quantitative framework for a systems-level understanding of virus-induced alterations of the cellular immune interaction networks and how they cooperate across organ systems to elicit intricate immune responses. Performance and interpretation of in vivo-derived interactions To assess the general reliability of interactions derived from in vivo settings, we performed ImageStream-based imaging flow cytometry on LCMV-infected spleens on day 7 post-infection. Despite its limitations in cellular throughput and number of measured markers, this method provides morphological information that can be used to distinguish single cells from interacting cells, making it a suitable benchmarking tool. Using a 6-plex panel focused on T and B cell interactions, we first applied the PICtR workflow on the ImageStream data without taking any morphological information into consideration (Figures 18A-C). Here, clustering and interaction identification relied solely on fluorescence intensity values as no forward scatter information is recorded with this method. We then compared the results from the Interact-omics workflow with data gained from both morphological imaging and / or fluorescence intensity data in order to: (i) differentiate singlets from interacting cells, (ii) identify interacting cell types, and (iii) assess LCMV-induced interaction changes. Comparing the results obtained from the Interact-omics- based approach to image segmentation-based classification showed high concordance in singlet and multiplet discrimination (Figure 18D, see Methods), further verified by manual inspection of images (Figure 18C). Immunophenotypic characterization of interacting populations via conventional gating also showed high agreement with Interact-omics-derived annotations (Figures 18E, G). Manual inspection of randomly selected images regarding the localized expression patterns of lineage-specific markers confirmed the expected types of interactions (Figure 18C). Finally, comparing LCMV-induced changes in cellular interactions derived from Interact-omics and conventional gating of imaging flow cytometry data, showed a high concordance (Figure 18F). These findings confirm the Interact-omics approach's accuracy in identifying and quantifying single and interacting cell landscapes in case-control settings. The presented framework measures cellular interactions following sample preparation ex vivo. Consequently, for in vivo applications, additional cellular interactions may be acquired during sample preparation. While this limitation applies to all cellular interaction mapping approaches that do not rely on specialized mouse models or measure co-localization in situ, it remains poorly characterized to what extent this occurs, whether newly acquired interactions are random or directed, and how representative the identified interactions are of the in vivo situation. To evaluate these questions, we utilized congenic mouse models differing in variants of the pan- hematopoietic cell marker CD45, allowing identification of respective immune cells as CD45.1 or CD45.2 using variant-sensitive antibodies (Figure 19A). First, we transferred LCMV-specific CD4 T cells (SMARTA: CD90.1-positive, CD45.2-positive) into CD45.2 mice, followed by LCMV infection (group A, Figure 19A). Non-infected control CD45.2 mice formed group B. In 821 parallel, we infected CD45.1 mice with LCMV (group C) or left them untreated (group D). On day 7 post-infection, spleens from group A (infected, CD45.2) and group B (non-infected, CD45.2) were either processed individually or mixed with spleens from group C (infected, CD45.1) or group D (non-infected, CD45.1) before tissue homogenization and processing. Applying the Interact-omics workflow to these individual and mixed samples resulted in single-cell and interacting cell landscapes of populations that were either single-positive or double-positive for CD45.1 and CD45.2 (Figures 19B-E). In this setting, double-positive interactions that must have arisen during processing ex vivo can be quantified. Notably, we observed a substantial fraction of newly acquired double-positive interactions during sample processing (Figure 19F). However, newly acquired interactions did not occur randomly but were highly correlated with interactions induced upon infection (Figure 19G). In particular, newly acquired interactions in mixed spleens from infected mice compared to non-infected controls were highly correlated with infection-induced single positive interactions in both mixed and non-mixed spleens (Figure 19G). This suggests that while new interactions can be acquired during sample preparation, they are not random but directed and reflect actual biological effects. Notably, a comparison with imaging and in situ interaction mapping in an LCMV infection study revealed highly similar interaction types and confirmed key LCMV-induced changes identified by the Interact-omics approach, including antigen-specific T cell interactions and transient monocyte- B cell interactions. Overall, these observations suggest that the measured interactions in the Interact-omics approach have all occurred in vivo, the interactions reflect biological effects, and are a proxy for cellular interactions occurring in vivo. Application to existing cytometry datasets To demonstrate the applicability of our approach for analyzing cellular interactions in previously generated datasets, we applied the PICtR workflow to a publicly available cytometry dataset on juvenile idiopathic arthritis (JIA54). JIA is an autoimmune disease characterized by chronic joint inflammation, leading to pain, swelling, and eventual joint damage. While it is hypothesized that abnormal interactions among immune cells – specifically T cells, B cells, and myeloid cells – contribute to the production of inflammatory cytokines and autoantibodies that drive the disease, the precise interaction processes remain poorly understood. In their study, Attrill and colleagues interrogated PBMC samples from healthy donors, JIA patients with active and inactive disease, and synovial fluid samples from JIA patients with active disease. We downloaded and preprocessed the FCS files as described in the Methods section and ran the FlowAI QC algorithm on all FCS files to exclude those with anomalous flow rates from further analysis. High-quality samples were then processed using the PICtR workflow. Notably, we were able to reproduce the single-cell landscape described by Attrill et al., and discovered a range of quantitative and qualitative changes in cellular interactions in the blood of patients with inactive versus active disease, as well as between the blood and synovial fluid of affected joints (Figure 20). Interestingly, in patients with inactive disease, T cells interacting with B cells, predominantly displayed a FoxP3-expressing regulatory T cell phenotype (Figure 20I). In contrast, in patients with active disease, these interactions shifted to an inflammatory, non-regulatory phenotype. Similarly, major qualitative differences of interactions between CD4 T cells and monocytes were observed between blood and synovial fluid of patients with active disease (Figure 20L). These findings provide an initial quantitative framework for understanding changes in immune cell interactions that may contribute to disease progression and could help identify targets for therapeutic intervention. Moreover, these analyses demonstrate that our approach can also be applied to existing datasets. Identifying cellular interactions between non-immune cells: To demonstrate the present invention’s ability to identify cellular interactions involving non- immune cells, we applied the workflow described herein to a spectral flow cytometry dataset (described in Funk, M. C. et al. Aged intestinal stem cells propagate cell-intrinsic sources of inflammaging in mice. Dev. Cell 58, 2914-2929.e7; 2023), from the proximal intestine of young and old mice (Figure 21A). The inventive approach identified interactions among EpCAM+ epithelial cells, Lgr5+ intestinal stem cells (ISCs), and various immune cell types (Figure 21B, C). In line with previous reports, our analysis confirmed an age-associated increase in CD4 and CD4- CD8 double-positive T cells, a decrease in CD8+TCRαβ+ / TCRγδ+ populations, and a notable increase in interactions between CD4+TCRαβ+T cells and epithelial cells in the aged group (Figure 21D, E). The latter supports the reported upregulation of the MHCII machinery in the aged intestine, suggesting enhanced antigen-presentation to CD4 T cells. These analyses demonstrate that the inventive approach can also be applied to existing datasets and non-immune cell types. REFERENCES Armingol, E., Officer, A., Harismendy, O. & Lewis, N. E. Deciphering cell-cell interactions and communication from gene expression. Nat. Rev. Genet.22, 71–88 (2021). Armingol, E., Baghdassarian, H. M. & Lewis, N. E. The diversification ofmethods for studying cell- cell interactions and communication. Nat. Rev. Genet. (2024) doi:10.1038 / s41576-023-00685-8. Cooper, G. M. Cell-Cell Interactions. in The Cell: A Molecular Approach. (ed. Sunderland) (2000). Bechtel, T. J., Reyes-Robles, T., Fadeyi, O. O. & Oslund, R. C. Strategies for monitoring cell-cell interactions. Nat. Chem. Biol.17, 641–652 (2021). Yu, J., Peng, J. & Chi, H. Systems immunology: Integrating multi-omics data to infer regulatory networks and hidden drivers of immunity. Current Opinion in Systems Biology 15, 19–29 (2019). Chen, K. H., Boettiger, A. N., Moffitt, J. R., Wang, S. & Zhuang, X. Spatially resolved, highly multiplexed RNA profiling in single cells. Science 348, aaa6090 (2015). Borm, L. E. et al. Scalable in situ single-cell profiling by electrophoretic capture of mRNA using EEL FISH. Nat. Biotechnol.41, 222–231 (2023). Shah, S., Lubeck, E., Zhou, W. & Cai, L. In situ transcription profiling of single cells reveals spatial organization of cells in the mouse hippocampus. Neuron 92, 342–357 (2016). Gyllborg, D. et al. Hybridization-based in situ sequencing (HybISS) for spatially resolved transcriptomics in human and mouse brain tissue. Nucleic Acids Res.48, e112 (2020). Biancalani, T. et al. Deep learning and alignment of spatially resolved single- cell transcriptomes with Tangram. Nat. Methods 18, 1352–1362 (2021). Sountoulidis, A. et al. SCRINSHOT enables spatial mapping of cell states in tissue sections with single-cell resolution. PLoS Biol.18, e3000675 (2020). Rodriques, S. G. et al. Slide-seq: A scalable technology for measuring genome- wide expression at high spatial resolution. Science 363, 1463–1467 (2019). Stickels, R. R. et al. Highly sensitive spatial transcriptomics at near-cellular resolution with Slide- seqV2. Nat. Biotechnol.39, 313–319 (2021). Ståhl, P. L. et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science 353, 78–82 (2016). Goltsev, Y. et al. Deep Profiling of Mouse Splenic Architecture with CODEX Multiplexed Imaging. Cell 174, 968-981.e15 (2018). Viratham Pulsawatdi, A. et al. A robust multiplex immunofluorescence and digital pathology workflow for the characterisation of the tumour immune microenvironment. Mol. Oncol.14, 2384– 2402 (2020). Baccin, C. et al. Combined single-cell and spatial transcriptomics reveal the molecular, cellular and spatial bone marrow niche organization. Nat. Cell Biol.22, 38–48 (2020). García-Guerrero, E. et al. Selection of Tumor-Specific Cytotoxic T Lymphocytes in Acute Myeloid Leukemia Patients Through the Identification of T-Cells Capable to Establish Stable Interactions With the Leukemic Cells: “Doublet Technology”. Front. Immunol.9, 1971 (2018). Andrews, N. et al. An unsupervised method for physical cell interaction profiling of complex tissues. Nat. Methods 18, 912–920 (2021). Wheeler, M. A. et al. Droplet-based forward genetic screening of astrocyte- microglia cross-talk. Science 379, 1023–1030 (2023). Giladi, A. et al. Dissecting cellular crosstalk by sequencing physically interacting cells. Nat. Biotechnol.38, 629–637 (2020). Halpern, K. B. et al. Paired-cell sequencing enables spatial gene expression mapping of liver endothelial cells. Nat. Biotechnol.36, 962–970 (2018). Manco, R. et al. Clump sequencing exposes the spatial expression programs of intestinal secretory cells. Nat. Commun.12, 3074 (2021). Boisset, J.-C. et al. Mapping the physical network of cellular interactions. Nat. Methods 15, 547– 553 (2018). Pasqual, G. et al. Monitoring T cell-dendritic cell interactions in vivo by intercellular enzymatic labelling. Nature 553, 496–500 (2018). Nakandakari-Higa, S. et al. Universal recording of cell-cell contacts in vivo for interaction-based transcriptomics. BioRxiv (2023) doi:10.1101 / 2023.03.16.533003. Zhang, S. et al. Monitoring of cell-cell communication and contact history in mammals. Science 378, eabo5503 (2022). Medaglia, C. et al. Spatial reconstruction of immune niches by combining photoactivatable reporters and scRNA-seq. Science 358, 1622–1626 (2017). Schraivogel, D. et al. High-speed fluorescence image-enabled cell sorting. Science 375, 315–320 (2022). Hao, Y. et al. Dictionary learning for integrative, multimodal and scalable single- cell analysis. Nat. Biotechnol. (2023) doi:10.1038 / s41587-023-01767-y. Lenth, R. V. emmeans: Estimated Marginal Means, aka Least-Squares Means. (CRAN, 2022). Scrucca, L., Fraley, C., Murphy, T. B. & Adrian E., R. Model-Based Clustering, Classification, and Density Estimation Using mclust in R. (Chapman and Hall / CRC, 2023). doi:10.1201 / 9781003277965. Triana, S. et al. Single-cell proteo-genomic reference maps of the hematopoietic system enable the purification and massive profiling of precisely defined cell states. Nat. Immunol.22, 1577– 1589 (2021). Gayoso, A. et al. Joint probabilistic modeling of single-cell multi-omic data with totalVI. Nat. Methods 18, 272–282 (2021). Bonilla, D. L., Reinin, G. & Chua, E. Full spectrum flow cytometry as a powerful technology for cancer immunotherapy research. Front. Mol. Biosci.7, 612801 (2020). Diggins, K. E., Greenplate, A. R., Leelatian, N., Wogsland, C. E. & Irish, J. M.Characterizing cell subsets using marker enrichment modeling. Nat. Methods 14, 275–278 (2017). PDQ Pediatric Treatment Editorial Board. Childhood acute lymphoblastic leukemia treatment (PDQ®): health professional version. in PDQ cancer information summaries (National Cancer Institute (US), 2002). Queudeville, M. et al. Blinatumomab in pediatric patients with relapsed / refractory B-cell precursor acute lymphoblastic leukemia. Eur. J. Haematol.106, 473–483 (2021). Queudeville, M. & Ebinger, M. Blinatumomab in Pediatric Acute Lymphoblastic Leukemia-From Salvage to First Line Therapy (A Systematic Review). J. Clin. Med.10, (2021). Yin, H. et al. Patient’s T Cell Functionality Determines Blinatumomab-Mediated Killing of B-ALL Leukemia Cells. Blood 132, 4079–4079 (2018). Wei, A. H. et al. Biomarkers associated with blinatumomab outcomes in acute lymphoblastic leukemia. Leukemia 35, 2220–2231 (2021). Zhao, Y. et al. Tumor-intrinsic and -extrinsic determinants of response to blinatumomab in adults with B-ALL. Blood 137, 471–484 (2021). Klinger, M. et al. Immunopharmacologic response of patients with B-lineage acute lymphoblastic leukemia to continuous infusion of T cell-engaging CD19 / CD3- bispecific BiTE antibody blinatumomab. Blood 119, 6226–6233 (2012). Duell, J. et al. Frequency of regulatory T cells determines the outcome of the T-cell-engaging antibody blinatumomab in patients with B-precursor ALL. Leukemia 31, 2181–2190 (2017). Bücklein, V. et al. Predictors of Efficacy for Blinatumomab in BCP-ALL Patients: Non-Responders Show Impaired CD19-BiTE®-Mediated Cytotoxicity in Vitro. Blood 134, 2632–2632 (2019). Zinkernagel, R. M. Lymphocytic choriomeningitis virus and immunology. Curr. Top. Microbiol. Immunol.263, 1–5 (2002). Oldstone, M. B. A. Biology and pathogenesis of lymphocytic choriomeningitis virus infection. Curr. Top. Microbiol. Immunol.263, 83–117 (2002). Olson, M. R., McDermott, D. S. & Varga, S. M. The initial draining lymph node primes the bulk of the CD8 T cell response and influences memory T cell trafficking after a systemic viral infection. PLoS Pathog.8, e1003054 (2012). Feuerer, M. et al. Bone marrow as a priming site for T-cell responses to blood- borne antigen. Nat. Med.9, 1151–1157 (2003). Thomsen, A. R., Pisa, P., Bro-Jørgensen, K. & Kiessling, R. Mechanisms of lymphocytic choriomeningitis virus-induced hemopoietic dysfunction. J. Virol.59, 428–433 (1986). Binder, D., Fehr, J., Hengartner, H. & Zinkernagel, R. M. Virus-induced transient bone marrow aplasia: major role of interferon-alpha / beta during acute infection with the noncytopathic lymphocytic choriomeningitis virus. J. Exp. Med.185, 517–530 (1997). Schürch, C. M., Riether, C. & Ochsenbein, A. F. Cytotoxic CD8+ T cells stimulate hematopoietic progenitors by promoting cytokine release from bone marrow mesenchymal stromal cells. Cell Stem Cell 14, 460–472 (2014). Sammicheli, S. et al. Inflammatory monocytes hinder antiviral B cell responses. Sci. Immunol.1, (2016). Pircher, H., Bürki, K., Lang, R., Hengartner, H. & Zinkernagel, R. M. Tolerance induction in double specific T-cell receptor transgenic mice varies with antigen. Nature 342, 559–561 (1989). Oxenius, A., Bachmann, M. F., Zinkernagel, R. M. & Hengartner, H. Virus- specific MHC-class II- restricted TCR-transgenic mice: effects on humoral and cellular immune responses after viral infection. Eur. J. Immunol.28, 390–400 (1998). Gottschlich, A. et al. Single-cell transcriptomic atlas-guided development of CAR-T cells for the treatment of acute myeloid leukemia. Nat. Biotechnol.41, 1618– 1632 (2023). Hataye, J., Moon, J. J., Khoruts, A., Reilly, C. & Jenkins, M. K. Naive and memory CD4+ T cell survival controlled by clonal abundance. Science 312, 114–116 (2006). Zajac, A. J. et al. Viral immune evasion due to persistence of activated T cells without effector function. J. Exp. Med.188, 2205–2213 (1998). Hahne, F. et al. flowCore: a Bioconductor package for high throughput flow cytometry. BMC Bioinformatics 10, 106 (2009). Finak, G., Jiang, W. & Gottardo, R. CytoML for cross-platform cytometry data sharing. Cytometry A 93, 1189–1196 (2018). Greg Finak, M. J. flowWorkspace. Bioconductor (2017) doi:10.18129 / b9.bioc.flowworkspace. R Foundation for Statistical Computing, R. C. T. R: A Language and Environment for Statistical Computing. (2021). Kuhn, M. Building Predictive Models in R Using the caret Package. J. Stat. Softw.28, (2008). Parks, D. R., Roederer, M. & Moore, W. A. A new “Logicle” display method avoids deceptive effects of logarithmic scaling for low signals and compensated data. Cytometry A 69, 541–551 (2006). Parks, B. BPCells: Single Cell Counts Matrices to PCA. (2023). Ashhurst, T. M. et al. Integration, exploration, and analysis of high-dimensional single-cell cytometry data using Spectre. Cytometry A 101, 237–253 (2022). Venables, W. N. & Ripley, B. D. Modern Applied Statistics with S. (Springer, 2002). Otsu, N. A Threshold Selection Method from Gray-Level Histograms. IEEE Trans. Syst. Man Cybern.9, 62–66 (1979). Waldman, A. D., Fritz, J. M. & Lenardo, M. J. A guide to cancer immunotherapy: from T cell basic science to clinical practice. Nature Reviews Immunology vol.20651–668 (2020). Schaft, N. et al. The future of affordable cancer immunotherapy. Frontiers in Immunology vol.14 1248867 (2023). Pilard, C. et al. Cancer immunotherapy: it’s time to better predict patients’ response. British Journal of Cancer vol.125927–938 (2021). Schiepers, A., van ’t Wout, M.F.L., Greaney, A.J. et al. Molecular fate -mapping of serum antibody responses to repeat immunization. Nature 615, 482–489 (2023). Monaco, G. et al. flowAI: automatic and interactive anomaly discerning tools for flow cytometry data. Bioinformatics 32, 2473–2480 (2016). Landini, G., Randell, D. A., Fouad, S. & Galton, A. Automatic thresholding from the gradients of region boundaries. J. Microsc.265, 185–195 (2017). van der Walt, S. et al. scikit-image: image processing in Python. PeerJ 2, e453 (2014). Stringer, C., Wang, T., Michaelos, M. & Pachitariu, M. Cellpose: a generalist algorithm for cellular segmentation. Nat. Methods 18, 100–106 (2021). McInnes, L., Healy, J., Saul, N. & Großberger, L. UMAP: uniform manifold approximation and projection. JOSS 3, 861 (2018). Hornik, K. A CLUE for cluster ensembles. J. Stat. Softw.14, (2005). Otsu, N. A Threshold Selection Method from Gray-Level Histograms. IEEE Trans. Syst. Man Cybern.9, 62–66 (1979). Therneau, T. & Atkinson, B. rpart: Recursive Partitioning and Regression Trees. The R Foundation (2024) doi:10.32614 / cran.package.rpart. Attrill, M. H. et al. The immune landscape of the inflamed joint defined by spectral flow cytometry. Clin. Exp. Immunol. (2024) doi:10.1093 / cei / uxae071.

Claims

CLAIMS 1. A method for determining physically interacting cells in a cell population, comprising a. providing a flow-cytometry data set of said cell population, said data set comprising data on scattering properties and expression of cell-type specific surface markers for the cells in said cell population, wherein said scattering properties comprise at least one measurement of forward and / or side scatter, b. applying an automated clustering algorithm to said flow-cytometry data set, thereby assigning the cells to a cluster according to a degree of similarity of the scattering properties and the expression of cell-type specific surface markers, c. wherein the expression of one or more cell-type specific surface marker(s) and / or combinations of cell-type specific surface markers in a cluster is indicative of a specific cell type(s) in the cluster, and wherein the scattering properties of a cluster are indicative of the presence of a single cell or two or more physically interacting cells in the cluster.

2. The method according to any one of the preceding claims, wherein the expression of one or more cell-type specific surface marker(s) and / or combinations of cell-type specific surface markers in a cluster is indicative of the number and / or frequency of one or more specific cell type(s) and / or cell state(s) in the cluster.

3. The method according to any one of the preceding claims, wherein the flow-cytometry data set comprises data on the expression of at least 5 cell-type specific surface markers (preferably at least 10 cell-type specific surface markers, more preferably at least 25 cell- type specific surface markers).

4. The method according to any one of the preceding claims, wherein the expression of at least two distinct cell-type specific surface markers and / or combinations of cell-type specific surface markers in a cluster, said cluster having a scattering property above a threshold value, is indicative of the presence of at least two physically interacting cells of different cell types and / or cell states in the cluster.

5. The method according to any one of the preceding claims, wherein said scattering properties comprise forward scatter area (FSC-A), forward scatter width (FSC-W), side scatter width (SSC-W) and / or forward scatter height (FSC-H).

6. The method according to any one of the preceding claims, wherein the scattering properties comprise a ratio of the signal intensity of FSC-A and FSC-H (FSC-ratio).

7. The method according to any one of the preceding claims, wherein the automated clustering algorithm is an unsupervised or semi-supervised clustering algorithm.

8. The method according to any one of the preceding claims, wherein the automated clustering algorithm comprises performing a hierarchical, non-hierarchical and / or graph- based clustering, preferably Louvain clustering, Leiden clustering, FlowSOM clustering, Phenograph clustering and / or K-means clustering.

9. The method according to any one of the preceding claims, wherein the scattering properties comprise the FSC-ratio and a threshold value for the FSC-ratio is determined by an automated thresholding technique, preferably by Otsu thresholding.

10. The method according to any one of the preceding claims, additionally comprising a. a non-uniform sampling of said data set prior to applying the clustering algorithm, preferably by applying a sampling algorithm based on dictionary learning, b. spectral compensation, transformation and scaling of said data set prior to performing non-uniform sampling, and / or c. comprising adjusting the clustered data for a number of free interacting partners, preferably comprising calculation of the expected rate of physically interacting cells within the cell population by calculating the harmonic mean of the rates of the cell types present in said cell population.

11. The method according to any one of the preceding claims, wherein providing the flow cytometry data set comprises a. providing a sample of a subject comprising a cell population, b. preparing a cell suspension from the sample, c. addition of multiple labels (preferably fluorophore-coupled binding reagents), each binding a distinct cell-type specific surface marker (preferably adding at least 15, more preferably at least 20 labels for distinct cell-type specific surface markers), and d. determining scattering properties of the cells and intensities of said labels by flow cytometry, wherein the fluorescence intensity of said labels is indicative of the expression of surface markers of said cell population.

12. The method according to any one of the preceding claims, wherein the physical interaction of the cells comprises a. an antigen-antibody and / or receptor-ligand interaction, b. an immunotherapy-induced interaction, preferably induced by a CAR-T cell, an antibody or a bi-specific antibody, and / or c. an infection-induced interaction, preferably induced by a viral infection.

13. The method according to any one of the preceding claims, wherein the assignment of the cells to clusters according to the scattering properties and the expression of cell-type specific surface markers represents a state of the cell population (“cellular interaction landscape”).

14. The method according to the preceding claim, wherein the method is performed on multiple flow-cytometry data sets obtained from cell populations of multiple subjects, wherein each subject has received a (medical or other) treatment, and wherein the method comprises comparing the states of each cell population.

15. The method according to claim 13 or 14, wherein the state of a cell population is indicative of therapeutic success of a medical treatment, the medical treatment preferably comprising a. administering a bispecific binding agent (e.g., antibody), and the composition and state of a cell population is indicative of therapeutic success of said medical treatment, and / or b. administering a therapeutic immune cell (e.g., a CAR-T cell), and the composition and state of a cell population is indicative of therapeutic success of said medical treatment.

Images