A robust classifier for identifying subtypes of sepsis

Abstract

The present disclosure provides gene expression-based methods for determining whether a subject with sepsis has an inflammatory, adaptive, or coagulopathic phenotype. Kits for performing the methods are also provided.

Description

Technical Field

[0001] Cross-reference This application claims the benefit of U.S. Provisional Application No. 62 / 636,096, filed Feb. 27, 2018, the entire disclosure of which is incorporated herein by reference.

[0002] Government Rights This invention was made with government support under Contract No. AI057229 and AI109662 awarded by the National Institutes of Health. The government has certain rights in this invention.

Background Art

[0003] Sepsis is defined as life-threatening organ dysfunction caused by dysregulation of the immune response to infection (1). Despite its association with nearly half of all in-hospital deaths, there is still no approved treatment specific to sepsis (2, 3). This is partly because the clinical syndrome of sepsis includes a significant degree of heterogeneity and may encompass many different subtypes similar to those established among cancer patients (4, 5). Current classification of sepsis is based on clinical criteria such as the presence of shock, source of infection, or organ failure, but such classification may not represent the biological drivers of the host response. These also fail to appropriately match patients to novel interventions. If the heterogeneity of sepsis truly reflects heterogeneity of the host response, characterization of the types of host responses underlying this is fundamental to enabling precision sepsis treatment (6).

[0004] In unsupervised analysis, data are classified into subgroups ("clusters") that are defined only internally, without referring to external "supervised" outcomes such as mortality or severity. Instead, structures inherent within the data are used to define the subgroups. Such data-driven analyses have been successful in defining validated, clinically relevant disease subtypes in multiple diseases (4, 5, 7, 8). Because whole blood gene expression reflects the temporal state of circulating leukocytes, at least two academic groups have applied unsupervised clustering to whole blood transcriptome profiles of sepsis patients to study "host responses" in a data-driven framework (9-13). Their results identified subtypes with higher mortality accompanied by evidence of immune exhaustion and reduced glucocorticoid receptor signaling, as well as subtypes with lower mortality accompanied by conventional pro-inflammatory signaling (9-13).

[0005] Clustering analysis often yields non-reproducible results for one of two reasons: either multiple arbitrary choices are used in the methodology so that minor changes to the analysis yield new results, or the clustered dataset is too small to represent the broad heterogeneity of the disease. However, recent advances in meta-clustering and data pooling can help address both problems (14–16). Coupled with an unprecedented amount of publicly available transcriptome data on sepsis (17, 18), the hypothesis that robustly reproducible host response subtypes (clusters) of sepsis exist across the broad and heterogeneous spectrum of clinical sepsis was tested. [Overview of the project]

[0006] Based on transcriptome data, subjects with sepsis can be assigned to one of three clusters: the "inflammatory disorder" cluster associated with high innate immune signaling / reduced adaptive immune signaling, the "adaptive" cluster associated with reduced innate immune signaling / high adaptive immune signaling and low mortality, and the "coagulation disorder" cluster exhibiting both clinical and molecular irregularities in the coagulation and complement systems.

[0007] In some embodiments, a method is provided for determining whether a subject with sepsis has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype. In these embodiments, the method (a) measures the amount of RNA transcripts encoded by at least two of ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB in an RNA sample obtained from the subject in order to obtain gene expression data, (b) Provide a report indicating, based on gene expression data, whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype. (i) Increases in ARG1, LCN2, LTF, and / or OLFM4, and / or decreases in HLA-DMB, indicate that the subject has an inflammatory phenotype. (ii) Increases in YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and / or FRS2, and / or decreases in GADD45A, CD24, S100A12, and / or STX1A indicate that the subject has an adaptive phenotype, and (iii) Increases in KCNMB4, CRISP2, HTRA1, and / or PPL, and / or decreases in RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and / or RELB, indicate that the subject has a coagulation disorder phenotype.

[0008] In some embodiments, a method is provided for treating a subject having sepsis. In these embodiments, the method is (a) Receiving a report indicating whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation disorder phenotype, the report being based on gene expression data obtained by measuring the amount of RNA transcripts encoded by at least two of the following in RNA samples obtained from the subject: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB. (i) Increases in ARG1, LCN2, LTF, and / or OLFM4, and / or decreases in HLA-DMB, indicate that the subject has an inflammatory phenotype. (ii) Increases in YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and / or FRS2, and / or decreases in GADD45A, CD24, S100A12, and / or STX1A indicate that the subject has an adaptive phenotype, and (iii) Increases in KCNMB4, CRISP2, HTRA1, and / or PPL, and / or decreases in RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and / or RELB, indicate that the subject has a coagulation disorder phenotype, and the report is received. (b) This may include treating the subject based on whether the subject is indicated to have an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype.

[0009] A kit for performing this method is also provided.

[0010] The present invention will be best understood from the following detailed description when read in conjunction with the accompanying drawings. Following common practice, it should be emphasized that various features in the drawings are not to scale. Rather, the dimensions of various features are arbitrarily enlarged or reduced for clarity. The drawings include the following figures. [Brief explanation of the drawing]

[0011] [Figure 1] This outlines the overall research. [Figure 2A-B] Using all 8,946 genes present in the COCONUT conormalized data, we display the first two principal components (PCs) of the exploratory clustering results (both (A) with and (B) without, representing 16% of samples that were not clustered in the final analysis, shown in gold). Here, as demonstrated by the first two principal components, we see that the cluster assignments reconstructed using the unsupervised method are clearly separated in high-dimensional space. [Figure 3A]This section displays the correlation between the average 500 gene expression vectors across assigned clusters in the exploration and validation datasets and heatmaps of gene ontology (GO) codes that were found to be overexpressed in different clusters. (A) Correlation between the average 500 gene expression vectors across assigned clusters in the exploration and validation datasets, with correlation coefficients indicated by color (legend on the right side of the figure). In particular, samples from the inflammatory dysfunction cluster are positively correlated with inflammatory dysfunction samples from other datasets and negatively correlated with adaptive samples from other datasets (and vice versa). The coagulation dysfunction clusters are positively correlated with each other, although with low cohesiveness. (B) Heatmaps of gene ontology (GO) codes were found to be overexpressed in different clusters, color-coded at a significant level. In both (A) and (B), the pooled “core” exploration dataset is represented by a single column for each cluster, while each cluster in each validation dataset is represented by a separate column. Both subfigures show block structures that indicate molecular similarity across datasets between clusters of the same type. [Figure 3B] This section displays the correlation between the average 500 gene expression vectors across assigned clusters in the exploration and validation datasets and heatmaps of gene ontology (GO) codes that were found to be overexpressed in different clusters. (A) Correlation between the average 500 gene expression vectors across assigned clusters in the exploration and validation datasets, with correlation coefficients indicated by color (legend on the right side of the figure). In particular, samples from the inflammatory dysfunction cluster are positively correlated with inflammatory dysfunction samples from other datasets and negatively correlated with adaptive samples from other datasets (and vice versa). The coagulation dysfunction clusters are positively correlated with each other, although with low cohesiveness. (B) Heatmaps of gene ontology (GO) codes were found to be overexpressed in different clusters, color-coded at a significant level. In both (A) and (B), the pooled “core” exploration dataset is represented by a single column for each cluster, while each cluster in each validation dataset is represented by a separate column. Both subfigures show block structures that indicate molecular similarity across datasets between clusters of the same type. [Figure 4] The principal component analysis of the exploration dataset before and after COCONUT is displayed. Before COCONUT conormalization, the exploration dataset is completely separated by technical batch effects. These technical effects are removed after COCONUT, as evidenced by the general overlap of the exploration dataset in the first two principal components. [Figure 5A-D] The output from two consensus clustering algorithms is shown: K-means (A, B) and partitioning around medioids (C, D). (A, C) is the cumulative density function of consensus assignment by number of clusters. (B, D) is the consensus mapping by cluster. 1 = inflammatory, 2 = adaptive, 3 = coagulation disorder. [Figure 6] The COMMUNAL map of cluster optimality is displayed. The X-axis represents the number of clusters, the Y-axis represents the number of genes included, and the Z-axis represents the mean validity score (higher is better). Red and blue dots indicate the optimal conditions automatically assigned for each number of genes included. COMMUNAL automatically selected five validity measures: gap statistics, connectivity, mean contour width, g3 metric, and Pearson's gamma. The resulting map shows the average of the standardized values ​​for each validity measure across the entire tested space. Stable optimal conditions with K=3 clusters are observed for most of the tested space, indicating a strong and consistent biological signal with this number of clusters. Red arrows indicate selected clustering (stable K[3] with the minimum number of genes

[0500] ). [Figure 7]The principal component analysis (PCA) of the exploratory clustering results is displayed (in gold, including 16% of samples that were not clustered in the final analysis), using either all 8,946 genes present in the conormalized COCONUT data or the 500 genes actually used in the clustering analysis. PCA is an unsupervised dimensionality reduction technique that enables visualization of high-dimensional data. Here, it is shown that the unsupervised reconstructed cluster assignments are clearly separated in high-dimensional space, as indicated by the first three principal components. Adaptive samples appear to be separated from inflammatory and coagulation-impairing samples along PC1 and 2, but PC3 shows a much larger separation of inflammatory and coagulation-impairing samples. [Figure 8] This simply displays a heatmap of 500 genes included in the clustering analysis of the search cluster, using hierarchical clustering of genes for visualization purposes. [Figure 9] This displays a comparison of raw predicted probabilities for cluster assignment in the exploratory data. The probability histograms show clear decisions by the adaptive model, but inflammatory and coagulation disorders have lower predicted certainty. [Modes for carrying out the invention]

[0012] The implementation of this invention will, unless otherwise specified, utilize conventional methods of pharmacology, chemistry, biochemistry, recombinant DNA technology, and immunology within the scope of the art possessed by those skilled in the art. Such techniques are fully described in the references. For example, see Handbook of Experimental Immunology, Vols. I-IV (DMWeir and CC Blackwell eds., Blackwell Scientific Publications); ALLehninger, Biochemistry (Worth Publishers, Inc., current addition); Sambrook, et al., Molecular Cloning: A Laboratory Manual (2nd Edition, 1989); Methods In Enzymology (S. Colowick and N. Kaplan eds., Academic Press, Inc.).

[0013] All publications, patents, and patent applications cited herein, whether above or below, are incorporated herein by reference in their entirety.

[0014] Where a range of values ​​is provided, unless otherwise explicitly stated in the context, it is understood that each intermediate value between the upper and lower limits of that range, up to one-tenth of the lower limit unit, is also specifically disclosed. Each smaller range between any stated value or intermediate value and any stated value or intermediate value within which the stated range is included in the Invention is included in the Invention. The upper and lower limits of these smaller ranges may be independently included in or excluded from this range, and each range in which either or both of these limits are included in the smaller range, or neither, is also included in the Invention and subject to any specifically excluded limitations within the stated range. If a stated range includes one or both of these limits, a range that excludes either or both of those included limits is also included in the Invention.

[0015] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar to or equivalent to those described herein can be used in the practice or testing of the present invention, several possible preferred methods and materials are described herein. All publications mentioned herein are hereby incorporated by reference herein for the purpose of disclosing and describing the methods and / or materials in connection with which they are cited. It is understood that the present disclosure supersedes any disclosure of an incorporated publication, to the extent there is a conflict.

[0016] As will be apparent to those skilled in the art upon reading this disclosure, each of the individual embodiments described and illustrated herein can be readily separated from or combined with the features of any of several other embodiments without departing from the scope or spirit of the present invention. Any recited method can be performed in the order of recited events or in any other order that is logically possible.

[0017] As used in this specification and the appended claims, the singular forms "a", "an", and "the" are to be construed to include the plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "an agonist" includes mixtures of two or more agonists and the like.

[0018] The publications discussed herein are provided solely for their disclosure prior to the filing date of the present application. Nothing herein is to be construed as an admission that the inventors are not entitled to antedate such disclosure by virtue of prior invention. Further, the presented publication dates may be different from the actual publication dates, which may need to be independently confirmed.

[0019] Diagnostic method As described above, a method is provided for determining whether a subject with sepsis (i.e., a subject diagnosed with sepsis, or a subject with sepsis that has not yet been diagnosed) has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype. In some embodiments, the method is: (a) To obtain gene expression data, measure the amount of RNA transcripts encoded by at least two of the following (e.g., at least two, at least three, at least five, at least ten, at least fifteen, at least twenty, at least thirty, or all of them) in the RNA sample obtained from the subject, and (b) Provide a report indicating, based on gene expression data, whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype, (i) Increases in ARG1, LCN2, LTF, and / or OLFM4, and / or decreases in HLA-DMB, indicate that the subject has an inflammatory phenotype. (ii) Increases in YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and / or FRS2, and / or decreases in GADD45A, CD24, S100A12, and / or STX1A indicate that the subject has an adaptive phenotype, and (iii) Increases in KCNMB4, CRISP2, HTRA1, and / or PPL, and / or decreases in RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and / or RELB, indicate that the subject has a coagulation disorder phenotype.

[0020] The measurement process can be carried out using any suitable method. For example, the amount of RNA transcript in a sample may be measured by RNA-seq (see, e.g., Morin et al BioTechniques 2008 45:81-94; Wang et al 2009 Nature Reviews Genetics 10:57-63), by RT-PCR (Freeman et al BioTechniques 1999 26:112-22,124-5), or by labeling the RNA or cDNA produced therefrom and hybridizing the labeled RNA or cDNA into an array. The array may contain spatially addressable or optically addressable sequence-specific oligonucleotide probes that specifically hybridize to the transcript to be measured or the cDNA produced therefrom. Spatially addressable arrays (commonly referred to in the art as “microarrays”) are described, for example, by Sealfon et al. (see, e.g., Methods Mol Biol. 2011;671:3-34). Optically addressable arrays (commonly referred to in the art as “bead arrays”) use beads that have been internally stained with fluorophores of different colors, intensities, and / or ratios so that the beads can be distinguished from one another, and these beads are also attached to oligonucleotide probes. Exemplary bead-based assays are described by Dupont et al. (J. Reprod Immunol. 2005 66:175-91) and Khalifian et al. (J Invest Dermatol. 2015 135:1-5). The amount of transcript in the sample can also be analyzed by quantitative RT-PCR or isothermal amplification methods, such as those described by Gao et al. (J. Virol Methods. 2018 255:71-75), Pease et al. (Biomed Microdevices (2018) 20:56) or Nixon et al. (Biomol. Det. and Quant 2014 2:4-10). Many other methods for measuring the amount of RNA transcript in a sample are known in the art.

[0021] RNA samples obtained from subjects may include, for example, RNA isolated from whole blood, leukocytes, neutrophils, or buffy coat. Methods for producing total RNA, poly(A+) RNA, RNA depleted of rich transcripts, and RNA enriched with the transcript to be measured are well known (see, e.g., Hitchen et al J Biomol Tech. 2013 24:S43-S44). If the method produces cDNA from RNA, then the cDNA may be produced using oligo(d)T primers, random primers, or a population of gene-specific primers that hybridize to the transcript to be analyzed.

[0022] Transcript measurement can determine the absolute amount of each transcript or the amount of each transcript compared to one or more control transcripts. An increase or decrease in transcript volume may be related to the amount of transcript (e.g., the average amount of transcript) in a control sample (e.g., in blood samples taken from a population of at least 100, at least 200, or at least 500 subjects with sepsis).

[0023] In some embodiments, the method may include providing a report indicating whether a subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype, based on a measurement of the amount of transcript. In some embodiments, this step involves calculating three scores (one for each phenotype) based on the weight of each transcript, the scores being correlated with the phenotype and may be numbers such as probability, likelihood, or a score out of 10. In these embodiments, the method may include inputting the amount of each transcript into one or more algorithms, running the algorithms, and receiving a score for each phenotype based on the calculations. In these embodiments, other measurements from the subject may be input into the algorithms, such as whether the subject is male, the subject's age, white blood cell count, neutrophil count, band count, lymphocyte count, monocyte count, whether the subject is immunosuppressed, and / or whether Gram-negative bacteria are present.

[0024] In some embodiments, the method generates a report, for example in electronic format, indicating the subject's inflammatory age, and forwards the report to a physician or other healthcare professional to help identify appropriate choices to be made, such as identifying a suitable treatment for the subject. The report may be used in conjunction with other measures as a diagnosis to determine whether the subject has a disease or condition.

[0025] In any embodiment, a report may be transferred to a “remote location,” where “remote location” means a location other than where the image is examined. For example, a remote location may be another location in the same town (e.g., an office, a lab, etc.), another location in a different town, another location in a different state, another location in a different country, etc. Thus, when one item is indicated to be “remote” to another item, it means that the two items are at least one mile, ten miles, or at least 100 miles apart, either in the same room but separate, or in at least different rooms or different buildings. To “communicate” information means to transmit data representing that information as electrical signals over a suitable communication channel (e.g., a private or public network). To “transfer” an item means any means of moving the item from one location to another, whether by physically transporting the item or by any other means (if possible), and at least in the case of data, includes physically transporting a medium containing the data or communicating the data. Examples of communication media include wireless or infrared transmission paths, network connections to other computers or network devices, and the Internet, including email transmission and information recorded on websites, etc. In certain embodiments, the report may be analyzed by an MD or other qualified medical professional, and the report based on the results of the image analysis may be forwarded to the subject from which the sample was obtained.

[0026] In computer-related embodiments, the system may include a computer comprising a processor, a storage component (i.e., memory), a display component, and other components typically found in a general-purpose computer. The storage component stores processor-accessible information, including instructions that can be executed by the processor and data that can be retrieved, manipulated, or stored by the processor.

[0027] The memory component includes instructions for determining whether a subject has inflammation of the inflammatory phenotype, adaptive phenotype, or coagulation phenotype, using the above measurements as input. A computer processor is coupled to the memory component and configured to receive patient data according to one or more algorithms and execute instructions stored in the memory component to analyze the patient data. A display component can display information related to the patient's diagnosis.

[0028] The storage component is any type capable of storing information accessible by the processor, such as a hard drive, memory card, ROM, RAM, DVD, CD-ROM, USB flash drive, writable memory, and read-only memory. The processor may be any well-known processor, such as an Intel Corporation processor. Alternatively, the processor may be a dedicated controller such as an ASIC.

[0029] Instructions can be a set of instructions that are executed directly (such as in machine code) or indirectly (such as in a script) by the processor. In this regard, the terms “instruction,” “process,” and “program” may be used interchangeably herein. Instructions can be stored in object code form for direct processing by the processor, or in any other computer language, including scripts or collections of independent source code modules that are interpreted on demand or pre-compiled.

[0030] The data can be read, stored, or modified by the processor according to instructions. For example, a diagnostic system is not limited to any particular data structure, but the data may be stored in a computer register of a relational database as a table with multiple different fields and records, an XML document, or a flat file. The data can also be formatted in a computer-readable format such as binary values, ASCII, or Unicode. Furthermore, the data may contain any information sufficient to identify relevant information, such as numbers, descriptions, unique codes, pointers, references to data stored in other memory (including other network locations), or information used by functions to compute the relevant data.

[0031] treatment method Treatment methods are also provided. In some embodiments, these methods may include identifying a subject having a phenotype using the methods described above, and treating the subject based on whether the subject is indicated to have an inflammatory phenotype, an adaptive phenotype, or a coagulation disorder phenotype. In some embodiments, this method may be a method for treating a subject having sepsis. In these embodiments, the method is to (a) receive a report indicating whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation disorder phenotype, wherein the report indicates ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A This is based on gene expression data obtained by measuring the amount of RNA transcripts encoded by at least two of the following (e.g., at least two, at least three, at least five, at least ten, at least fifteen, at least twenty, at least thirty, or) CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB, and all (i) Increases in ARG1, LCN2, LTF, and / or OLFM4, and / or decreases in HLA-DMB, indicate that the subject has an inflammatory phenotype. (ii) Increases in YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and / or FRS2, and / or decreases in GADD45A, CD24, S100A12, and / or STX1A indicate that the subject has an adaptive phenotype, and (iii) Increases in KCNMB4, CRISP2, HTRA1, and / or PPL, and / or decreases in RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and / or RELB, indicate that the subject has a coagulation disorder phenotype, and the report is received. (b) Treating the subject based on whether the subject is indicated to have an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype.

[0032] In some embodiments, the treatment involves (a) measuring the amount of RNA transcripts encoded by at least two of the following in an RNA sample obtained from a subject, in order to obtain gene expression data: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB, or the measurement of such amounts. (b) Based on gene expression data, identify subjects as having an inflammatory phenotype, adaptive phenotype, or coagulation disorder phenotype, (i) Increases in ARG1, LCN2, LTF, and / or OLFM4, and / or decreases in HLA-DMB, indicate that the subject has an inflammatory phenotype. (ii) Increases in YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and / or FRS2, and / or decreases in GADD45A, CD24, S100A12, and / or STX1A indicate that the subject has an adaptive phenotype, and (iii) Increases in KCNMB4, CRISP2, HTRA1, and / or PPL, and / or decreases in RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and / or RELB, indicate that the subject has a coagulation disorder phenotype, to be identified. (c) Treatment may include treating the patient accordingly, as described below. Treatment may vary depending on whether the subject is shown to have an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype.

[0033] For example, subjects shown to have inflammatory or adaptive phenotypes include abatacept, Abetimus, abrilumab, adalimumab, afelimomab, aflibercept, alefacept, anakinra, andecaliximab, anifrolumab, anrukinzumab, antilymphocyte globulin, antithymocyte globulin, antifolate, apolizumab, apremilast, aselizumab, and atezolizumab. zumab), atorolimumab, avelumab, azathioprine, basiliximab, belatacept, belimumab, benralizumab, bertilimumab, besilesomab, bleserumab, blisibimod, brazikumab, briakinumab, brodalumab, canakinumab, carlumab, cedelizumab, certolizumab pegol pegol), chloroquine, clazakizumab, clenoliximab, corticosteroids, cyclosporine, daclizumab, dupilumab, durvalumab, eculizumab, efalizumab, eldelumab, elsilimomab,Emapalumab, Enokizumab, Epratuzumab, Erlizumab, Etanercept, Etrolizumab, Everolimus, Fanolesomab, Faralimomab, Fezakinumab, Fletikumab, Font Fontolizumab, Fresolimumab, Galiximab, Gavilimomab, Gevokizumab, Gilvetmab, Golimumab, Gomiliximab, Guselkumab, Gusperimus, Hydroxychloroquine Ibalizumab, immunoglobulin E, inebilizumab, infliximab, inolimomab, integrin, interferon, ipilimumab, itolizumab, ixekizumab, keriximab, lampalizumab ab) Lanadelumab, Lebrikizumab, Leflunomide, Lemalesomab, Lenalidomide, Lenzilumab, Lerdelimumab, Letolizumab, Ligelizumab, Lirilumab, Lulizumab pegol, Lumiliximab, Maslimomab, Mavrilimumab, Mepolizumab,Metelimumab, methotrexate, minocycline, mogamulizumab, morolimumab, muromonab-CD3, mycophenolic acid, namilumab, natalizumab, nerelimomab, nivolumab, obinutuzumab, ocrelizumab, odulimomab, oleclumab, orokizumab, omalizumab, otelyxizumab elixizumab, oxelumab, ozoralizumab, pamrevlumab, pascolizumab, pateclizumab, PDE4 inhibitors, pegsunercept, pembrolizumab, perakizumab, pexelizumab ), pidilizumab, pimecrolimus, placulumab, prozalizumab, pomalidomide, priliximab, purine synthesis inhibitors, pyrimidine synthesis inhibitors, quilizumab, reslizumab, ridafololimus, rilo Nacept (Rilonacept), rituximab, rontalizumab, roasterizumab, ruplizumab, samalizumab, sarilumab, secukinumab, sifalimumab, siplizumab, sirolimus,Sirukumab, Sulesomab, sulfasalazine, Tabalumab, Tacrolimus, Talizumab, Telimomab aritox), temsirolimus, teneliximab, teplizumab, teriflunomide, tezepelumab, tildrakizumab, tocilizumab, tofacitinib, tooralizumab, tralokinumab, tregalizumab, tremelimumab Treatment can be with urocuplumab, umirolimus, urelumab, ustekinumab, bapaliximab, varlilumab, batelizumab, vedolizumab, bepalimomab, visilizumab, vobarilizumab, zanolimmab, zolimomab aritox, zotarolimus, or recombinant human cytokines, such as rh-interferon-gamma, either natural or adaptive immunomodulatory agents.

[0034] In another example, subjects shown to have an inflammatory or adaptive phenotype can be treated with any of the following blockers or signaling modulators: PD1, PDL1, CTLA4, TIM-3, BTLA, TREM-1, LAG3, VISTA, or human clusters of differentiation, including CD1, CD1a, CD1b, CD1c, CD1d, CD1e, CD2, CD3, CD3d, CD3e, CD3g, CD4, CD5, CD6, CD7, CD8, CD8a, CD8b, CD9, CD10, CD11a, CD11b, CD11c, CD11d, CD13, CD14, CD15, CD16, CD16a, CD16b, CD17, CD18, CD19, CD20, CD21, CD22, CD23, CD24, CD25, CD26, CD27, CD28, CD29, CD30, CD31, CD32A, CD32B, CD33, CD34, CD35, CD36, CD37, CD38, CD39, CD40, CD41 , CD42, CD42a, CD42b, CD42c, CD42d, CD43, CD44, CD45, CD46, CD47, CD48, CD49a, CD49b, CD49c, CD49d, CD49e, CD49f, CD50, CD51, CD52, CD53, CD54, CD55, CD56, CD57, CD58, CD59, CD60a, CD60b, CD60c, CD61, CD62E, CD62L, CD62P, CD63, CD64a, CD65, CD65s, CD66a, CD66b, C D66c, CD66d, CD66e, CD66f, CD68, CD69, CD70, CD71, CD72, CD73, CD74, CD75, CD75s, CD77, CD79A, CD79B, CD80, CD81, CD82, CD83, CD84, CD85A, CD85B, CD85C, CD85D, CD85F, CD85G, CD85H, CD85I, CD85J, CD85K, CD85M, CD86, CD87, CD88, CD89, CD90, CD91, CD92, CD93, CD94, CD95, CD96, CD97, CD98, CD99, CD100, CD101, CD102, CD103, CD104, CD105, CD106, CD107, CD107a, CD107b, CD108, CD109, CD110, CD111,CD112、CD113、CD114、CD115、CD116、CD117、CD118、CD119、CD120、CD120a、CD120b、CD121a、CD121b、CD122、CD123、CD124、CD125、CD126、CD127、CD129、CD130、CD131、CD132、CD133、CD134、CD135、CD136、CD137、CD138、CD139、CD140A、CD140B、CD141、CD142、CD143、CD144、CDw145、CD146、CD147、CD148、CD150、CD151、CD152、CD153、CD154、CD155、CD156、CD156a、CD156b、CD156c、CD157、CD158、CD158A、CD158B1、CD158B2、CD158C、CD158D、CD158E1、CD158E2、CD158F1、CD158F2、CD158G、CD158H、CD158I、CD158J、CD158K、CD159a、CD159c、CD160、CD161、CD162、CD163、CD164、CD165、CD166、CD167a、CD167b、CD168、CD169、CD170、CD171、CD172a、CD172b、CD172g、CD173、CD174、CD175、CD175s、CD176、CD177、CD178、CD179a、CD179b、CD180、CD181、CD182、CD183、CD184、CD185、CD186、CD187、CD188、CD189、CD190、CD191、CD192、CD193、CD194、CD195、CD196、CD197、CDw198、CDw199、CD200、CD201、CD202b、CD203c、CD204、CD205、CD206、CD207、CD208、CD209、CD210、CDw210a、CDw210b、CD211、CD212、CD213a1、CD213a2、CD214、CD215、CD216、CD217、CD218a、CD218b、CD219、CD220、CD221、CD222、CD223、CD224、CD225、CD226、CD227、CD228、CD229、CD230、CD231、CD232、CD233、CD234、CD235a、CD235b、CD236、CD237、CD238、CD239, CD240CE, CD240D, CD241, CD242, CD243, CD244, CD245, CD246, CD247, CD248, CD249, CD250, CD251, CD252, CD253, CD254, CD255, CD256, CD257, CD258, CD259, CD260, CD261, CD262, CD263, CD264, CD265, CD266, CD267, CD268, CD269, CD27 0, CD271, CD272, CD273, CD274, CD275, CD276, CD277, CD278, CD279, CD280, CD281, CD282, CD283, CD284, CD285, CD286, CD 287, CD288, CD289, CD290, CD291, CD292, CDw293, CD294, CD295, CD296, CD297, CD298, CD299, CD300A, CD300C, CD301, CD3 02, CD303, CD304, CD305, CD306, CD307, CD307a, CD307b, CD307c, CD307d, CD307e, CD308, CD309, CD310, CD311, CD312, CD 313, CD314, CD315, CD316, CD317, CD318, CD319, CD320, CD321, CD322, CD323, CD324, CD325, CD326, CD327, CD328, CD329, Includes CD330, CD331, CD332, CD333, CD334, CD335, CD336, CD337, CD338, CD339, CD340, CD344, CD349, CD351, CD352, CD353, CD354, CD355, CD357, CD358, CD360, CD361, CD362, CD363, CD364, CD365, CD366, CD367, CD368, CD369, CD370, or CD371.

[0035] In another example, subjects shown to have a coagulation disorder phenotype may be treated with one or more drugs that modify the coagulation cascade or platelet activation, such as those targeting albumin, antihemophilia globulin, AHF A, C1 inhibitors, Ca++, CD63, Christmas factor, AHF B, endothelial growth factor, or epidermal growth factor. Factors, AHF B, Endothelial Growth Factor, Epidermal Growth Factor, Factors V, XI, XIII, Fibrin Stabilizing Factor, Laki-Lorand Factor, Fibrinase, Fibrinogen, Fibronectin, GMP33, Hageman Factor, High Molecular Weight Kininogen, IgA, IgG, IgM, Interleukin-1B, Multimelin, P-Selectin, Plasma Thromboplastin Progenitor, AHF C, Plasminogen Activator Inhibitor 1, Platelet Factor, Platelet-Derived Growth Factor, Prekallikrein, Proaccelerin, Proconvertin, Protein C, Protein M, Protein S, Prothrombin, Stuart-Prower Factor, TF, Thromboplastin, Thrombospondin, Tissue Factor Pathway Inhibitor, Transforming Growth Factor-β, Vascular Endothelial Growth Factor, Vitronectin, von Willebrand Treatment can be performed with one or more agents that modify the coagulation cascade or platelet activation, such as Willebrand factor, α2-antiplasmin, α2-macroglobulin, β-thromboglobulin, or agents that target other members of the coagulation or platelet activation cascade.

[0036] In another example, subjects with a coagulation disorder phenotype may be treated with blood products, heparin, low molecular weight heparin, apixaban, dabigatran, rivaroxaban, dalteparin, fondaparinux, warfarin, activated protein C, recombinant coagulation cascade protein, tranexamic acid, or other coagulation modifiers.

[0037] The method of administering the above-mentioned therapeutic agent and the dosage for administration may be known in the art or derived from the art.

[0038] In some embodiments, the subject may also be treated for sepsis. For example, the present patent may also be used in combination therapy with a broad-spectrum antibiotic, such as meropenem, imipenem, piperacillin-tazobactam, or tigecycline, or metronidazole, in addition to the compounds described above, and one of levofloxacin, aztreonam, cefepime, or ceftriaxone.

[0039] kit This disclosure also provides a kit for carrying out the method as described above. In some embodiments, the kit may include reagents for measuring the amount of RNA transcript encoded by at least two of the following (e.g., at least two, at least three, at least five, at least ten, at least fifteen, at least twenty, at least thirty, or all of them). The kit may include, for each RNA transcript, a sequence-specific oligonucleotide that hybridizes to the transcript. In some embodiments, the sequence-specific oligonucleotide may be biotinylated and / or labeled with an optically detectable moiety. In some embodiments, the kit may include, for each RNA transcript, a pair of PCR primers that amplify a sequence from the RNA transcript or from a cDNA derived therefrom. In some embodiments, the kit may include an array of oligonucleotide probes, the array of which includes, for each RNA transcript, at least one sequence-specific oligonucleotide that hybridizes to the transcript. The oligonucleotide probes may be, for example, spatially addressable on the surface of a planar support or may be coupled to optically addressable beads.

[0040] The various components of the kit may be in separate containers, or, if necessary, specific compatible components may be pre-assembled in a single container.

[0041] In addition to the components described above, the kit may further include instructions for using the kit's components to carry out the method.

[0042] Embodiment 1. A method for determining whether a subject with sepsis has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype, (a) In order to obtain gene expression data, the amount of RNA transcripts encoded by at least two of the following in the RNA sample obtained from the subject is measured: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB. (b) Provide a report indicating, based on gene expression data, whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype, including, (i) Increases in ARG1, LCN2, LTF, and / or OLFM4, and / or decreases in HLA-DMB, indicate that the subject has an inflammatory phenotype. (ii) Increases in YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and / or FRS2, and / or decreases in GADD45A, CD24, S100A12, and / or STX1A indicate that the subject has an adaptive phenotype, and (iii) Increases in KCNMB4, CRISP2, HTRA1, and / or PPL, and / or decreases in RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and / or RELB, indicating that the subject has a coagulation disorder phenotype.

[0043] 2. The method according to Embodiment 1, wherein the measurement step is performed by sequence determination.

[0044] 3. The method according to Embodiment 1, wherein the measurement step is performed by RT-PCR.

[0045] 4. The method according to Embodiment 1, wherein the measurement step is performed by labeling RNA or cDNA produced therefrom, and hybridizing the labeled RNA or cDNA to a support, such as an array or beads.

[0046] 5. The method according to any one of Embodiments 1 to 4, wherein the sample comprises RNA isolated from whole blood, leukocytes, neutrophils, or buffy coat.

[0047] 6. A method for treating a subject with sepsis, (a) Acceptance of a report indicating whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation disorder phenotype, wherein the report is based on gene expression data obtained by measuring the amount of RNA transcripts encoded by at least two of the following in RNA samples obtained from the subject: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB. (i) Increases in ARG1, LCN2, LTF, and / or OLFM4, and / or decreases in HLA-DMB, indicate that the subject has an inflammatory phenotype. (ii) Increases in YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and / or FRS2, and / or decreases in GADD45A, CD24, S100A12, and / or STX1A indicate that the subject has an adaptive phenotype, and (iii) Increases in KCNMB4, CRISP2, HTRA1, and / or PPL, and / or decreases in RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and / or RELB, indicate that the subject has a coagulation disorder phenotype, and accept the report. (b) A method comprising treating a subject based on whether the subject is indicated to have an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype.

[0048] 7. Subjects with inflammatory or adaptive phenotypes are treated with abatacept, abetimus, abrilumab, adalimumab, afelimomab, aflibercept, alefacept, anakinra, andecaliximab, anifrolumab, anrukinzumab, antilymphocyte globulin, antithymocyte globulin, antifolate, apolizumab, apremilast, aselizumab, and atezolizumab. b) Atorolimumab, Avelumab, azathioprine, basiliximab, Belatacept, Belimumab, Benralizumab, Bertilimumab, Besilesomab, Bleserumab, Blisibimod, Brazikumab, Briakinumab, Brodalumab, Canakinumab, Carlumab, Cedelizumab, Certolizumab pegol pegol), chloroquine, clazakizumab, clenoliximab, corticosteroids, cyclosporine, daclizumab, dupilumab, durvalumab, eculizumab, efalizumab, eldelumab, elsilimomab, emapalumab,Enokizumab, Epratuzumab, Erlizumab, Etanercept, Etrolizumab, Everolimus, Fanolesomab, Faralimomab, Fezakinumab, Fletikumab, Fontolizu mab), Fresolimumab, Galiximab, Gavilimomab, Gevokizumab, Gilvetmab, Golimumab, Gomiliximab, Guselkumab, Gusperimus, Hydroxychloroquine, Ivalizumab (I balizumab), immunoglobulin E, inebilizumab, infliximab, inolimomab, integrin, interferon, ipilimumab, itolizumab, ixekizumab, keriximab, lampalizumab, Lanadelumab, Lebrikizumab, Leflunomide, Lemalesomab, Lenalidomide, Lenzilumab, Lerdelimumab, Letolizumab, Ligelizumab, Lirilumab, Lulizumab pegol, Lumiliximab, Maslimomab, Mavrilimumab, Mepolizumab, Metelimumab,Methotrexate, minocycline, mogamulizumab, morolimumab, muromonab-CD3, mycophenolic acid, namilumab, natalizumab, nerelimomab, nivolumab, obinutuzumab, ocrelizumab, odulimomab, oleclumab, olokizumab, omalizumab zumab), otelixizumab, oxelumab, ozoralizumab, pamrevlumab, pascolizumab, pateclizumab, PDE4 inhibitors, pegsunercept, pembrolizumab, perakiz Mab (Perakizumab), Pexelizumab, Pidilizumab, Pimecrolimus, Placulumab, Prozalizumab, Pomalidomide, Priliximab, Purine synthesis inhibitors, Pyrimidine synthesis inhibitors, Quilizumab izumab), Reslizumab, Ridaforolimus, Rilonacept, Rituximab, Rontalizumab, Rovelizumab, Ruplizumab, Samalizumab, Sarilumab, Secukinumab, Sifalimumab, Siplizumab, Sirolimus, Sirukumab,Sulesomab, sulfasalazine, tabalumab, tacrolimus, talizumab, telimomab aritox aritox), temsirolimus, teneliximab, teplizumab, teriflunomide, tezepelumab, tildrakizumab, tocilizumab, tofacitinib, tooralizumab, tralokinumab, tregalizumab, tremelimumab The method according to Embodiment 6, in which the patient is treated with a natural or adaptive immunomodulatory agent such as urocuplumab, umiromus, urelumab, ustekinumab, vapaliximab, varlilumab, batelizumab, vedolizumab, bepalimomab, visilizumab, vobarilizumab, zanolimmab, zolimomab aritox, zotarolimus, or recombinant human cytokines, such as rh-interferon-gamma.

[0049] 8. Subjects with inflammatory or adaptive phenotypes are treated with PD1, PDL1, CTLA4, TIM-3, BTLA, TREM-1, LAG3, VISTA, or human clusters of differentiation (HIS) inhibitors or signaling modulators, including CD1, CD1a, CD1b, CD1c, CD1d, CD1e, CD2, CD3, CD3d, CD3e, CD3g, CD4, CD5, CD6, CD7, CD8, CD8a, CD8b, CD9, CD10, CD11a, CD11b, CD11c, CD11d, CD13, CD14, CD15, CD16, CD16a, CD16b, CD17, CD18, CD19, CD20, CD21, CD22 , CD23, CD24, CD25, CD26, CD27, CD28, CD29, CD30, CD31, CD32A, CD32B, CD33, CD34, CD35, CD36, CD37, CD38, CD39, CD40, CD41, CD42, CD42 a, CD42b, CD42c, CD42d, CD43, CD44, CD45, CD46, CD47, CD48, CD49a, CD49b, CD49c, CD49d, CD49e, CD49f, CD50, CD51, CD52, CD53, CD54, CD 55, CD56, CD57, CD58, CD59, CD60a, CD60b, CD60c, CD61, CD62E, CD62L, CD62P, CD63, CD64a, CD65, CD65s, CD66a, CD66b, CD66c, CD66d, CD 66e, CD66f, CD68, CD69, CD70, CD71, CD72, CD73, CD74, CD75, CD75s, CD77, CD79A, CD79B, CD80, CD81, CD82, CD83, CD84, CD85A, CD85B, CD8 5C, CD85D, CD85F, CD85G, CD85H, CD85I, CD85J, CD85K, CD85M, CD86, CD87, CD88, CD89, CD90, CD91, CD92, CD93, CD94, CD95, CD96, CD97, C D98, CD99, CD100, CD101, CD102, CD103, CD104, CD105, CD106, CD107, CD107a, CD107b, CD108, CD109, CD110, CD111, CD112, CD113, CD114,CD115、CD116、CD117、CD118、CD119、CD120、CD120a、CD120b、CD121a、CD121b、CD122、CD123、CD124、CD125、CD126、CD127、CD129、CD130、CD131、CD132、CD133、CD134、CD135、CD136、CD137、CD138、CD139、CD140A、CD140B、CD141、CD142、CD143、CD144、CDw145、CD146、CD147、CD148、CD150、CD151、CD152、CD153、CD154、CD155、CD156、CD156a、CD156b、CD156c、CD157、CD158、CD158A、CD158B1、CD158B2、CD158C、CD158D、CD158E1、CD158E2、CD158F1、CD158F2、CD158G、CD158H、CD158I、CD158J、CD158K、CD159a、CD159c、CD160、CD161、CD162、CD163、CD164、CD165、CD166、CD167a、CD167b、CD168、CD169、CD170、CD171、CD172a、CD172b、CD172g、CD173、CD174、CD175、CD175s、CD176、CD177、CD178、CD179a、CD179b、CD180、CD181、CD182、CD183、CD184、CD185、CD186、CD187、CD188、CD189、CD190、CD191、CD192、CD193、CD194、CD195、CD196、CD197、CDw198、CDw199、CD200、CD201、CD202b、CD203c、CD204、CD205、CD206、CD207、CD208、CD209、CD210、CDw210a、CDw210b、CD211、CD212、CD213a1、CD213a2、CD214、CD215、CD216、CD217、CD218a、CD218b、CD219、CD220、CD221、CD222、CD223、CD224、CD225、CD226、CD227、CD228、CD229、CD230、CD231、CD232、CD233、CD234、CD235a、CD235b、CD236、CD237、CD238、CD239、CD240CE、CD240D、CD241, CD242, CD243, CD244, CD245, CD246, CD247, CD248, CD249, CD250, CD251, CD252, CD253, CD254, CD255, CD256, CD 257, CD258, CD259, CD260, CD261, CD262, CD263, CD264, CD265, CD266, CD267, CD268, CD269, CD270, CD271, CD272, CD273 , CD274, CD275, CD276, CD277, CD278, CD279, CD280, CD281, CD282, CD283, CD284, CD285, CD286, CD287, CD288, CD289, CD 290, CD291, CD292, CDw293, CD294, CD295, CD296, CD297, CD298, CD299, CD300A, CD300C, CD301, CD302, CD303, CD304, CD 305, CD306, CD307, CD307a, CD307b, CD307c, CD307d, CD307e, CD308, CD309, CD310, CD311, CD312, CD313, CD314, CD315 , CD316, CD317, CD318, CD319, CD320, CD321, CD322, CD323, CD324, CD325, CD326, CD327, CD328, CD329, CD330, CD331, CD The method according to Embodiment 6, which includes CD332, CD333, CD334, CD335, CD336, CD337, CD338, CD339, CD340, CD344, CD349, CD351, CD352, CD353, CD354, CD355, CD357, CD358, CD360, CD361, CD362, CD363, CD364, CD365, CD366, CD367, CD368, CD369, CD370, or CD371.

[0050] 9. Patients with a coagulation disorder phenotype may be treated with one or more drugs that modify the coagulation cascade or platelet activation, such as those targeting albumin, antihemophilia globulin, AHF A, C1 inhibitors, Ca++, CD63, Christmas factor, AHF B, endothelial growth factor, or epidermal growth factor. Factors, AHF B, Endothelial Growth Factor, Epidermal Growth Factor, Factors V, XI, XIII, Fibrin Stabilizing Factor, Laki-Lorand Factor, Fibrinase, Fibrinogen, Fibronectin, GMP33, Hageman Factor, High Molecular Weight Kininogen, IgA, IgG, IgM, Interleukin-1B, Multimelin, P-Selectin, Plasma Thromboplastin Progenitor, AHF C, Plasminogen Activator Inhibitor 1, Platelet Factor, Platelet-Derived Growth Factor, Prekallikrein, Proaccelerin, Proconvertin, Protein C, Protein M, Protein S, Prothrombin, Stuart-Prower Factor, TF, Thromboplastin, Thrombospondin, Tissue Factor Pathway Inhibitor, Transforming Growth Factor-β, Vascular Endothelial Growth Factor, Vitronectin, von Willebrand The method according to Embodiment 6, which is treated with one or more agents that modify the coagulation cascade or platelet activation, such as Willebrand factor, α2-antiplasmin, α2-macroglobulin, β-thromboglobulin, or those that target other members of the coagulation or platelet activation cascade.

[0051] 10. The method according to Embodiment 6, wherein a subject having a coagulation disorder phenotype is treated with a blood product, heparin, low molecular weight heparin, apixaban, dabigatran, rivaroxaban, dalteparin, fondaparinux, warfarin, activated protein C, recombinant coagulation cascade protein, tranexamic acid, or another coagulation modifier.

[0052] 11. The method according to any one of Embodiments 1 to 10, wherein the indicator of whether the subject has an inflammatory phenotype, adaptive phenotype, or coagulation phenotype is further based on whether the subject is male, the subject's age, white blood cell count, neutrophil count, band count, lymphocyte count, monocyte count, whether the subject is immunosuppressed, and / or whether Gram-negative bacteria are present.

[0053] 12. A method, A method comprising measuring the amount of RNA transcript encoded by at least two of the following: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB, in an RNA sample obtained from a subject.

[0054] 13. The method according to any one of Embodiments 1 to 12, wherein gene expression data includes measuring the amount of RNA transcripts encoded by at least three, at least five, at least ten, at least fifteen, at least 20, at least 30, or all of the following in an RNA sample obtained from a subject: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB.

[0055] 14. A kit comprising reagents for measuring the amount of RNA transcript encoded by at least two, at least three, at least five, at least ten, at least fifteen, at least twenty, at least thirty, or all of the following: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB.

[0056] 15. The kit according to Embodiment 14, wherein the reagent includes a sequence-specific oligonucleotide that hybridizes to each RNA transcript.

[0057] 16. The kit according to Embodiment 15, wherein a sequence-specific oligonucleotide is labeled with biotinylation and / or an optically detectable portion.

[0058] 17. The kit according to Embodiment 14, wherein the reagents include a pair of PCR primers for each RNA transcript that amplify a sequence from the RNA transcript or from a cDNA prepared therefrom.

[0059] 18. The kit according to Embodiment 14, wherein the reagent comprises an array of oligonucleotide probes, the array comprising, for each RNA transcript, at least one sequence-specific oligonucleotide that hybridizes to the transcript. [Examples]

[0060] The following examples are described to provide those skilled in the art with a complete disclosure and methods of manufacturing and using the present invention, and are not intended to limit the scope of what the inventors consider to be their invention, nor are they intended to indicate that the following experiments are all or only experiments are performed. Efforts have been made to ensure accuracy for the figures used (e.g., quantity, temperature, etc.), but some degree of experimental error and deviation should be taken into account. Unless otherwise indicated, parts are parts by weight, molecular weight is weight-average molecular weight, temperature is in Celsius, and pressure is atmospheric pressure or near atmospheric pressure. Standard abbreviations can be used, such as room temperature (RT), base pair (bp), kilobase (kb), picoliters (pl), seconds (s or sec), minutes (m or min), hours (h or hr), days (d), weeks (wk or wks), nanoliters (nl), microliters (ul), milliliters (ml), liters (L), nanograms (ng), micrograms (μg), milligrams (mg), grams (g in the sense of mass), kilograms (kg), gravitational equivalent (g, in the sense of centrifugation), nanomoles (nM), micromoles (μM), millimoles (mM), moles (M), amino acids (aa), kilobases (kb), base pairs (bp), nucleotides (nt), intramuscular (im), intraperitoneal (ip), subcutaneous (sc), etc.

[0061] overview Sepsis is not a single disease, but rather may be a spectrum consisting of several "endotypes" (also known as clusters or subclasses of the disease). Assuming there are sepsis clusters that are widespread across patients with sepsis, we used whole blood transcriptome data (gene expression microarrays and RNA sequencing) from a broad range of clinical settings to test this hypothesis.

[0062] A new bioinformatics technique has been published that relies on the assumption that health management across different trials is largely the same. Using this assumption, data from different trials can be pooled into a single framework in a bias-free manner (i.e., without any assumptions about cases of sepsis) and analyzed as if they were collected in a single large trial. Thus, all transcriptome trials of bacterial sepsis at admission were collected and split into all trials including healthy controls (used for sepsis cluster exploration) and all trials without healthy controls (used for validation of clusters found in the exploration).

[0063] Across exploratory data (700 patients from 14 datasets), advanced bioinformatics was used to determine that transcriptome data ideally divided into three clusters. Pathway analysis was performed on the gene expression profiles of subjects within the three clusters, with one cluster exhibiting high innate / adaptive immune signaling reduction ("inflammatory depressant"), one cluster exhibiting low innate / adaptive immune signaling reduction with low mortality ("adaptive"), and one cluster exhibiting both clinical and molecular irregularities in coagulation and complement systems ("coagulable depressant"). Cluster members were associated with significantly different ages, shock states, clinical severity, leukocyte differentiation, and mortality. However, the influence of age, shock states, and disease severity on cluster members was characterized and showed little explanation for why patients were assigned to a given cluster. This suggests that cluster members are not simply explained by obvious clinical variables.

[0064] To establish even a slight clinical relevance, some method is needed to determine cluster membership for any given new patient. In other words, some diagnostic blood test is needed to determine cluster membership when a patient presents with sepsis. Therefore, 33 genetic classifiers of the exploratory data were derived that had an 83% accuracy in reassigning exploratory patients to the same cluster as them. These 33 genetic classifiers were applied to nine external, independent datasets (N=600), retrospectively assigning each of the 600 patients to one of three clusters (inflammatory, adaptive, or coagulation-related).

[0065] These patients were retrospectively assigned to three clusters, requiring determination of whether they replicated the same clinical and biological characteristics as the original inflammatory, adaptive, and coagulation-related groups. The same relative patterns of age, severity, shock, and mortality were observed, on average, between the validation and exploratory clusters. The same pathways were also generally activated among patients across the cohort assigned to the same cluster.

[0066] The analysis shows that there are three distinct subtypes of sepsis: inflammatory, adaptive, and coagulative. These subtypes differ significantly in their clinical and molecular profiles. The study also created 33 genetic classifiers that can identify any new patient as belonging to one of these clusters. The endotype concept is clinically useful because it can be combined with endotype-specific therapies.

[0067] method Systematic search and dataset criteria As previously mentioned, a systematic search of GEO and ArrayExpress for gene expression studies in clinical trials of sepsis (16) was conducted. Individual datasets were re-normalized as previously described (18). Datasets were included only if whole blood gene expression was tested at admission or admission to the ICU (i.e., initial admission for sepsis). Because host responses differ significantly between bacterial and viral infections (15, 19), unsupervised analysis may result in grouping primarily based on the type of infection. All samples with microbiologically confirmed viral infections were removed unless microbiologically confirmed bacterial infections were also present (except in cases where only three confirmed co-infections were included). All studies that did not provide sample-level microbiological data but were identified in the manuscript as being taken primarily from patients with bacterial sepsis were treated as bacterial. Patients sampled more than 48 hours after diagnosis of sepsis were further removed, considering the potential impact of treatment on host response (20, 21). All data used here are anonymized and publicly available and therefore excluded from IRB review.

[0068] COCONUT for pooling data to enable clustering Recent developments in the COmbat CO-Normalization Using controls method (COCONUT) (15) enable bias-free correction of batch effects between multiple microarray datasets, allowing for pooled analysis under the condition that healthy controls are present. The core assumption is that the healthy controls across the entire dataset originate from the same statistical distribution. This assumption allows for the calculation of correction factors that eliminate technical differences across the entire pooled dataset without bias towards the number or type of affected samples present.

[0069] Specifically, to enable the use of the COCONUT method, the dataset was split into "exploratory" and "validation" groups based on whether healthy controls were present in the dataset. Since including healthy controls in a particular dataset is inherently random, it was not expected that the exploration / validation split would introduce bias. Using the COCONUT method, the exploration dataset was conormalized into a single pool, and all healthy controls were removed from subsequent analyses.

[0070] Clustering of exploration data using COMMUNAL To determine the number of clusters present in the conormalized exploratory data using COCONUT, we used the COMMUNAL (Combined Mapping of Multiple clustering algorithms) method, which integrates data from multiple clustering algorithms and effectiveness metrics across the range of variables included to identify the most robust number of clusters present in the data (see Supplementary Materials, Methods and Supplementary Results) (14). The top 5,000 genes across the exploratory dataset were ranked using algorithms that account for both intra-dataset and inter-dataset variability (16). For these robustness in large, noisy datasets, COMMUNAL was run using consensus clustering versions of two algorithms: K-means clustering and Partitioning Around Medioids (PAM). Both methods were run across a variety of variables ranging from 100 to up to 5,000 genes (in ranked order). COMMUNAL then integrated these data (with default parameters) to create an optimized clustering map. In the resulting map, the most stable optimal value was considered to represent the most robust clustering.

[0071] Using COMMUNAL to select the optimal clustering resulted in unified sample assignments across clustering algorithms (i.e., clusters to which PAM and K-means algorithms assigned samples). The COMMUNAL method assigned all samples that matched the search clusters of the clustering algorithms and removed all samples that had mismatches between PAM and K-means as "unclustered." The hypothesis is that not all samples will be perfectly assigned to a given cluster (for example, some samples may exhibit biology suggesting two clusters). Since classifiers trained on low-error data are more robust, removing these uncertain samples improves the accuracy of the classifier. Note that the classifier built for validation does not create "unclustered" assignments (see Supplementary Materials, Methods and Supplementary Results).

[0072] To determine whether the exploratory clusters appeared to be separated in the gene expression space, we used both heatmaps and principal component analysis for visualization. We examined the clinical differences between the exploratory clusters using pooled sample-level demographic and phenotypic data.

[0073] Biological and clinical studies Details of handling complex clinical variables, including disease severity, immunosuppression, and coagulation disorders, are described in the following Supplementary Materials and Methods and Supplementary Results sections. Gene ontology analysis (22), cluster classifier construction (23), and validation dataset testing are described in the Supplementary Materials and Methods and Supplementary Results sections.

[0074] result Selection of included studies, COCONUT conormalization, and COMMUNAL clusters. It was initially hypothesized that a robust molecular subgroup exists among patients with bacterial sepsis. Integrated clustering was performed across 14 bacterial sepsis exploration datasets from eight different countries using COCONUT conormalization (24–37) (N=700, Table 1a). Nine validation datasets were identified from five different countries that met the inclusion criteria but lacked healthy controls (N=600, Table 1b and Figure 1) (12, 38–43). [Table 1-1] [Table 1-2]

[0075] The 14 exploratory datasets were first conormalized into a single pooled cohort using the COCONUT method (15) to provide batch-corrected pooled sepsis data across various clinical conditions (Figure 4). There were 8,946 genes measured across all 14 pooled exploratory datasets. The pooled data were then clustered using the COMMUNAL algorithm with 11 test points spanning the top 100 to 5,000 genes, using consensus K-means and consensus PAM clustering (results for individual clustering algorithms are shown in Figure 5) (14). A visual inspection of the COMMUNAL optimality map showed clear and stable optimal values ​​for K=3 clusters from 500 genes to 5,000 genes (Figure 6). Furthermore, clustering with 500 genes was selected as the optimal clustering assignment under the assumption that using the fewest number of genes would result in the least amount of noise or redundant signal. Based on the gene ontology analysis described below, the three clusters were named "inflammatory," "adaptive," and "coagulation-impairing" to facilitate their understanding.

[0076] To visualize their general separability, principal component analysis was performed on the exploration clusters using all genes, both with and without "unclustered" samples (Figures 2A–2B). Details of the cluster assignments in the exploration dataset are available in Supplementary Results, Table 12, and Supplementary Figures 4–5.

[0077] Gene ontology across different clusters To better understand the biology represented by clusters, overrepresentation analysis of gene ontology was used. Each of the 500 genes was assigned to one of three exploratory clusters based on absolute effect size (i.e., each gene was assigned to the cluster most different from the other two clusters). For each of the resulting three gene lists, the significance of gene ontology (GO) terms was tested. The inflammatory hyperinflammation cluster was significant for standard pro-inflammatory signaling pathways such as IL-1 receptor, pattern recognition receptor activity, and complement activation. The adaptive cluster was significant for several pathways related to adaptive immunity and interferon signaling. The third cluster was named clotting hyperinflammation because it was significant for terms related to clotting and coagulation, such as platelet degranulation, glycosaminoglycan binding, and coagulation cascades.

[0078] Clinical findings across various clusters We examined the differences between exploratory clusters for demographic and clinical variables for which participant-level data (Table 2) were available. [Table 2]

[0079] The following were identified: significant differences in age (both overall distribution and proportion of patients over 70 years), severity (measured by the proportion of patients with clinical severity scores above the dataset mean, and / or septic shock), and 30-day mortality. It was also found that the inflammatory and damaging disease cohort had a larger bandemia and a lower lymphocyte ratio relative to leukocyte percentage, although percentages are only available in a single cohort. This suggests that the adaptive cluster consists of patients with less disease and fewer older patients, while the inflammatory and damaging and coagulation clusters divide patients by disease stage into younger and older groups. The addition of “unclustered” patients showed that they had a balanced phenotype with respect to age and shock, and their addition did not substantially alter demographic or clinical findings (Table 4). Since unsupervised clustering did not take clinical data into account at all, the finding of significant differences in mortality indicates that the clusters represent different pathophysiological states of clinical relevance. [Table 4] [Table 3]

[0080] Regression models were run on cluster members (in a "one vs. other" format) to assess the joint ability of their interactions to predict age, shock, severity, and cluster members. In all cases, exploration revealed that the proportions of variance explained by age, shock, and severity were 9.7%, 6.4%, and 0.7% in the inflammatory, adaptive, and coagulation disorders groups, respectively (N=251, Table 5). Sensitivity analysis showed that these results could only be explained by unmeasured confounding variables with substantially larger effect sizes than the variables included (Table 5). Thus, while age, shock, and severity differ significantly between groups, cluster assignment is far more complex than these three factors alone. [Table 5] [Table 4]

[0081] Validation of cluster classifiers in independent datasets Since sepsis clusters were characterized in the exploratory dataset, it was hypothesized that these same clusters could be recovered in an independent validation dataset using a separate classifier. To test the cluster hypothesis and apply it to an external validation dataset, a gene expression-based classifier for cluster assignment was created. Briefly, the classifier assigns three scores (one for each cluster type) to each sample and applies multi-class regression to output the final cluster assignment (Table 6A-B). The classifier used a total of 33 genes and yielded an overall accuracy of 83% in leave-one-out reassignment of trained samples (Table 6C). The largest classifier inaccuracy was distinguishing inflammatory disorder patients from coagulation disorder patients (Figure 9). The classifier was applied to nine bacterial sepsis validation datasets (Table 7)(12, 38-44), and the accuracy of the classifier was judged by its ability to recover clusters with similar molecular and clinical phenotypes to the exploratory clusters. Since the nine validation datasets were independent of each other, they were examined both using a pooled method (Table 3) and by processing each dataset individually, using the same demographic and clinical variables as the exploratory cluster. Because individual datasets may lack the ability to detect differences, statistical tests were performed on the pooled data, and the same patterns of significance were observed compared to the exploratory cluster. The coagulation disorder cluster had a significantly higher proportion of patients over 70 years of age (P<0.05), while the adaptive cluster had fewer patients with shock (P<0.01), fewer patients with high clinical severity (P<0.05), and a lower mortality rate (P=0.01). [Table 3] [Table 5] [Table 6] [Table 7] [Table 8]

[0082] The coagulation disorder cluster was also associated with clinical coagulation disorders, including disseminated intravascular coagulation (P<0.05, Tables 9–10 and Supplementary Results). [Table 9] [Table 10] [Table 11]

[0083] Molecular similarities between clusters identified through exploration and validation. Since the validation cluster was assigned information from only 33 genes, we examined whether similar biology existed in the complete gene expression profiles across the exploration and validation clusters. First, we calculated the mean gene expression profiles for all 500 clustered genes and tested the correlations between clusters. Significant correlations indicated that the classifier had captured most of the information from the original clustering, and therefore the 33 genes used in the classifier were excluded from this analysis to avoid bias. The Pearson correlations of the mean gene expression profiles within the assigned clusters were high (inflammatory dysfunction cluster, 0.59±0.18; adaptive cluster, 0.67±0.19; coagulation dysfunction cluster, 0.20±0.21, Figure 3A). These correlations were significant between the exploration and validation clusters for all inflammatory dysfunction datasets, all adaptive datasets, and five of the nine coagulation dysfunction datasets (P<0.01). For comparison, mean correlations of 0.01–0.02 were obtained for 1000 random samples of the 500 genes.

[0084] Next, we tested whether the same gene ontology (GO) codes were overexpressed across validation clusters compared to the exploration clusters (Figure 3B). On average, 68%, 87%, and 61% of codes were found to be significant (p<0.01) in the exploration clusters (inflammatory, adaptive, and coagulation disorders) and significant (p<0.05) within the same clusters in validation. Furthermore, block structures were observed within the same cluster types, indicating that pathway enrichments are generally shared within each cluster type.

[0085] Comparison with previously established septic endotypes Two groups had previously undergone clustering using sepsis transcriptome profiles: Wong et al. (9-11) and Davenport et al. (12, 13). Current cluster assignments were compared with previously issued assignments and showed significant overlap with inflammatory and adaptive clusters (Supplementary Results and Table 10).

[0086] Consideration This study performed unsupervised clustering analysis on pooled transcriptome profiles (N=700) from 14 datasets from a broad range of subjects with bacterial sepsis, revealing three robust sepsis clusters (or "endotypes"). Based on molecular and clinical profiles, these clusters were named inflammatory disorder (higher mortality, innate immune activation), adaptive (lower mortality, adaptive immune activation), and coagulation disorder (higher mortality, older age, and with clinical and molecular evidence of coagulation disorder). Next, it was shown that 33 genetic classifiers assigning subjects to these three clusters could recover clinical and molecular phenotypes from nine independent validation datasets (N=600). Finally, it was shown that these clusters could largely explain clusters derived by independent groups using different methods (9, 12). In summary, these results indicate that the host response to septic syndrome can be broadly defined by these three robust clusters.

[0087] In particular, each validation dataset has individual inclusion / exclusion criteria, providing a kind of sensitivity analysis in which identified clusters appear not only in both pooled settings (such as exploration) but also in a more uniform and carefully considered phenotypic cohort. For example, samples from pediatric and adult datasets were pooled during exploration, but this method did not simply cluster patients by age; rather, in subsequent validation, two datasets were pediatric and seven were adult, but all datasets contained a mixture of all three sepsis clusters. The fact that the same broad phenotypic and molecular differences were re-demonstrated in these independent applications of the cluster classifier is strong evidence that cluster members exist across the entire population.

[0088] Despite the differences in outcomes among the three clusters, their clinical utility extends beyond their ability to risk stratify with respect to mortality. Mortality prediction is better achieved using dedicated classifiers, as shown with these same data (18). Instead, the hypothesis underlying the search for sepsis clusters is that “sepsis” represents multiple distinct conditions that manifest in many different ways (3, 6, 45). Thus, the aim of this study was to identify these asymptomatic clusters using a very large pool of sepsis patients across a wide range of clinical conditions. By revealing and defining this heterogeneity, the search and validation of treatments that are beneficial to only one sepsis cluster but neutral or even harmful to others can be more successful (11). For example, both molecular and clinical data suggest that the coagulative cluster is associated with functional coagulation disorders. Given the association between sepsis and clinical coagulation disorders, further testing of the coagulative cluster is needed despite (or perhaps because of) the failure of most therapeutic interventions for sepsis coagulation disorders (3, 46, 47). Similarly, drugs tested in sepsis that are known to modulate the innate or adaptive immune system (such as anti-IL-1 or anti-PD-L1 therapies (48, 49)) should find efficacy in the inflammatory or adaptive cluster, respectively.

[0089] The pathology of the clusters was inferred by assigning each gene to the cluster showing the greatest difference in change compared to the other clusters. For example, the association of innate immune pathways in the inflammatory cluster indicates hyperactivation of the innate immune system, or a relative lack of activation of adaptive immune genes in inflammatory patients compared to other sepsis patients, rather than "normal" innate immune activation. Similarly, the relatively high activation of adaptive immune genes in the adaptive cluster is associated with its lower mortality. Viewed through this lens, the three sepsis clusters offer biological insights that, to some extent, reflect clinical intuition. The relative lack of these changes and the expansion of the adaptive immune response are linked to better outcomes (50), while early hyperactivation of the innate immune system or coagulation cascade in sepsis is linked to higher mortality. Furthermore, since genes were selected based on absolute effect size, the similarity of the gene ontology pathway analysis between the inflammatory and adaptive clusters may reflect the opposite regulation of similar pathways, which is further suggested by the strong inverse correlation between the inflammatory and adaptive clusters in Figures 2A–2B. As described above, these biological insights can be used to guide the treatment of different subtypes.

[0090] Two independent trials identified subgroups of sepsis, one focusing on pediatric sepsis in a US-based cohort (9, 10) and the other on adult sepsis in a UK-based cohort (12, 13). Notably, the two subgroups did not overlap significantly. Comparing the three clusters to previous clusters yielded some interesting findings. First, using comparisons at the level of control, patients assigned to the inflammatory disorder cluster were mostly assigned to endotype B (11) or SRS1 (12). However, endotype B had lower childhood mortality compared to endotype A, while SRS1 had higher adult mortality compared to SRS2. Still, it was encouraging that these independent trials identified patients in the same groups using entirely different methods. Similarly, patients assigned to the adaptive cluster were primarily assigned to SRS2, which in both trials was confirmed to be a group with lower mortality associated with interferon signaling. A third (coagulation disorder) cluster was also identified. Compared to previous studies, the substantially larger sample size and heterogeneity of the exploratory cohort enabled the detection of this third cluster of coagulation disorders.

[0091] Supplementary method COMMUNAL algorithm Conventional clustering methods determine clusters by applying a single clustering algorithm (e.g., K-means clustering) and a single validation metric (e.g., gap statistic(1)) to a single number of variables (e.g., 1,000 genes, usually arbitrarily selected). However, this method can produce unstable and irreproducible results(2). Here, in this study, we use the COMMUNAL method, which combines data from multiple clustering algorithms and a validation metric across a wide range of variables to identify the most robust number of clusters present in the data(2).

[0092] Unsupervised clustering uses high-dimensional distance calculations between samples to identify subgroups within a dataset. Therefore, to improve the signal-to-noise ratio, it is crucial to include variables that are likely to be informative (in this case, genes) while minimizing non-informative variables. Typical single-dataset clustering algorithms usually use several measures of variance to rank variables. However, across multiple co-clustered datasets, this metric may be less useful because technical differences between datasets can lead to larger variances. The top 5,000 genes across the exploration datasets were ranked using an algorithm that accounts for both intra-dataset and inter-dataset variance (measured from mean absolute deviation) (3). The algorithm works as follows: Median absolute deviation (MAD) is first used to rank all genes within each dataset, with the gene with the highest MAD being ranked highest. The median of the overall ranking is calculated across the datasets. However, because cluster distributions may differ from dataset to dataset, this meta-ranking may reduce the weight of informative genes from heterogeneously distributed datasets. Therefore, the top 20 genes from each individual dataset are also included (this number is arbitrary but is set as the default by the creators of the original algorithm). The final meta-ranking algorithm combines the top individual and pooled gene rankings into a single list. Further details can be found in the original current work by Planey & Gevaert (Genome Med, 2016) and the accompanying software package "Coincide" (https: / / github.com / kplaney / CoINcIDE). These ranked genes were then progressively incorporated into the COMMUNAL algorithm.

[0093] Genetic ontology testing To verify whether different clusters exhibit different biological characteristics, each gene used in the final clustering was assigned to the cluster with the highest absolute effect size using microarray significance analysis (SAM) (4). Since these genes generally have high variance across samples, higher differential expression of genes within a particular cluster suggests that this contributes to the identity of that cluster. Gene ontology (GO) enrichment was performed on the resulting gene list using ToppGene (5). A Benjamini-Hochberg corrected p-value less than 0.05 was used as the significance threshold.

[0094] Application in cluster classifiers and validation datasets External validation is a key component of any exercise in clustering. However, in validation, it is important to switch to a supervised method (classification) rather than simply continuing to use unsupervised clustering with new validation datasets. There are two main reasons for this. Firstly, de novo clustering does not generate labels. If clustering is performed for each new dataset and three clusters are created (let's call them A, B, and C), there is no way to match the new cluster to the exploration (dysinflammatory / dyscoagulable / adaptive) cluster. Instead, one must rely on trying to "pattern match" the closest phenotypic and molecular profile (e.g., C=dysinflammatory, A=dyscoagulable, B=adaptive), which obviously introduces a large bias. On the other hand, a classifier directly generates labels and can therefore directly ask whether a validation sample classified as "dysinflammatory" matches the more relevant clinical question of exploring the phenotypic and molecular profile of "dysinflammatory". The second reason for deriving a classifier is that without one, there is no way to assign a new patient to a sample in a clinical setting. This is because clustering can only be performed retrospectively, as it relies on the existence of an entire cohort to establish relative distances between samples. In contrast, classification allows for the predictive determination of a single patient's subtype assignment. For example, a validated classifier is needed when it is necessary to determine which cluster a patient belongs to upon admission.

[0095] Therefore, the gene expression-based classifiers for the resulting clusters were constructed using a two-step process with a one-to-other round-robin approach for all clusters using all genes. First, SAM was used to find genes that were statistically significant to a particular cluster, using all genes (not just those used in clustering). Greedy forward search was used to find the set of genes that maximally separated a particular cluster from all other clusters (6). If there are K clusters, this method produces K scores, and therefore a multiclass logistic regression model was fitted to the K scores as the final classifier using the R package nnet. Thus, to apply the classifier to an external dataset, it is necessary to calculate the score for each cluster and apply the multiclass logistic regression model to the set of scores to obtain the assigned results (see Main Methods). The classifiers for the exploration data were tested using one-out cross-validation to estimate accuracy for validation purposes.

[0096] Testing the validation dataset The classifier was applied to validation datasets, and for each validation dataset, demographic and phenotypic data were calculated for each assigned cluster. Since these datasets were tested individually, the data is presented both as the output from each validation dataset and as a pooled output.

[0097] Next, we determined whether the cluster assignments within the validation clusters represented the same biology as the matching clusters in the exploration data. First, we scaled each gene within its local dataset, and then obtained the mean value for each gene within each assigned cluster in each dataset. This left us with a mean difference vector for each gene within each cluster. These mean difference vectors were correlated between the exploration clusters and all validation clusters, and the results were plotted as a heatmap.

[0098] Pathway-based methods were also employed to confirm biological consistency between the exploration and validation clusters. Within each validation cohort, the same SAM method described above was used to assign overrepresented GO pathways to each validation cluster. Next, all GO pathways that were found to be significant in the exploration cluster were tested in all validation clusters, and the resulting significance levels were plotted as a heatmap, with the order of the columns (GO) determined by the significance level in the exploration cluster.

[0099] E-value sensitivity analysis To address the potential for confounding not measured in sample assignment, a sensitivity analysis based on the “E-value” was performed (7). In this method, the “E-value” is determined using a given risk ratio (RR), which is the effect size that unmeasured confounding factors must have on both the explanatory and outcome variables in order to somehow explain the observed RR. To put the range of E-values ​​into context, the RRs of already measured potential confounding variables (here, age, shock, and severity) were tested for both the explanatory variable (cluster assignment) and the outcome variable (mortality). The resulting RRs were then compared to the calculated E-value to determine the magnitude of the effect that potential confounding factors must have in order to explain the observed effect. This application is useful for testing the relationship between cluster assignment and mortality.

[0100] Cluster classifier Because the diverse datasets covered a wide range of microarray types, a two-stage classification method was constructed, running a generative model (regression) on the output of a parameterless algorithm (differential gene expression) to overcome technical differences between microarrays. Thus, the classifier has two stages. In the first stage, three cluster-specific scores are created by examining differences in gene expression. Each of the three cluster-specific scores is calculated by calculating the geometric mean of the "up" genes in a particular cluster and subtracting the geometric mean of the "down" genes. For example, the "inflammatory disorder" score is calculated as follows: (ARG1*LCN2*LTF*OLFM4)^(1 / 4)-(HLADMB). In the second stage, a multi-class regression algorithm takes each of these three scores for each sample and creates a final prediction. This two-stage process was necessary to utilize the full range of data across a wide variety of microarrays.

[0101] Clinical parameters One of the key clinical variables is clinical severity, measured by a standardized score. Because each of the different datasets used a different clinical severity score (e.g., APACHE II, SOFA, SAPS, PRISM), these scores could not be pooled across datasets. Instead, for each dataset, a mean clinical severity score was calculated, and patients were labeled as either “high clinical severity” (greater than average) or “low clinical severity” (less than average). The proportion of patients with “high clinical severity” within each cluster in each dataset was then calculated as a way to test clinical severity across the various datasets.

[0102] Immunosuppressed status was available in two datasets (GSE63042 and GSE66099). In both cases, the categories were binary (regardless of whether the individual was immunosuppressed or not), as retrospectively documented by the registration team. In the case of GSE66099, no precise criteria existed, and in the case of GSE63042, the composite categories included absolute neutropenia, AIDS, chronic immunosuppressants or corticosteroids, chemotherapy, or "other immunosuppressants." Therefore, the categories are considered heterogeneous.

[0103] General method All analyses were performed using the R statistical computing language, version 3.1.1. Categorical data were tested using chi-squared or Fisher's exact test, and continuous data were tested using ANOVA. Unless otherwise specified, significance was set to P<0.05.

[0104] Supplementary results Assignment of exploration clusters For 500 genes, there was an 84% agreement between the K-means algorithm and the PAM algorithm when assigning samples to three clusters. Of the samples with mismatched assignments (N=112), 16% were removed as "unclustered," and the remaining samples were assigned to exploratory clusters. The cluster distribution across the dataset was diverse (Table 12), which was expected given the varying registration criteria assumed for each dataset. [Table 12]

[0105] Principal component analysis (PCA) was performed to assess whether the 500 gene subsets captured the majority of the variance across all measured genes in the exploration data. Both PCAs showed clear separation between the three clusters, with the "unclustered" samples distributed across the three clusters (Figure 7). Heatmaps of the 500 genes used for clustering also showed clear differences between the clusters, as expected (Figure 8).

[0106] Clinical coagulation disorders in the coagulation disorder cluster To examine whether there was functional evidence of coagulation disorder in the coagulation disorder cluster, we looked at standard measures of coagulation disorder and tested whether they were distributed differently across the three clusters. In the only cohort for which we had access to these data (pediatric ICU, GSE66099), disseminated intravascular coagulation (DIC) was significantly associated with the coagulation disorder cluster (P < 0.05, Table 9). In another dataset (adults, CAPSOD, GSE63042), the crossover of thrombocytopenia (platelets < 100,000) and long-term INR (> 1.3) was also significantly associated with the coagulation disorder cluster, although the parameters themselves were not significantly associated with cluster type (Table 10). These findings suggest that the coagulation disorder cluster may be associated with more severe forms of coagulation disorder, such as DIC, but may also be associated only with thrombocytopenia.

[0107] Comparison with previously established septic endotypes Two groups have previously undergone clustering using sepsis transcriptome profiles. Wong et al. (exploratory dataset GSE66099) derived three endotypes of pediatric sepsis, and have since validated two: endotype A (high mortality, adaptive immunosuppression, and reduced glucocorticoid receptor signaling) and endotype B (low mortality) (9-11). Davenport et al. (validation dataset EMTAB-4421.51) derived two clusters of adult sepsis: sepsis response signature (SRS) 1 (high mortality, endotoxin signaling, T cell suppression, and NF-κB activation) and SRS2 (low mortality, T cell activation, and interferon signaling) (12, 13). For each subject in these two cohorts, current cluster assignments were compared to previously published assignments (Table 10). Most samples in the inflammatory cluster (60 out of 69) were endotype B as identified by Wong et al. However, the reverse was not true, with 22 endotype B samples in the adaptive cluster and 13 in the coagulation disorder cluster. Consistent with these findings, the Goodman and Kruskal lambda showed unidirectional significance, indicating that the current clusters have the ability to explain Wong's clusters, but not the other way around. The correlation was stronger in Davenport et al.'s cluster, with the inflammatory disorder cluster and the adaptive cluster representing SRS1 and 2, respectively. The lambda between the current cluster and the Davenport cluster was significant in both directions. These results suggest that endotype B and SRS1 may represent the inflammatory disorder cluster, and SRS2 may represent the adaptive cluster.

[0108] Therefore, the above description merely illustrates the principles of the present invention. Those skilled in the art will understand that various configurations embodying the principles of the present invention and falling within its spirit and scope can be devised, although these are not explicitly described or illustrated herein. Furthermore, all examples and conditional statements enumerated herein are intended primarily to assist the reader in understanding the principles of the present invention and the concepts to which the inventors have contributed to the advancement of the art, and should be interpreted not as limitations to such specifically enumerated examples and conditions. In addition, all descriptions herein enumerating the principles, aspects, and embodiments of the present invention, as well as specific examples thereof, are intended to encompass both their structural and functional equivalents. Moreover, such equivalents are intended to include both currently known equivalents and those to be developed in the future, i.e., any development elements that perform the same function regardless of their structure. Therefore, the scope of the present invention is not intended to be limited to the exemplary embodiments illustrated and described herein. Rather, the scope and spirit of the present invention are embodied in the appended claims.

[0109] References 1.Singer M, Deutschman CS, Seymour CW, et al.The Third International Consensus Definitions for Sepsis and Septic Shock(Sepsis-3).JAMA 2016;315(8):801-810. 2.Liu V, Escobar GJ, Greene JD, et al.Hospital Deaths in Patients With Sepsis From 2 Independent Cohorts.JAMA 2014. 3.Opal SM,Dellinger RP,Vincent JL,et al.The next generation of sepsis clinical trial designs:what is next after the demise of recombinant human activated protein C?*.Crit Care Med 2014;42(7):1714-1721. 4.van’t Veer LJ,Dai H,van de Vijver MJ,et al.Gene expression profiling predicts clinical outcome of breast cancer.Nature 2002;415(6871):530-536. 5.Golub TR,Slonim DK,Tamayo P,et al.Molecular classification of cancer:class discovery and class prediction by gene expression monitoring.Science 1999;286(5439):531-537. 6.Prescott HC,Calfee CS,Thompson BT,et al.Toward Smarter Lumping and Smarter Splitting:Rethinking Strategies for Sepsis and Acute Respiratory Distress Syndrome Clinical Trial Design.Am J Respir Crit Care Med 2016;194(2):147-155. 7.Haldar P,Pavord ID,Shaw DE,et al.Cluster analysis and clinical asthma phenotypes.Am J Respir Crit Care Med 2008;178(3):218-224. 8.Famous KR,Delucchi K,Ware LB,et al.Acute Respiratory Distress Syndrome Subphenotypes Respond Differently to Randomized Fluid Management Strategy.Am J Respir Crit Care Med 2017;195(3):331-338. 9.Wong HR,Cvijanovich N,Lin R,et al.Identification of pediatric septic shock subclasses based on genome-wide expression profiling.BMC Med 2009;7:34. 10.Wong HR,Cvijanovich NZ,Allen GL,et al.Validation of a gene expression-based subclassification strategy for pediatric septic shock.Crit Care Med 2011;39(11):2511-2517. 11.Wong HR,Cvijanovich NZ,Anas N,et al.Developing a clinically feasible personalized medicine approach to pediatric septic shock.Am J Respir Crit Care Med 2015;191(3):309-315. 12.Davenport EE,Burnham KL,Radhakrishnan J,et al.Genomic landscape of the individual host response and outcomes in sepsis:a prospective cohort study.Lancet Respir Med 2016. 13.Burnham KL,Davenport EE,Radhakrishnan J,et al.Shared and Distinct Aspects of the Sepsis Transcriptomic Response to Fecal Peritonitis and Pneumonia.Am J Respir Crit Care Med 2016. 14.Sweeney TE,Chen AC,Gevaert O.Combined Mapping of Multiple clUsteriNg ALgorithms(COMMUNAL):A Robust Method for Selection of Cluster Number,K.Sci Rep 2015;5:16971. 15.Sweeney TE,Wong HR,Khatri P.Robust classification of bacterial and viral infections via integrated host gene expression diagnostics.Sci Transl Med 2016;8(346):346ra391. 16.Planey CR,Gevaert O.CoINcIDE:A framework for discovery of patient subtypes across multiple datasets.Genome Med 2016;8(1):27. 17.Sweeney TE,Khatri P.Benchmarking Sepsis Gene Expression Diagnostics Using Public Data.Crit Care Med 2016. 18.Sweeney TE,Perumal TM,Henao R,et al.Mortality prediction in sepsis via gene expression analysis:a community approach.Nature Communications,Accepted,2018. 19.Andres-Terre M,McGuire HM,Pouliot Y,et al.Integrated,Multi-cohort Analysis Identifies Conserved Transcriptional Signatures across Multiple Respiratory Viruses.Immunity 2015;43(6):1199-1211. 20.Sweeney TE,Shidham A,Wong HR,et al.A comprehensive time-course-based multicohort analysis of sepsis and sterile inflammation reveals a robust diagnostic gene set.Sci Transl Med 2015;7(287):287ra271. 21.Seok J,Warren HS,Cuenca AG,et al.Genomic responses in mouse models poorly mimic human inflammatory diseases.Proc Natl Acad Sci U S A 2013;110(9):3507-3512. 22.Chen J,Bardes EE,Aronow BJ,et al.ToppGene Suite for gene list enrichment analysis and candidate gene prioritization.Nucleic Acids Res 2009;37(Web Server issue):W305-311. 23.Tusher VG,Tibshirani R,Chu G.Significance analysis of microarrays applied to the ionizing radiation response.Proc Natl Acad Sci U S A 2001;98(9):5116-5121. 24.Irwin AD,Marriage F,Mankhambo LA,et al.Novel biomarker combination improves the diagnosis of serious bacterial infections in Malawian children.BMC Med Genomics 2012;5:13. 25.Almansa R,Heredia-Rodriguez M,Gomez-Sanchez E,et al.Transcriptomic correlates of organ failure extent in sepsis.J Infect 2015;70(5):445-456. 26.Emonts M.Polymorphisms in Immune Response Genes in Infectious Diseases and Autoimmune Diseases [Polymorphisms in Immune Response Genes in Infectious Diseases and Autoimmune Diseases; Ph.D.thesis]:Erasmus University Rotterdam; 2008. 27.Pankla R,Buddhisa S,Berry M,et al.Genomic transcriptional profiling identifies a candidate blood biomarker signature for the diagnosis of septicemic melioidosis.Genome Biol 2009;10(11):R127. 28.Parnell GP,McLean AS,Booth DR,et al.A distinct influenza infection signature in the blood transcriptome of patients with severe community-acquired pneumonia.Crit Care 2012;16(4):R157. 29.Sutherland A,Thomas M,Brandon RA,et al.Development and validation of a novel molecular biomarker diagnostic test for the early detection of sepsis.Crit Care 2011;15(3):R149. 30.Ahn SH,Tsalik EL,Cyr DD,et al.Gene expression-based classifiers identify Staphylococcus aureus infection in mice and humans.PLoS One 2013;8(1):e48979. 31.Parnell GP,Tang BM,Nalos M,et al.Identifying key regulatory genes in the whole blood of septic patients to monitor underlying immune dysfunctions.Shock 2013;40(3):166-174. 32.Lill M,Koks S,Soomets U,et al.Peripheral blood RNA gene expression profiling in patients with bacterial meningitis.Front Neurosci 2013;7:33. 33.Cazalis MA,Lepape A,Venet F,et al.Early and dynamic changes in gene expression in septic shock patients:a genome-wide approach.Intensive Care Med Exp 2014;2(1):20. 34.Scicluna BP,Klein Klouwenberg PM,van Vught LA,et al.A molecular biomarker to diagnose community-acquired pneumonia on intensive care unit admission.Am J Respir Crit Care Med 2015;192(7):826-835. 35.Wong HR,Shanley TP,Sakthivel B,et al.Genome-level expression profiles in pediatric septic shock indicate a role for altered zinc homeostasis in poor outcome.Physiol Genomics 2007;30(2):146-155. 36.Wong HR,Cvijanovich N,Allen GL,et al.Genomic expression profiling across the pediatric systemic inflammatory response syndrome,sepsis,and septic shock spectrum.Crit Care Med 2009;37(5):1558-1566. 37.Kwan A,Hubank M,Rashid A,et al.Transcriptional instability during evolving sepsis may limit biomarker based risk stratification.PLoS One 2013;8(3):e60501. 38.Howrylak JA,Dolinay T,Lucht L,et al.Discovery of the gene signature for acute lung injury in patients with sepsis.Physiol Genomics 2009;37(2):133-139. 39.Dolinay T,Kim YS,Howrylak J,et al.Inflammasome-regulated cytokines are critical mediators of acute lung injury.Am J Respir Crit Care Med 2012;185(11):1225-1234. 40.Tsalik EL,Henao R,Nichols M,et al.Host gene expression classifiers diagnose acute respiratory illness etiology.Sci Transl Med 2016;8(322):322ra311. 41.Kangelaris KN,Prakash A,Liu KD,et al.Increased expression of neutrophil-related genes in patients with early sepsis-induced ARDS.Am J Physiol Lung Cell Mol Physiol 2015;308(11):L1102-1113. 42.McHugh L,Seldon TA,Brandon RA,et al.A Molecular Host Response Assay to Discriminate Between Sepsis and Infection-Negative Systemic Inflammation in Critically Ill Patients:Discovery and Validation in Independent Cohorts.PLoS Med 2015;12(12):e1001916. 43.Khoo SK,Petillo D,Parida M,et al.Host response transcriptional profiling reveals extracellular components and ABC(ATP-binding cassette) transporters gene enrichment in typhoid fever-infected Nigerian children.BMC Infect Dis 2011;11:241. 44.Langley RJ,Tsalik EL,van Velkinburgh JC,et al.An integrated clinico-metabolomic model improves prediction of death in sepsis.Sci Transl Med 2013;5(195):195ra195. 45.Cohen J,Vincent JL,Adhikari NK,et al.Sepsis:a roadmap for future research.Lancet Infect Dis 2015;15(5):581-614. 46.Bernard GR,Vincent JL,Laterre PF,et al.Efficacy and safety of recombinant human activated protein C for severe sepsis.N Engl J Med 2001;344(10):699-709. 47.Allingstrup M,Wetterslev J,Ravn FB,et al.Antithrombin III for critically ill patients:a systematic review with meta-analysis and trial sequential analysis.Intensive Care Med 2016;42(4):505-520. 48.Shakoory B,Carcillo JA,Chatham WW,et al.Interleukin-1 Receptor Blockade Is Associated With Reduced Mortality in Sepsis Patients With Features of Macrophage Activation Syndrome:Reanalysis of a Prior Phase III Trial.Crit Care Med 2016;44(2):275-281. 49.Shindo Y,McDonough JS,Chang KC,et al.Anti-PD-L1 peptide improves survival in sepsis.J Surg Res 2017;208:33-39. 50.Angus DC,van der Poll T.Severe sepsis and septic shock.N Engl J Med 2013;369(9):840-851. 51.Oved K,Cohen A,Boico O,et al.A novel host-proteome signature for distinguishing between acute bacterial and viral infections.PLoS One 2015;10(3):e0120012.

[0110] The dressing is a good way to dress the dressing gown 1.Tibshirani R,Walther G,Hastie T.Estimating the number of clusters in a data set via the gap statistic.Journal of the Royal Statistical Society Series B-Statistical Methodology 2001; 63:411-4 2.Sweeney TE,Chen AC,Gevaert O.Combined Mapping of Multiple clUsteriNg ALgorithms(COMMUNAL):A Robust Method for Selection of Cluster Number,K.Sci Rep 2015; 5:16971. 3.Planey CR,Gevaert O.CoINcIDE:A framework for discovery of patient subtypes across multiple datasets.Genome Med 2016; 8:27. 4.Tusher VG,Tibshirani R,Chu G.Significance analysis of microarrays applied to the ionizing radiation response.Proc Natl Acad Sci U S A 2001; 98:5116-5121. 5.Chen J,Bardes EE,Aronow BJ,Jegga AG.ToppGene Suite for gene list enrichment analysis and candidate gene prioritization.Nucleic Acids Res 2009; 37:W305-311. 6.Sweeney TE,Shidham A,Wong HR,Khatri P.A comprehensive time-course-based multicohort analysis of sepsis and sterile inflammation reveals a robust diagnostic gene set.Sci Transl Med 2015; 7:287ra271. 7.VanderWeele TJ,Ding P.Sensitivity Analysis in Observational Research:Introducing the E-Value.Ann Intern Med 2017.

Claims

1. A method for providing a physician or other healthcare professional with a report on whether a subject with bacterial sepsis has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype, wherein the method comprises: (a) In order to obtain gene expression data, a step of measuring the amount of RNA transcript encoded by at least two of the following in an RNA sample obtained from the subject: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB, (b) Providing the physician or other healthcare professional with the report indicating, based on the gene expression data, whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype, (i) Increased expression of at least one of ARG1, LCN2, LTF, and OLFM4 compared to the average amount of the transcript in the control sample (i.e., reference value), decreased expression of HLA-DMB compared to the reference value, or a combination thereof, indicates that the subject has an inflammatory phenotype. (ii) An increase in the expression of at least one of YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and FRS2 compared to the reference value, a decrease in the expression of at least one of GADD45A, CD24, S100A12, and STX1A compared to the reference value, or a combination thereof, indicates that the subject has an adaptive phenotype. (iii) A step in which an increase in the expression of at least one of KCNMB4, CRISP2, HTRA1, and PPL compared to a reference value, a decrease in the expression of at least one of RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB compared to a reference value, or a combination thereof, indicates that the subject has a coagulation disorder phenotype. Methods that include...

2. The method according to claim 1, wherein the measurement step is performed by sequence determination.

3. The method according to claim 1, wherein the measurement step is performed by RT-PCR.

4. The method according to claim 1, wherein the measurement step is performed by labeling the RNA or cDNA produced therefrom, and hybridizing the labeled RNA or cDNA to a support, for example, an array or beads.

5. The method according to any one of claims 1 to 4, wherein the sample comprises RNA isolated from whole blood, leukocytes, neutrophils, or buffy coat.

6. A method for assigning subjects having bacterial sepsis to a subgroup, wherein the method is: (a) A step of receiving a report indicating whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype, wherein the report indicates that the RNA sample obtained from the subject contains ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP 1. Based on gene expression data obtained by measuring the amount of RNA transcripts encoded by at least two of the following: PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB. (i) Increased expression of at least one of ARG1, LCN2, LTF, and OLFM4 compared to the average amount of the transcript in the control sample (i.e., reference value), decreased expression of HLA-DMB compared to the reference value, or a combination thereof, indicates that the subject has an inflammatory phenotype. (ii) An increase in the expression of at least one of YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and FRS2 compared to the reference value, a decrease in the expression of at least one of GADD45A, CD24, S100A12, and STX1A compared to the reference value, or a combination thereof, indicates that the subject has an adaptive phenotype. (iii) A step of receiving a report indicating that an increase in the expression of at least one of KCNMB4, CRISP2, HTRA1, and PPL compared to a reference value, a decrease in the expression of at least one of RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB compared to a reference value, or a combination thereof, indicates that the subject has a coagulation disorder phenotype. (b) A step of assigning the subject to a specific subgroup based on whether the subject is indicated to have an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype, Methods that include...

7. The method according to claim 6, wherein a subject having an inflammatory phenotype or an adaptive phenotype is treated with a natural or adaptive immunomodulatory agent.

8. The method according to claim 6, wherein a subject having an inflammatory phenotype or adaptive phenotype is treated with a blocker or signaling modulator of any of the following: PD1, PDL1, CTLA4, TIM-3, BTLA, TREM-1, LAG3, VISTA, or the human surface antigen classification.

9. The method according to claim 6, wherein a subject having a coagulation disorder phenotype is treated with one or more agents that modify the coagulation cascade or platelet activation.

10. The method according to claim 6, wherein a subject having a coagulation disorder phenotype is treated with a blood product, heparin, low molecular weight heparin, apixaban, dabigatran, rivaroxaban, dalteparin, fondaparinux, warfarin, activated protein C, recombinant coagulation cascade protein, tranexamic acid, or another coagulation modifier.

11. The method according to any one of claims 1 to 10, wherein the indicator of whether the subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype is further based on whether the subject is male, the subject's age, white blood cell count, neutrophil count, band count, lymphocyte count, monocyte count, whether the subject is immunosuppressed, whether Gram-negative bacteria are present, or any combination thereof.

12. A method for assigning subjects with bacterial sepsis to subgroups, wherein the method is: (a) In order to obtain gene expression data, a step of measuring the amount of RNA transcript encoded by at least two of the following in an RNA sample obtained from the target: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB, (b) A step of analyzing the expression levels of each measured RNA transcript along with their respective reference ranges, (i) Increased expression of at least one of ARG1, LCN2, LTF, and OLFM4 compared to the average amount of the transcript in the control sample (i.e., reference value), decreased expression of HLA-DMB compared to the reference value, or a combination thereof, indicates that the subject has an inflammatory phenotype. (ii) An increase in the expression of at least one of YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, and FRS2 compared to the reference value, a decrease in the expression of at least one of GADD45A, CD24, S100A12, and STX1A compared to the reference value, or a combination thereof, indicates that the subject has an adaptive phenotype. (iii) A step in which an increase in the expression of at least one of KCNMB4, CRISP2, HTRA1, and PPL compared to a reference value, a decrease in the expression of at least one of RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB compared to a reference value, or a combination thereof, indicates that the subject has a coagulation disorder phenotype. (c) A step of assigning the subject to a specific subgroup depending on whether the subject has the inflammatory phenotype, the adaptive phenotype, or the coagulation phenotype, Methods that include...

13. A kit for use in a method for assigning subjects having bacterial sepsis according to any one of claims 6 to 12 to a subgroup, wherein the kit comprises: ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19 , SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, STX1A, KCNMB4, CRISP2, A kit comprising reagents for measuring the amount of RNA transcripts encoded by at least two of HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB, and instructions for determining whether a subject has an inflammatory phenotype, an adaptive phenotype, or a coagulation phenotype.

14. The kit according to claim 13, wherein the reagent comprises a sequence-specific oligonucleotide that hybridizes to each RNA transcript.

15. The kit according to claim 14, wherein the sequence-specific oligonucleotide is biotinylated, labeled with an optically detectable moiety, or a combination thereof.

16. The kit according to claim 13, wherein the reagent comprises a pair of PCR primers for each RNA transcript, for amplifying a sequence from the RNA transcript or from a cDNA prepared therefrom.

17. The kit according to claim 13, wherein the reagent comprises an array of oligonucleotide probes, and the array comprises, for each RNA transcript, at least one sequence-specific oligonucleotide that hybridizes to the transcript.

Images