Cancer patient stratification biomarkers

Abstract

The present invention relates to a method of predicting the responsiveness to Immune Checkpoint Inhibitor (ICI)-based treatments in cancer subjects at baseline based on the relative abundances of at least 2 of 8 microbial taxa in a stool sample from the patient. The method produces a binary classification of patients as responders and non-responders to ICI-based treatments. More particularly, the stratification method relates to stratifying hepatocellular carcinoma patients into 2 classes. The invention also relates to ICI-based treatments that can be administered to responders. While the identified biomarkers partially overlap the previously reported ones, the novelty of this invention is based on a) a combination of mixed-level taxonomy (combining genus- and family-level taxonomic information) and b) development of an original machine learning pipeline to leverage this type of input. Also provided is a method of normalisation and look-up table for carrying out the invention on arbitrary datasets, shown to exactly reproduce results from a deep learning model. As a result, while the markers are partially known, their context of use such as combination with markers of different nature and in-house machine learning architecture resulted in the predictive performance that was never seen before for this task and data type.

Description

[0001] CANCER PATIENT STRATIFICATION BIOMARKERS

[0002] This application claims priority from US 63 / 762,204 filed 24 February 2025, the contents and elements of which are herein incorporated by reference for all purposes.

[0003] FIELD OF THE INVENTION

[0004] The invention relates to a method of stratifying cancer patients by predicting their responsiveness to certain treatments that have Immune Checkpoint Inhibitors (ICIs) as a component. More particularly, this method stratifies cancer patients based on certain gut microbiota signatures of the patients.

[0005] BACKGROUND OF THE INVENTION

[0006] Cancer has had a profound impact on societies worldwide with a recent report by the WHO indicating that, in 2022 alone, 20 million new cases of cancer emerged, whilst the number of cancer deaths was 9.7 million. The projected cancer burden is predicted to increase by 77% over the 2022 numbers in 2050.

[0007] Recently, the use of cancer therapies that harness a patient’s immune system has shown remarkable results. This group of therapies, named Immune checkpoint inhibitors (ICIs), interferes with mechanisms that cancer cells use to evade the host’s immune response. The common mechanisms of action of these therapies include blocking cytotoxic T lymphocyte-associated protein 4 (CTLA-4), programmed cell death protein 1 (PD-1) and programmed death ligand 1 (PD-L1), resulting in a re-activation of previously suppressed cancer-killing T-cells.

[0008] However, the response rate of patients to ICIs is in the region of 20-40% with factors such as the tumor’s genetic profile, the host’s baseline immune response and the tumor microenvironment contributing to the varying degree of responsiveness to an ICI therapy. In the specific case of Hepatocellular Carcinoma (HCC), the objective response rates to anti-PD1 agents such as nivolumab and pembrolizumab are in the region of 15-20% (Haber eta / ., 2022). Given this discrepancy in treatment outcomes, biomarker tools to stratify patients into responders and non-responders to ICIs will benefit clinicians in 2 ways. Firstly, ICI therapies place an immense cost burden on healthcare systems and stratification tools can reduce this burden. Secondly, ICIs can trigger severe auto-immune reactions in non-responders - a sideeffect that further exacerbates costs and the condition of cancer patients.

[0009] Currently, host-derived biomarkers such as genomic alterations, protein expressions or specific host cell enrichment in the tumor microenvironment have been used to predict responses to ICIs. Examples of the first group include tumor mutational burden (TMB),microsatellite instability and gene mutations in a number of specific pathways that may be associated with either therapy response or non-response (Bai etal., 2020). The second group encompasses both expression level of PD-L1 itself (Hao et al., 2023) and the abundance of CD8+ T cells (Tumeh et al., 2014). However, these biomarkers often have either limited predictive power or / and are invasive to obtain, requiring tissue biopsies.

[0010] Recently, exploring gut microbiome-based biomarkers to stratify cancer patients into responder and non-responder categories to ICIs has attracted the attention of researchers. The analysis can be performed on stool samples, eliminating the need for tumor biopsies. While associations between gut microbiota and responsiveness have been evaluated in advanced melanoma (Lee etal., 2022), HCC (Wu etal., 2022) and Non-small cell lung cancer (Derosa et al., 2024), this burgeoning field of gut microbiota-based biomarkers is yet to result in a robust tool that can have any meaningful clinical application. It is an object of this invention to address this unmet need.

[0011] SUMMARY OF THE INVENTION

[0012] In a first aspect, the invention provides a method of predicting the responsiveness to an Immune Checkpoint Inhibitor (ICI)-based treatment in a cancer patient at baseline, said method comprising:

[0013] (a) assessing the relative abundance of at least 2 microbial taxa in a stool sample derived from the patient at baseline, wherein at least 2 taxa are selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus; and

[0014] (b) stratifying the patient as a responder to the ICI-based treatment based on higher relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus; or, stratifying the patient as a non-responder to the ICI-based treatment based on higher relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus and / or lower relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus;

[0015] wherein determining the relative abundance as higher or lower is made dynamically within the workings of the highly non-linear model without reference to strictly determined thresholds.The advantages of this method are several-fold. This method allows the stratification of cancer patients to an ICI-based treatment regimen at baseline, wherein the patients have not received any treatments. This not only reduces the cost burden on the patients and the healthcare system but non-responders can also be saved from any severe side-effects of an unnecessary ICI-based treatment. This results in an improvement in the efficacy of the cancer care system whereby only patients who have been stratified as responders need to be administered an ICI-based treatment. This combined with the non-invasive stool sampling and the parsimonious model of the microbial taxa panel (maximum of 8 taxa) without requiring expensive strain-level data makes this a robust stratification method that can aid in clinical decision-making that saves lives and costs. The utility of 3 taxa at a higher taxonomic level -Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family is unexpected since markers at this level have not yielded patient stratification predictions that have any meaningful clinical utility in the literature due to the comparability of their relative abundances in the 2 patient strata of responders and non-responders. This further underlines the novelty of our non-linear prediction method that managed to meaningfully incorporate such information that have low-to-none predictive value by itself but improves the predictor’s performance while being a part of the model.

[0016] In some embodiments, the method of predicting the responsiveness to an Immune Checkpoint Inhibitor (ICI)-based treatment in a cancer patient at baseline, comprises:

[0017] (a) assessing the relative abundance of at least 2 microbial taxa in a stool sample derived from the patient at baseline, wherein the at least 2 taxa are selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus; and

[0018] (b) stratifying the patient as a responder to the ICI-based treatment based on higher relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus; or, stratifying the patient as a non-responder to the ICI-based treatment based on higher relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus and / or lower relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus;

[0019] wherein determining the relative abundance as high or low is made by the nonlinear model in a context-dependent manner.In some embodiments, the method of predicting the responsiveness to an Immune Checkpoint Inhibitor (ICI)-based treatment in a cancer patient at baseline, comprises:

[0020] assessing the relative abundance of at least 2 microbial taxa in a stool sample derived from the patient at baseline, wherein the at least 2 taxa are selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus; and (i) stratifying the patient as a responder to the ICI-based treatment based on higher relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus; or, stratifying the patient as a non-responder to the ICI-based treatment based on higher relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus and / or lower relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus; wherein determining the relative abundance as high or low is made by a non-linear model in a context-dependent manner, and / or

[0021] (ii) stratifying the patient as a responder to the ICI-based treatment based on a probability score being higher or lower than a pre-determined threshold, wherein the probability score is determined by using a look-up table generated from a non-linear model to match the relative abundance of the at least 2 microbial taxa to a probability score, wherein a cancer patient with a probability score above the pre-determined threshold is stratified as a responder, and a cancer patient with a probability score below the pre-determined threshold is stratified as a non-responder, and / or

[0022] (iii) stratifying the patient as a responder to the ICI-based treatment based on a probability score being higher or lower than a pre-determined threshold, wherein the probability score is determined using non-linear model to generate a probability score from the relative abundance values, wherein a cancer patient with a probability score above the predetermined threshold is stratified as a responder, and a cancer patient with a probability score below the pre-determined threshold is stratified as a non-responder.

[0023] In some embodiments, the look-up table is Table 3.

[0024] In some embodiments, the pre-determined threshold is 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, or 0.9. In some embodiments, the pre-determined threshold is a value selected within the range 0.1-0.9. In some embodiments, the pre-determined threshold is a value selected within the range 0.2-0.8. In some embodiments, the pre-determined threshold is a value selected withinthe range 0.3-0.7. Preferably, the pre-determined threshold is a value selected within the range 0.4-0.6. More preferably, the pre-determined threshold is 0.5.

[0025] In some embodiments, assessing the relative abundance comprises determining and / or measuring a relative abundance value as described herein for at least 2 microbial taxa in a stool sample derived from the patient at baseline, wherein the at least 2 taxa are selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus. In some embodiments the at least 2 taxa comprise all of the following taxa: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

[0026] In some embodiments, a method as described herein may further comprise a method of normalising microbial count data to determine the relative abundance of at least 2 microbial taxa described herein. The method of normalising microbial count data may comprise one or more of:

[0027] (a) receiving a plurality of microbial count data (features) associated with a biological sample,

[0028] (b) applying a centred log-ratio (CLR) transformation to the plurality of microbial count features,

[0029] (c) selecting a subset of the transformed features, and / or

[0030] (d) scaling each feature in the subset to a bounded numerical range using feature-specific scaling parameters predetermined from a reference dataset.

[0031] In some embodiments, applying a centred log-ratio transformation may comprise computing, for each feature of the data, a log-transformed value adjusted by an average of log-transformed values across all features in the sample. The CLR transformation method, as well as alternative transformation methods, will be apparent to the skilled person.

[0032] In some embodiments, applying the centred log-ratio transformation comprises computing, for each feature i: CLRi = log(counti + c) - meanj(log(countj + c)), wherein c is a pseudocount and the mean is computed over all features in the sample. In some embodiments, the pseudocount c is a constant selected to avoid undefined log values. In some embodiments, c = 0.5.

[0033] In some embodiments, the mean is computed across some or all taxa present in the microbial count data, irrespective of whether such taxa form part of the selected subset.

[0034] In some embodiments, selecting the subset comprises extracting a predetermined set of model features from the CLR-transformed feature vector. In some embodiments, the predetermined set comprises features selected during a model training stage. In someembodiments, the predetermined set of model features comprises at least 2 taxa selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

[0035] In some embodiments, scaling each feature to the bounded range comprises applying min-max normalisation using a feature-specific minimum and maximum value derived from a discovery or training cohort. In some embodiments, the scaling is performed according to x_normalised = clamp((CLRi - mini) I (maxi - min), 0, 1), wherein min and maxi are frozen values corresponding to the discovery cohort. In some embodiments, this clamping function prevents scaled values from exceeding the predetermined numerical range. In preferred embodiments, the bounded numerical range is [0,1], In some embodiments, the feature-specific minimum and maximum values correspond to the CLR boundaries defined in Table 2.

[0036] In some embodiments, a method as described herein further comprises quantizing microbial count data, preferably the normalised microbial count data. In some embodiments, the normalized abundance values are subjected to a frugal adaptive quantization procedure configured to identify informative regions of each feature's value distribution, optionally wherein the quantization intervals are non-uniform across all features. Preferably, the method comprises performing a frugal quantization of the normalised abundance values using the key abundance values defined in Table 1.

[0037] In some embodiments, the method is a computer-implemented method.

[0038] Also provided herein is a system or non-transitory computer-readable medium storing instructions configured to perform one or more methods described herein.

[0039] In some embodiments, the patient may have hepatocellular carcinoma (HCC). Given the objective response rates of HCC to ICI-based treatments such as nivolumab and pembrolizumab in the art are in the range of 15-20%, adoption of the disclosed method contributes to the improvement of treatment outcomes in the field of ICI-based treatments for HCC.

[0040] In some embodiments of the invention, the method predicts responsiveness to an ICI-based treatment wherein the ICI-based treatment comprises at least one ICI to treat cancer patients, wherein, in some preferred embodiments, the patient may have HCC. More preferably, the at least one ICI is an anti-PD1 agent. Most preferably, the anti-PD1 agent may be nivolumab or pembrolizumab.In some embodiments, the method predicts responsiveness in cancer patients such as HCC patients to an ICI-based treatment wherein the ICI-based treatment has at least one ICI and that ICI may be an anti-PDL1 agent In some embodiments, the at least one ICI may be an anti-CTLA agent. In some embodiments, the ICI-based treatment may comprise a combination of 2 ICIs such as an anti-PD1 agent and an anti-CTLA-4 agent.

[0041] In some embodiments, the method predicts responsiveness in cancer patients to an ICI-based treatment wherein the ICI-based treatment may further comprise Molecular Targeted Therapy (MTT) in addition to the at least 1 ICI. Preferably, the MTT comprises one or more therapies selected from a group consisting of: sorafenib, lenvatinib, apatinib and anlotinib. Preferably, the cancer patients that are predicted for their responsiveness to such ICI-based treatments have HCC.

[0042] In some embodiments, the method predicts responsiveness to an ICI-based treatment wherein the ICI-based treatment may further comprise a Transarterial Chemoembolisation (TACE) therapy in addition to the at least 1 ICI. In some embodiments, the ICI-based treatment may further comprise a combination of MTT and TACE therapy in addition to the at least 1 ICI. Preferably, the cancer patients whose responsiveness to such ICI-based treatments is predicted have HCC.

[0043] In some embodiments, the method may further comprise the step of treating the cancer patients such as HCC patients stratified as responders to the ICI-based treatment with an effective ICI-based treatment. Preferably, the effective ICI-based treatment comprises an anti-PD1 agent. More preferably, the anti-PD1 agent is nivolumab or pembrolizumab.

[0044] In some embodiments, the method may comprise stratifying the HCC patient at baseline as a responder to the ICI-based treatment based on lower relative abundance of Acidaminococcaceae family and lower relative abundance of Ligilactobacillus genus in the stool sample and stratifying the patient as a non-responderto the ICI-based treatment is based on higher relative abundance of Acidaminococcaceae family and higher relative abundance of Ligilactobacillus genus in the stool sample, wherein determining the relative abundance as higher or lower is made by the non-linear model in a context-dependent manner.

[0045] In a second aspect of the invention, there is provided a method of treatment comprising, administering to a cancer patient with an effective ICI-based treatment upon stratifying the patient at baseline as a responder to the ICI-based treatment based on at least 2 of the following: higher relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus in a stoolsample from the patient at baseline, wherein determining the relative abundance as higher or lower is made by the non-linear model in a context-dependent manner.

[0046] This method of treatment offers several advantages. Firstly, only patients who have been predicted to be responders to effective ICI-based treatments are treated with an effective ICI-based treatment thereby reducing the cost burden on patients who may unnecessarily be administered these treatments. This method of treatment eliminates the pool of nonresponders from ICI-based treatments, thereby reducing the occurrence of severe autoimmune reactions in such patients that may occur with ICI-based treatment administration. Furthermore, the treatment method arms clinicians with an ICI-based treatment intervention that can be administered with confidence to responders since the stratification steps are performed on the patients at baseline and are based on the innate composition of the gut microbiome, rather than treatment- induced microbiome perturbations.

[0047] Also provided herein is a method of treatment comprising, administering to a cancer patient an effective ICI-based treatment, wherein the patient has been:

[0048] (i) stratified at baseline as a responder to the ICI-based treatment based on at least 2 of the following: higher relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus in a stool sample from the patient at baseline, wherein determining the relative abundance as high or low is made by the non-linear model in a context-dependent manner, and / or

[0049] (ii) stratified at baseline as a responder to the ICI-based treatment based on a probability score being higher or lower than a pre-determined threshold, wherein the probability score is determined using a look-up table generated from a non-linear model to match relative abundance values of least 2 microbial taxa in a stool sample derived from the patient to a probability score, the at least 2 taxa selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus;

[0050] wherein a cancer patient with a probability score above the pre-determined threshold is stratified as a responder, and a cancer patient with a probability score below the predetermined threshold is stratified as a non-responder.

[0051] Also provided is an ICI-based medicament, such as an ICI, for use in a method of treating cancer, the method comprising administering to a cancer patient the ICI-based treatment, wherein the patient has been:(i) stratified at baseline as a responder to the ICI-based treatment based on at least 2 of the following: higher relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus in a stool sample from the patient at baseline, wherein determining the relative abundance as high or low is made by the non-linear model in a context-dependent manner, and / or

[0052] (ii) stratified at baseline as a responder to the ICI-based treatment based on a probability score being higher or lower than a pre-determined threshold, wherein the probability score is determined by using a look-up table generated from a non-linear model to match relative abundance values of least 2 microbial taxa in a stool sample derived from the patient to a probability score, the at least 2 taxa selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

[0053] In some embodiments of the method of treatment, or the medicament for use, the look-up table is Table 3.

[0054] In some embodiments of the method of treatment or the medicament for use, the predetermined threshold is a value selected within the range 0.4-0.6. Preferably, the predetermined threshold is 0.5.

[0055] Importantly, the exact threshold that would be chosen in a particular clinical application is a question of clinical values rather than statistical properties. While Figure 5 shows an example - formally statistically optimized - choice of the specificity I sensitivity tradeoff based on maximizing the Youden index (Fluss et al., 2005), and that optimization corresponded to predicted probability threshold of 0.578, this approach treats the value of false positives and false negatives equally, and this deviates from the clinical dilemma of balancing non-detected responders that might benefit from the treatment but would be missed, on one hand, and nonresponders falsely called responders and subjected to wasteful procedure. The specific operational decision threshold should be chosen by the clinician based on consideration of patients’ quality-of-life values.

[0056] In some embodiments, interpolation between lookup-table vertices is performed directly in the bounded (0, 1) probability space. In some embodiments, interpolation between lookup-table vertices is performed in unbounded (-«, +°°) logit space. In some embodiments, the stored outcome probability values are converted to logits (log-probabilities) prior to N-linear interpolation, and the interpolated result is mapped back to a valid probability distribution via the Softmax transformation (Bishop, 2006).In some embodiments, the patient has HCC. Given the objective response rates of HCC to ICI-based treatments such as nivolumab and pembrolizumab in the art are in the 15-20% bracket, adoption of the disclosed method of treatment contributes to the improvement of objective response rates since the treatment is based on a patient stratification method that can select for responders to ICI-based treatments.

[0057] In some embodiments of the invention, the effective ICI-based treatment method comprises at least one ICI to treat cancer patients such as HCC patients. In some preferred embodiments, the effective ICI-based treatment is administered to patients that have HCC. More preferably, the at least one ICI is an anti-PD1 agent. Most preferably, the anti-PD1 agent may be nivolumab or pembrolizumab.

[0058] In some embodiments, the ICI-based treatment method has at least one ICI. In some embodiments, that ICI may be an anti-PDL1 agent. In some embodiments, the at least one ICI may be an anti-CTLA agent. In some embodiments, the ICI-based treatment may comprise a combination of 2 ICIs such as an anti-PD1 agent and an anti-CTLA-4 agent. In a preferred embodiment, the patients administered such therapies have HCC.

[0059] In some embodiments, the ICI-based treatment may further comprise Molecular Targeted Therapy (MTT) in addition to the at least 1 ICI. Preferably, the MTT comprises one or more therapies selected from a group consisting of: sorafenib, lenvatinib, apatinib and anlotinib. Preferably, the cancer patients that are administered such treatments have HCC.

[0060] In some embodiments, the ICI-based treatment method may further comprise a Transarterial Chemoembolisation (TACE) therapy in addition to the at least 1 ICI. In some embodiments, the ICI-based treatment may further comprise a combination of MTT and TACE therapy in addition to the at least 1 ICI. Preferably, the cancer patients administered such treatments have HCC.

[0061] A third aspect of the invention provides for the use of at least 1 ICI in the manufacture of a medicament for the therapeutic treatment of cancer upon stratifying a cancer patient at baseline as a responder to an ICI-based treatment based on at least 2 of the following: higher relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus in a stool sample from the patient at baseline, wherein determining the relative abundance as high or low is made by the non-linear model in a context-dependent manner. More preferably, at least 1 ICI that is used in the manufacture of a medicament is an anti-PD1 agent and the cancer patient that is therapeutically treated by such a medicament has HCC.BRIEF DESCRIPTION OF THE FIGURES

[0062] Figure 1 - Biomarkers importance derived from SHAP analysis (A) Comparison of the 8 biomarkers identified in this study with the 18 biomarkers from Wu et al. Biomarkers are arranged in decreasing importance, with exact matches, cross-rank matches, and high phylogenetic closeness (with a possibility of the de facto exact match given the phylogenetic uncertainty and evolving classification) indicated. Our 8 biomarkers include Acidaminococcaceae, Ligilactobacillus, Pasteurellaceae, Dorea, Barnesiella, Tannerellaceae, Ruminococcus, and Atopobium, highlighting overlapping and distinct taxa compared to Wu et al. (B) SHAP analysis beeswarm plot illustrating the impact of the 8 biomarkers on the model’s output. Each point represents a SHAP value for a single prediction, with color indicating biomarker abundance after transformation and min-max normalization (purple for high abundance and blue for low abundance). Higher positive SHAP values indicate a stronger contribution towards predicting responder status, while higher negative SHAP values indicate a stronger contribution towards predicting non-responder status. The points are an aggregation of predictions from the validation samples, generated from the top 37 models out of 150 models, independently trained with heuristic optimized bootstrapped train dataset. Figure 2 - (A) Violin plots illustrating the distribution of AU ROC scores across 100 models for different sample groups: all samples, samples before intervention (baseline), samples before the third dose of anti-PD-1, and samples from the third dose onwards. The horizontal lines within the violins indicate the median performance. (B) Heatmap of the predicted responder(R) / non-responder (NR) status based on the probability score provided by the model with a threshold of 0.5. The probabilities are the mean of 100 trials. The heatmap includes 155 samples from all 35 patients in both the train and test sets. The bottom horizontal heatmap shows the ground truth responder status based on clinical evaluation, with pink denoting responders and blue denoting non-responders. The patient identifiers are listed along the x-axis, and the number of doses is represented on the y-axis

[0063] Figure 3 - ROC curves of the original model (black) and LUT-based predictions. (A) the entire dataset from Wu et al., 2022, (B) test split only.

[0064] Figure 4 - Scatter plots of the LUT-based prediction values versus the values output by the original model. Wu et al., 2022 dataset. (A) all data, (B): test split only. Dots are colored by the reported therapy response. Non-responder (NR) pink. Responder (R) blue.

[0065] Figure 5 - An end-to-end LUT-based prediction pipeline test: from separate unnormalized tables of genus and family-level counts to each sample’s associated therapy response probability.DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] Definitions

[0067] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as is commonly understood by one of skill in the art to which the invention belongs. Certain terms employed in the specification, examples and appended claims are collected here for convenience.

[0068] As used herein, the term “comprising” does not preclude the presence of additional steps or substances in the methods and compositions, respectively, of the invention, and is understood to include within its scope the terms ‘consisting of’ and ‘consisting essentially of’ features defined in the claimed invention.

[0069] The terms “patient” and “patients” include references to mammalian (e.g. human) patients. An Immune Checkpoint Inhibitor (ICI) desginates any drug, molecule or composition that inhibits or blocks an immune checkpoint. It encompasses anti-PD1 agents, anti-PDL1- agents, anti-PDL2 agents and anti-CTLA-4 agents. Anti-PD1 agents, anti-PDL1 agents and anti-CTLA-4 agents encompass any active molecules that angonize PD1, PDL1 and CTLA-4 respectively. These agents may be antibodies such as Nivolumab (anti-PD1 agent) , Pembrolizumab (anti-PD1 agent), Atezolizumab (anti-PDL1 agent), Avelumab (anti-PDL1 agent) and Ipilimumab (anti-CTLA-4 agent). These agents also include molecules such as antibody fragments or aptamers that antagonize PD1, PDL1 and CTLA-4.

[0070] An “ICI-based treatment” encompasses all cancer treatments I cancer treatment plans I cancer treatment regimens that include one or more doses of an ICI. ICI-based treatment includes all combination therapies that have at least 1 ICI as a component, irrespective of the sequence of administration of such combination therapies.

[0071] “Responders” are defined as patients with complete response, partial response and stable disease lasting more than 6 months, while “non-responders” were defined as patients with progressive disease and stable disease with less than 6 months.

[0072] “Molecular Targeted Therapy (MTT)” encompasses all drugs or other substances to target specific molecules that cancer cells need to survive and spread. MTTs use a variety of mechanisms to stop cancer cells from growing such as interrupting signals that cause them to grow and divide, stopping signals that help form blood vessels, delivering cell-killing substances to cancer cells, starving cancer cells of hormones they need to grow or helping the immune system kill cancer cells. Some examples of MTTs are monoclonal antibodies such as bevacizumab that targets VEGF, trastuzumab that targets HER2 and cetuximab that targetsEGFR, rituximab that targets CD20. Other MTTs include small molecules such as dasatinib, imatinib, lapatinib, lenvatinib, nilotinib, regorafenib, sorafenib, sunitinib, vemurafenib, cabozantinib, ceritinib, gefitinib, palbociclib, pazopanib, ponatinib, ruxolitinib, tofacitinib, trametinib, everolimus, sirolimus, temsirolimus, afatinib, ibrutinib and lenvatinib.

[0073] The term "cancer" includes but is not limited to, breast cancer, large intestinal cancer, lung cancer, small cell lung cancer, gastric (stomach) cancer, liver cancer, hepatocellular cancer, blood cancer, bone cancer, pancreatic cancer, skin cancer, head and / or neck cancer, cutaneous or intraocular 26 melanoma, uterine sarcoma, ovarian cancer, rectal or colorectal cancer, anal cancer, colon cancer, fallopian tube carcinoma, endometrial carcinoma, cervical cancer, vulval cancer, squamous cell carcinoma, vaginal carcinoma, Hodgkin's disease, nonHodgkin's lymphoma, esophageal cancer, small intestine cancer, endocrine cancer, thyroid cancer, parathyroid cancer, adrenal cancer, soft tissue tumour, urethral cancer, penile cancer, prostate cancer, chronic or acute leukemia, lymphocytic lymphoma, bladder cancer, kidney cancer, ureter cancer, renal cell carcinoma, renal pelvic carcinoma, CNS tumour, glioma, astrocytoma, glioblastoma multiforme, primary CNS lymphoma, bone marrow tumour, brain stem nerve gliomas, pituitary adenoma, uveal melanoma (also known as intraocular melanoma), testicular cancer, oral cancer, pharyngeal cancer or a combination thereof.

[0074] “Baseline” refers to a sample, patient, or subject before to intervention with an ICI based treatment. That is, before treatment with an ICI based composition, or composition comprising an ICI.

[0075] Herein, a “probability score” refers to a numerical value representing the likelihood of an event occurring, expressed on a scale from 0 (impossible) to 1 (certain). Likelihood is the quantitative measure of how probable it is that a particular event or outcome will occur. For example, a probability score as to whether a sample, subject or patient is a responder or non-responder to an ICI-based treatment can be determined using a look-up table, which matches relative abundance of microbial taxa to a probability score. The look up table is generated using a non linear model.

[0076] All references to methods of treatment by therapy or surgery in the following description are to also be interpreted as references to compounds and compositions of the present invention for use in those methods.

[0077] The embodiments disclosed herein relate to a method of stratifying cancer patients at baseline into 2 classes, responders and non-responders with respect to their responsiveness to Immune Checkpoint lnhibitor(ICI)-based treatments. The stratification method is based on assessing the relative abundances of at least 2 of 8 microbial taxa - Acidaminococcaceaefamily, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

[0078] The stratification as a responder to the ICI-based treatment is made based on the presence at least 2 of the following taxa-related features: higher relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus in a stool sample of the patient at baseline. The determination of high or low abundance is made with reference to a threshold.

[0079] The stratification as a non-responder to the ICI-based treatment is made based on the presence of at least 2 of the following taxa-related features: higher relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus and / or lower relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus in a stool sample of the patient at baseline. The determination of high or low abundance is made by the non-linear model in a context-dependent manner.

[0080] The stratification of a cancer patient as a responder to the ICI-based treatment may be made in any of the possible combinations of 2 or 3 or 4 or 5 or 6 or 7 or all 8 of the taxa-related features based on assessing the relative abundance of the taxa in a stool sample derived from the cancer subject at baseline.

[0081] 1. Higher relative abundance of Pasteurellaceae family;

[0082] 2. Higher relative abundance of Dorea genus;

[0083] 3. Higher relative abundance of Barnesiella genus;

[0084] 4. Higher relative abundance of Ruminococcus genus;

[0085] 5. lower relative abundance of Acidaminococcaceae family;

[0086] 6. lower relative abundance of Ligilactobacillus genus;

[0087] 7. lower relative abundance of Tannerellaceae family;

[0088] 8. lower relative abundance of Atopobium genus

[0089] The stratification of a cancer patient as a non- responder to the ICI-based treatment may be made in any of the possible combinations of 2 or 3 or 4 or 5 or 6 or 7 or 8 of the following taxa-related features based on assessing the relative abundance of the taxa in a stool sample derived from the cancer subject at baseline:

[0090] 1. lower relative abundance of Pasteurellaceae family;

[0091] 2. lower relative abundance of Dorea genus;

[0092] 3. lower relative abundance of Barnesiella genus;4. lower relative abundance of Ruminococcus genus;

[0093] 5. higher relative abundance of Acidaminococcaceae family;

[0094] 6. higher relative abundance of Ligilactobacillus genus;

[0095] 7. higher relative abundance of Tannerellaceae family;

[0096] 8. higher relative abundance of Atopobium genus

[0097] An array of embodiments relate to various combinations of 2 or 3 or 4 or 5 or 6 or 7 or 8 taxa-related features recited in the invention whose relative abundances may be assessed in order to predict the responders amongst HCC patients to ICI-based treatments. Some embodiments that stratify the HCC patient as a responder to ICI-based treatments are based on the following features of the taxa present in the stool sample:

[0098] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus

[0099] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus

[0100] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus

[0101] higher relative abundance of Pasteurellaceae family, lower relative abundance of Acidaminococcaceae family

[0102] higher relative abundance of Pasteurellaceae family, lower relative abundance of Ligilactobacillus genus

[0103] higher relative abundance of Pasteurellaceae family, lower relative abundance of Tannerellaceae family

[0104] higher relative abundance of Pasteurellaceae family, lower relative abundance of Atopobium genus

[0105] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family

[0106] higher relative abundance of Dorea genus, lower relative abundance of Ligilactobacillus genus higher relative abundance of Dorea genus, lower relative abundance of Tannerellaceae family higher relative abundance of Dorea genus, lower relative abundance of Atopobium genushigher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus

[0107] higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family

[0108] higher relative abundance of Barnesiella genus, lower relative abundance of Ligilactobacillus genus

[0109] higher relative abundance of Barnesiella genus, lower relative abundance of Tannerellaceae family

[0110] higher relative abundance of Barnesiella genus, lower relative abundance of Atopobium genus higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family

[0111] higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus

[0112] higher relative abundance of Ruminococcus genus, lower relative abundance of Tannerellaceae family

[0113] higher relative abundance of Ruminococcus genus, lower relative abundance of Atopobium genus

[0114] lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0115] lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0116] lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0117] lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0118] lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0119] lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genushigher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus

[0120] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus

[0121] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family

[0122] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Ligilactobacillus genus

[0123] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Tannerellaceae family

[0124] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Atopobium genus

[0125] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus

[0126] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family

[0127] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Ligilactobacillus genus

[0128] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Tannerellaceae family

[0129] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Atopobium genus

[0130] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family

[0131] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus

[0132] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Tannerellaceae family

[0133] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Atopobium genushigher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus

[0134] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family

[0135] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Ligilactobacillus genus

[0136] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Tannerellaceae family

[0137] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Atopobium genus

[0138] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family

[0139] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus

[0140] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Tannerellaceae family

[0141] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Atopobium genus

[0142] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family

[0143] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus

[0144] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Tannerellaceae family

[0145] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Atopobium genus

[0146] higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0147] higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae familyhigher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0148] higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0149] higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0150] lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0151] lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0152] lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0153] lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0154] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus

[0155] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family

[0156] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Ligilactobacillus genus

[0157] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Tannerellaceae family

[0158] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Atopobium genus

[0159] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae familyhigher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus

[0160] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Tannerellaceae family

[0161] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Atopobium genus

[0162] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0163] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0164] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0165] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0166] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0167] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0168] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae familyhigher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus

[0169] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Tannerellaceae family

[0170] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Atopobium genus

[0171] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0172] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0173] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0174] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0175] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0176] higher relative abundance of Pasteurellaceae family, higher relative abundance of Barnesiella genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0177] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genushigher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0178] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0179] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0180] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0181] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family

[0182] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus

[0183] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Tannerellaceae family

[0184] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Atopobium genus

[0185] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0186] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae familyhigher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0187] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0188] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0189] higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0190] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0191] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0192] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0193] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0194] higher relative abundance of Dorea genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0195] higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae familyhigher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0196] higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0197] higher relative abundance of Dorea genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0198] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0199] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0200] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0201] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0202] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0203] higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0204] higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genushigher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0205] higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0206] higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0207] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family

[0208] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus

[0209] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Tannerellaceae family

[0210] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Atopobium genus

[0211] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0212] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0213] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genushigher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0214] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0215] higher relative abundance of Pasteurellaceae family, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0216] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0217] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0218] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0219] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0220] higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0221] higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genushigher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0222] higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0223] higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0224] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus

[0225] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0226] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0227] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0228] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0229] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genushigher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Tannerellaceae family

[0230] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Atopobium genus

[0231] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0232] higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0233] higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0234] higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus

[0235] higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family higher relative abundance of Pasteurellaceae family, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus higher relative abundance of Dorea genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0236] higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae familyhigher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Atopobium genus higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family

[0237] higher relative abundance of Barnesiella genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus higher relative abundance of Pasteurellaceae family, higher relative abundance of Dorea genus, higher relative abundance of Barnesiella genus, higher relative abundance of Ruminococcus genus, lower relative abundance of Acidaminococcaceae family, lower relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus.

[0238] A further array of embodiments pertain to methods of treating the HCC patients who have been stratified as a responder to ICI-based treatments based on any of the above combinations of features with an effective ICI-based treatment. Some examples of ICI-based treatments are treatments that comprise at least 1 ICI or combination therapies such as combinations of one or more ICIs and MTT or one or more ICIs and TACE therapy or combinations of one or more ICIs + MTT + TACE therapy. The ICI in such embodiments may be anti-PD1 agents I anti-PDL1 agents I anti-PDL2 agents / anti-CTLA-4 agents. The anti-PD1 agents may be nivolumab or pembrolizumab.

[0239] An array of embodiments relate to various combinations of 2 or 3 or 4 or 5 or 6 or 7 or 8 taxa-related features recited in the invention whose relative abundances may be assessed in order to predict the non-responders amongst HCC patients to ICI-based treatments. Some embodiments that stratify the HCC patient as a non-responder to ICI-based treatments are based on the following features of the taxa present in the stool sample:

[0240] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genuslower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus

[0241] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus

[0242] lower relative abundance of Pasteurellaceae family, higher relative abundance of Acidaminococcaceae family

[0243] lower relative abundance of Pasteurellaceae family, higher relative abundance of Ligilactobacillus genus

[0244] lower relative abundance of Pasteurellaceae family, higher relative abundance of Tannerellaceae family

[0245] lower relative abundance of Pasteurellaceae family, higher relative abundance of Atopobium genus

[0246] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus lower relative abundance of Dorea genus, higher relative abundance of Acidaminococcaceae family

[0247] lower relative abundance of Dorea genus, higher relative abundance of Ligilactobacillus genus lower relative abundance of Dorea genus, higher relative abundance of Tannerellaceae family lower relative abundance of Dorea genus, higher relative abundance of Atopobium genus lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus

[0248] lower relative abundance of Barnesiella genus, higher relative abundance of Acidaminococcaceae family

[0249] lower relative abundance of Barnesiella genus, higher relative abundance of Ligilactobacillus genus

[0250] lower relative abundance of Barnesiella genus, higher relative abundance of Tannerellaceae family

[0251] lower relative abundance of Barnesiella genus, higher relative abundance of Atopobium genuslower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family

[0252] lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus

[0253] lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0254] lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genus

[0255] higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus

[0256] higher relative abundance of Acidaminococcaceae family, higher relative abundance of Tannerellaceae family

[0257] higher relative abundance of Acidaminococcaceae family, higher relative abundance of Atopobium genus

[0258] higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family

[0259] higher relative abundance of Ligilactobacillus genus, higher relative abundance of Atopobium genus

[0260] higher relative abundance of Tannerellaceae family, higher relative abundance of Atopobium genus

[0261] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus

[0262] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus

[0263] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Acidaminococcaceae family

[0264] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Ligilactobacillus genus

[0265] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Tannerellaceae familylower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Atopobium genus

[0266] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus

[0267] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, higher relative abundance of Acidaminococcaceae family

[0268] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, higher relative abundance of Ligilactobacillus genus

[0269] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, higher relative abundance of Tannerellaceae family

[0270] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, higher relative abundance of Atopobium genus

[0271] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family

[0272] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus

[0273] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0274] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genus

[0275] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus

[0276] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Acidaminococcaceae family

[0277] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Ligilactobacillus genus

[0278] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Tannerellaceae family

[0279] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Atopobium genuslower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family

[0280] lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus

[0281] lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0282] lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genus

[0283] lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family

[0284] lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus

[0285] lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0286] lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genus

[0287] lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus

[0288] lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Tannerellaceae family

[0289] lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Atopobium genus

[0290] lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family

[0291] lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Atopobium genus

[0292] higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family

[0293] higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Atopobium genushigher relative abundance of Acidaminococcaceae family, higher relative abundance of Tannerellaceae family, higher relative abundance of Atopobium genus

[0294] higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family, higher relative abundance of Atopobium genus

[0295] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus

[0296] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Acidaminococcaceae family

[0297] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Ligilactobacillus genus

[0298] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Tannerellaceae family

[0299] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Atopobium genus lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family

[0300] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus

[0301] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0302] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genuslower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus

[0303] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Tannerellaceae family

[0304] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Atopobium genus

[0305] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family

[0306] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Atopobium genus

[0307] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Tannerellaceae family, higher relative abundance of Atopobium genus

[0308] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family

[0309] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus

[0310] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0311] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genuslower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus

[0312] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Tannerellaceae family

[0313] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Atopobium genus

[0314] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family

[0315] lower relative abundance of Pasteurellaceae family, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Atopobium genus

[0316] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family

[0317] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus

[0318] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0319] lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genus

[0320] lower relative abundance of Dorea genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genuslower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family

[0321] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus

[0322] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0323] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genus

[0324] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus

[0325] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Tannerellaceae family

[0326] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Atopobium genus

[0327] lower relative abundance of Pasteurellaceae family, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus

[0328] lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus

[0329] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genuslower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Tannerellaceae family

[0330] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Atopobium genus

[0331] lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus

[0332] higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family

[0333] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family

[0334] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Ligilactobacillus genus, lower relative abundance of Tannerellaceae family, lower relative abundance of Atopobium genus

[0335] lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family

[0336] lower relative abundance of Pasteurellaceae family, lower relative abundance of Dorea genus, lower relative abundance of Barnesiella genus, lower relative abundance of Ruminococcus genus, higher relative abundance of Acidaminococcaceae family, higher relative abundance of Ligilactobacillus genus, higher relative abundance of Tannerellaceae family, higher relative abundance of Atopobium genus.

[0337] The assessment of the relative abundance of at least 2 taxa is made by measuring the number of copies of at least 1,2,5,10,15,20 or 25 sequences specific for the at least 2 taxa. Methods to measure the relative abundance such as PCR, RT-PCR, QRT-PCR, sequencing methods ( NGS) or hybridization are known to persons skilled in the art. In some embodiments, an RT-PCR measures the relative abundance of at least 2 taxa through the use of primers that can amplify at least 1,2,5,10,20 or 25 sequences specific for the at least 2 taxa. The Ct values ofthe at least 2 taxa from an RT-PCR may be used to compute the relative abundances of the at least 2 taxa.

[0338] Some embodiments of the invention also pertain to responsiveness prediction kits that can predict responsiveness to ICI-based treatments based on an RT-PCR analysis of at least 2 out of the 8 taxa recited in this invention. Such kits comprise a panel of PCR primers and reagents to conduct an RT-PCR analysis. Given that different combinations are feasible in terms of the number of taxa to be assessed such as 2 out of 8, 3 out of 8, 4 out of 8, 5 out of 8, 6 out of 8, 7 out of 8 or 8 out of 8, a person skilled in the art will understand that the prediction kit can have PCR primer panels that are tailor-made for each of these embodiments. In some embodiments, such PCR primer panels may comprise primers that can amplify the omega subunit of RNA polymerase that are specific for 2 out of 8, 3 out of 8, 4 out of 8, 5 out of 8, 6 out of 8, 7 out of 8 or 8 out of the 8 taxa recited in this invention. In a preferred embodiment, the responsiveness prediction kit comprises PCR primers that can amplify sequences specific for any 2 or 3 or 4 or 5 or 6 or 7 or 8 of the recited taxa with which a person skilled in the art may conduct an RT-PCR measurement to stratify HCC patients as responders and nonresponders to ICI-based treatments.

[0339] While "higher abundance" and "lower abundance" statements are expected to imply certain quantitative thresholds, the peculiarity of the operation of our model which is based on highly non-linear relationships between the biomarkers and accounting for the significant person-to-person variations in the general taxonomy distribution of the gut microbes (partially mapping to the concepts of "enterotypes" and going beyond that), we are not providing the particular threshold values for any biomarker (that would be only possible of the relationships between the biomarker levels and the response prediction score were more simple than they are). Instead, we supply a look-up table (LUT) that takes the relative abundances of the biomarkers used by the deep model and outputs the resulting prediction score approximating the scores output by the actual Al model without exposing the internal workings of the model. The availability of LUT makes it straightforward to use the biomarker abundance information in clinical practice.

[0340] In some of the methods disclosed herein, stratifying the patient as a responder to the ICI-based treatment comprises determining a probabilistic score using a look-up table (LUT). That is, a probabilistic score can be determined using an input parameter of relative abundance of a microbial taxa in a stool sample and a predefined lookup table that maps the input values or ranges to corresponding probability values. The method enables rapid, repeatable probability determination without requiring real-time statistical computation. In some embodiments, the look-up table maps the relative abundance of at least two of : Acidaminococcaceae family,Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus to a probability score. In some embodiments, the lookup table provides outputs determined by a machine learning model. In some embodiments, the look-up table is generated from a non-linear model. In some embodiments, the look-up table values are as set out in Table 3.

[0341] In some embodiments, the patient is stratified, or has been stratified, based on whether the probability score is higher or lower than a pre-determined threshold value. In some embodiments, a cancer patient with a probability score above the pre-determined threshold is stratified as a responder, and a cancer patient with a probability score below the predetermined threshold is stratified as a non-responder. In some embodiments, the predetermined threshold is 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, preferably wherein the predetermined threshold is 0.5. That is, in some embodiments, a subject or patient having a probability score greater than 0.5, as determined by a look up table and / or non linear model, is stratified as a responder. In some embodiments, a subject or patient having a probability score less than 0.5, as determined by a look up table and / or non linear model, is stratified as a non-responder.

[0342] In some embodiments, high relative abundance of microbial taxa of, for example, Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus and / or lower relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus FAF1, is correlated with an increased likelihood of patient or subject having a probability score above a threshold value and so an increased likelihood of patient or subject being a responder to the ICI-based treatment.

[0343] In some embodiments, higher relative abundance of Acidaminococcaceae family, Ligilactobacillus genus, Tannerellaceae family and / or Atopobium genus and / or lower relative abundance of Pasteurellaceae family, Dorea genus, Barnesiella genus and / or Ruminococcus genus; is correlated with an increased likelihood of patient or subject having a probability score below a threshold value and so an increased likelihood of patient or subject being a non-responder to the ICI-based treatment.

[0344] Example 1 - METHODS

[0345] We used the publicly available data from Wu et al. (Wu et al. 2022), a study of 35 patients (31 males, 4 females, mean age of 53.8 years) with hepatocellular carcinoma (HCC) who undergo anti-PD-1 -based immunotherapy. The majority of the patients were in the advanced stage ofHCC, classified as Barcelona Clinic Liver Cancer stage C (32 / 35, 91.45%). The distinction between Responders (R, 46% of the cohort) and Non-responders (NR, 54%) was performed in the original publication on the basis of a combination of complete response (CR), a partial response (PR), or stable disease (SD) labels and the follow-up time. This 46% / 54% distribution shows a relatively balanced dataset, allowing for robust comparative analyses. Notably, the R and NR groups were also balanced by age, gender, liver function status, and underlying etiologies, with the exception of baseline alphafetoprotein (AFP) levels, immature granulocyte counts, and lymphocyte counts. The raw 16S rRNA gene sequencing data representing this study in the National Genomics Data Center (NGDC) Genome Sequence Archive (GSA) database (accession: CRA005195) is accompanied by a limited metadata file with only two columns containing the patients’ data: “Public description” with R / NR status, and “Collection date”, apart from other non-informative or sequencing-specific data. Applying the knowledge of the three-week interval between fecal sample collections, we were able to match the samples to all 35 patients.

[0346] Raw sequencing data (FASTQ files) representing the above dataset were downloaded from GSA and re-processed using a custom 16S amplicon analysis workflow, using the year 2023 update of the taxonomy databases. Low-level preprocessing of the data was performed with modifications relevant to the optimization of the predictive model’s performance. The feature abundance data were transformed, and data analysis was performed using a combination of R statistical environment and Python programming language, and a combination of custom scripting with open-source packages.

[0347] Custom ML model construction was performed using Python programming language and open-source packages. Like a typical modelling workflow, the ML model construction involved steps such as the Train Test Split procedure, feature selection, constructing a neural network, SHAP-based feature refinement and evaluating the model using AUROC scores.

[0348] EXAMPLE 2 - INSIGHTS INTO THE TAXA AND CLINICAL UTILITY

[0349] Our 8 biomarkers included 3 taxonomy features at the family level (Acidaminococcaceae, Pasteurellaceae and Tannerellaceae) and 5 - at the genus level (Ligilactobacillus, Dorea, Barnesiella, Ruminococcus and Atopobium). In contrast, Wu et al.’s model exclusively utilized biomarkers from the genus level. All three of our family-level biomarkers have corresponding genus-level biomarkers (one for each family) used in Wu et al.’s model (Figure 1A). These family-level markers have the potential to capture a broader range of bacterial species with similar biological properties and predictive values, overcoming the granularity of genus-level features across individuals.Notably, the most informative marker based on SHAP values in our study, Acidaminococcaceae, maps to two taxa at the genus level, Acidaminococcus and Phascolarctobacterium, with only the latter identified as a key microbial biomarker (though the least important) in Wu et al.’s study. This indicates that our all-taxonomic-rank approach could take advantage of features encompassing a broader microbial diversity.

[0350] For our genus-level biomarkers, three (Dorea, Barnesiella and Ruminococcus) overlapped with Wu et al.’s key microbial biomarkers, highlighting their significance and providing convergence in the identification of crucial biomarkers. Interestingly, our method identified Ligilactobacillus as the second most informative biomarker, whereas Wu et al.’s model identified Lactobacillus as the most important biomarker. Ligilactobacillus genus is taxonomically close to Lactobacillus and was historically reclassified from the latter. Similar to Lactobacillus, it also has a substantial track record of probiotic properties (Yang et al. 2024). Taxonomic closeness and the respective functional overlap between the two genera suggest that both models could be capturing the same underlying biological signal. The SHAP value plot provides insights into the contribution and impact of each biomarker on our model’s output (Figure 1B). The top contributors to the model’s predictions, Acidaminococcaceae and Ligilactobacillus, are shown to be negatively associated with responder status. Tannerellaceae and Atopobium also show a negative association with responders while biomarkers, as evidenced by their negative SHAP values. Conversely, Pasteurellaceae, Dorea, Barnesiella and Ruminococcus positively contributed to the model’s responder status.

[0351] The detection of Acidaminococcaceae as a family-level biomarker of non-response makes sense if we recall that one genus of this family - Phascolarctobacterium - was captured in the same context by Wu et al., while another genus of this family - Acidaminococcus - was reported to be associated with adverse effects of immune checkpoint inhibitor therapy, being enriched by nearly 200 times in the Adverse Events group (Hamada, 2023). The underlying functional nature of the similar direction of the influence of different genera of this family and the ability to generalize the biomarker to the family taxonomy cannot be easily tracked in the literature and requires a focused investigation. Despite the fact that selection of this familylevel marker by our procedure finds an instant echo in the above-mentioned relevant phenotypic associations involving the underlying genera, we prefer to avoid the temptation of over-analyzing the composition of our biomarker signature on the basis of the existing literature sentiment on health- or disease-associated role of individual taxonomy features. With the nature of the neural network model that highly relies on the complex non-linear relationships between the marker features abundance and clinical phenotypes, involvement of network-level relationships, other than “healthy” versus “disease-associated” traits ofindividual features is reasonably expected. This makes functional interpretation of the signature taxa a longer path that requires specific focused studies.

[0352] The practical advantage of the unveiled biomarker combination, despite its unclear functional background, comes from solving two outstanding problems: 1) overcoming biomarker sparseness and 2) enabling early detection of the therapy response signal, before the first dose of the drug. Indeed, the entire set of our 8 biomarkers was found in at least 12.3% of the patients, with individual marker detection showing the median prevalence of 67.8% across all patients. In comparison, the median prevalence of biomarkers from Wu et al.’s model was substantially lower, at 51%, with some of the markers (like Aggregatibacter and Parasutterella) dropping all the way to 1.94% of the patients, which is over 6 times less than the minimal prevalence observed for our model’s markers. This sparseness of the markers of the original model is expected to decrease the consistency of predictive power across different patient cohorts.

[0353] The robustness of our model was additionally confirmed by its performance across different stages of anti-PD-1 intervention: it showed a useful degree of predictive power even at early stages (Figure 2A), such as the baseline (ALIROC: 0.681 ± 0.056) and before the third dose of antiPD-1 (ALIROC: 0.688 ± 0.058). This qualitatively differs from the original model that performed randomly with an ALIROC of around 0.5 at any time before the third dose. The ability to stratify patients at these early stages, especially before the very first dose of the drug, is invaluable for clinicians to make timely and informed decisions about the course of treatment.

[0354] This result suggests that our model is able to detect the response signal that is not the result of drug induced microbiome perturbations but rather belongs to the persistent individual microbiome signature. To provide additional evidence in support of detecting individualized response signal in microbiomes by our predictor, independently of the drug-induced microbiome-restructuring effects (captured by other studies, including Wu et al.), we provide per-sample visualization of the prediction model’s calls (probabilities of therapy response), with coloring split at the probability 0.5 (Figure 2B), i.e. the lower values (shades of teal) suggested the absence of therapy response, while higher values (shades of pink) mark samples giving the “responder” predictive signature of various confidence. Those per-sample prediction visualizations are stacked by subject on top of the ground truth response indication to allow analysis of per-subject dynamics of the response prediction signals.

[0355] First, it is easily noticeable that in the absolute majority of the cases, the sign of the response prediction (color of the stacked squares), despite changes in the confidence level, stayed the same, with one prominent exception of subject #6 of the test set. This suggest that our modelcaptures the properties of the individual microbiomes per se, as opposed to respond to the drug-induced microbiota perturbations as seen by Wu et al.. At the same time, our model still captures the latter as a part of the overall signal. This is visible in the evolution of the confidence of the model calls following the increasing number of the drug doses, as seen in patients 24 and 31 (train set), as well as 6 and 13 (test set). This phenomenon contributes to overall increase in the classification performance after the 3rd dose in our case as well (Figure 2A), resulting in correct prediction of the responder status for the majority of the patients based on their last fecal sample (87.5% and 72.7% on train and test sets, respectively).

[0356] The above-mentioned case of subject #6 shows a rare instance when the initial microbiome composition of a non-responder subject was of the “responder” type as assessed by the model and this phenomenon even exacerbated after the first dose; however, subsequent doses resulted in the evolution of subsequent treatment- induced microbiome composition to converge to the correct - responder - signature. While it is early to speculate about the possibility of changing the responder status during the drug treatment course, given the limited data and restriction of its nature to the predictive model’s calls (that have their unavoidable degrees of uncertainty), such phenomenon, suggested by a remarkable switch of the prediction direction for the subject #6, may be theoretically considered and is worth investigating by plugging in clinical and metabolomic data in an integrative microbiota-host analysis. The latter may shed light on the nature of the possible dynamic therapy response alterations in the light of, for example, three distinct known HCC metabolic subtypes (Bidkhori, 2018) that may interact with gut microbiota in specific ways and result in either stable or evolving response to the drug intervention.

[0357] Despite the exception above, we should underline again the impressive overall stability of the R / NR binary calls by the model for individual subjects, capturing the individualized microbiome signatures as the main part of the detected signal. To add an intuitive numerical insight to this statement, please note that the subjects who received the longest treatment followed up by microbiome samples (14 doses in the training set’s subject #14 or 12 doses the test set’s subject #25) had, as the absolute majority of the subjects represented by multiple samples, single-direction response prediction in all the samples that belong to a subject, despite having a potential of 214=16,384 or 212=4,096 combinations of the binary classes if the classification was performed randomly.

[0358] We have shown that the issue of large inter-subject variation in the gut microbiome composition that usually limits the possibilities of creating a performant drug response predictor (that should balance informativeness and prevalence of individual features) can be overcome by selecting a representative mix of features across taxonomy ranks and processingthem as inputs to a custom-optimized neural network. The resulting model is advantageous compared to other models reported in the area because of a combination of 1) its capability to stratify the subjects before the drug intervention, relying largely on the innate composition of the individual microbiome rather than drug-induced microbiome perturbations and 2) doing so in a parsimonious manner, without requiring a long list of biomarkers or expensive strain-level data resolution. This combination of traits opens the doors to implementing the results of our original data processing decisions and the resulting model as a simple and affordable clinical test for subject stratification before planning the application of the ICI-baes treatment decision. EXAMPLE 3 - LUT Implementation and Standalone Deployment

[0359] To translate the ML predictive engine for ICI HCC treatment response, we generated a lookup table (LUT; Table 3) optimized to preserve the informativeness and predictive power of the original deep learning model while minimizing its size. For this purpose, we identified informative regions in the normalized abundance values for each of the eight features and created a compact (2,916 rows) lookup table that, after leveraging N-linear interpolation procedure (Salomon, 2011) can be used to match the predictive performance of the original deep learning model while being used within a standalone end-to-end prediction pipeline from matrix of raw counts to the therapy response probabilities.

[0360] The 2,916 key combinations of the input data come from the frugal adaptive quantization of the custom normalized feature abundance values (Table 1). To ensure reliable performance of the LUT predictor Table 3 in unseen datasets, the custom data normalization procedure should be followed as described herein.

[0361] Table 1. Frugal quantization of the frozen-minmax (see the description) normalized abundance values of the model’s features preserving the feature-level prediction informativeness.

[0362]

[0363]

[0364]

[0365] The normalization applies a Centered Log-Ratio (CLR) transform followed by min-max scaling to [0, 1], The normalization exactly reproduces the training-time pipeline, and the frozen CLR boundaries make the procedure transferable to an arbitrary dataset:

[0366] 1. CLR transform - computed over ALL taxa in the sample (not just the 8 model features):

[0367] CLR i = log ( count i + 0.5 ) - mean j ( lo ( count j + 0.5 ) ) for j = 1. . D

[0368] where D is the total number of taxa columns in the input.

[0369] 2. Feature selection - the 8 model features are extracted from the full CLR matrix.

[0370] 3. Min-max scaling - each of the 8 CLR values is scaled to [0, 1] using frozen min / max constants from the discovery cohort (n = 155):

[0371] x normalised = clamp ( (CLR i - CLR min i ) / (CLR max i CLR min i ) , 0 , 1 )

[0372] The frozen CLR boundaries are shown in Table 2:

[0373] Table 2. CLR boundaries

[0374] Feature

[0375]

[0376] Acidaminococcaceae -1.6448495014 6.1040720667 Ligilactobacillus -1.8308532563 8.0387307207 Pasteurellaceae -1.8308532563 7.4394776329 Dorea -1.7690356083 5.9186030758 Barnesiella -1.8308532563 4.5233744979 Tannerellaceae -1.6912515095 7.0896759889 Ruminococcus -1.7690356083 7.0477522430 Atopobium -1.9479242151 2.7531659628

[0377] ROC curves of the original model (black) and LUT-based predictions are shown in Figure 3 while scatter plots of the LUT-based prediction values versus the values output by the original model are shown in Figure 4. An end-to-end LUT-based prediction pipeline test: from separate unnormalized tables of genus and family-level counts to each sample’s associated therapy response probability is shown in Figure 5. These data show that LUT-based prediction values closely reflect the output from original model and so can be used to accurately implement the prediction methods disclosed herein.In some embodiments, interpolation between lookup-table vertices is performed in unbounded (-«, +oo) logit space rather than directly in the bounded (0, 1) probability space, to improve interpolation accuracy in the inherently non-linear probability domain (Salton, 1959). Specifically, the stored outcome probability values are converted to logits (log-probabilities) prior to N-linear interpolation, and the interpolated result is mapped back to a valid probability distribution via the Softmax transformation (Bishop, 2006).Table 3 - Look-up Table

[0378]

[0379]

[0380]

[0381]

[0382]

[0383]

[0384]

[0385]

[0386]

[0387]

[0388]

[0389]

[0390]

[0391]

[0392]

[0393]

[0394]

[0395]

[0396]

[0397]

[0398]

[0399]

[0400]

[0401]

[0402]

[0403]

[0404]

[0405]

[0406]

[0407]

[0408]

[0409]

[0410]

[0411]

[0412]

[0413]

[0414]

[0415]

[0416]

[0417]

[0418]

[0419]

[0420]

[0421]

[0422]

[0423]

[0424]

[0425]

[0426]

[0427]

[0428]

[0429]

[0430]

[0431]

[0432]

[0433]

[0434]

[0435]

[0436]

[0437]

[0438]

[0439]

[0440]

[0441]

[0442]

[0443]

[0444]

[0445]

[0446]

[0447]

[0448]

[0449]

[0450]

[0451]

[0452]

[0453]

[0454]

[0455]

[0456]

[0457]

[0458]

[0459]

[0460]

[0461]

[0462]

[0463]

[0464]

[0465]

[0466]

[0467]

[0468]

[0469]

[0470]

[0471]

[0472]

[0473]

[0474]

[0475]

[0476] REFERENCES

[0477] 1. Haber, P. K., Castet, F., Torres-Martin, M., Andreu-Oller, C., Puigvehi, M., Miho, M., Radu, P., Dufour, J. F., Verslype, C., Zimpel, C., Marquardt, J. II., Galle, P. R., Vogel, A., Bathon, M., Meyer, T., Labgaa, I., Digklia, A., Roberts, L. R., Ali, M. a. M., . . . Llovet, J. M. (2023). Molecular Markers of Response to Anti-PD1 Therapy in Advanced Hepatocellular Carcinoma. Gastroenterology, 164(1), 72-88. e18. https: / / doi.Org / 10.1053 / j.gastro.2022.09.005

[0478] 2. Bai, R., Z. Lv, D. Xu and J. Cui (2020). "Predictive biomarkers for cancer immunotherapy with immune checkpoint inhibitors." Biomark Res 8: 34

[0479] 3. Hao, L., S. Li, J. Deng, N. Li, F. Yu, Z. Jiang, J. Zhang, X. Shi and X. Hu (2023). "The current status and future of PD-L1 in liver cancer." Front Immunol 14: 1323581.

[0480] 4. Tumeh, P. C., C. L. Harview, J. H. Yearley, I. P. Shintaku, E. J. Taylor, L. Robert, B.

[0481] Chmielowski, M. Spasic, G. Henry, V. Ciobanu, A. N. West, M. Carmona, C. Kivork, E. Seja, G. Cherry, A. J. Gutierrez, T. R. Grogan, C. Mateus, G. Tomasic, J. A. Glaspy, R. O. Emerson, H. Robins, R. H. Pierce, D. A. Elashoff, C. Robert and A. Ribas (2014). "PD-1 blockade induces responses by inhibiting adaptive immune resistance." Nature 515(7528): 568-571.

[0482] 5. Lee, K. A., A. M. Thomas, L. A. Bolte, J. R. Bjork, L. K. de Ruijter, F. Armanini, F.

[0483] Asnicar, A. Blanco-Miguez, R. Board, N. Calbet-Llopart, L. Derosa, N. Dhomen, K. Brooks, M. Harland, M. Harries, E. R. Leeming, P. Lorigan, P. Manghi, R. Marais, J. Newton-Bishop, L. Nezi, F. Pinto, M. Potrony, S. Puig, P. Serra-Bellver, H. M. Shaw, S. Tamburini, S. Valpione, A. Vijay, L. Waldron, L. Zitvogel, M. Zolfo, E. G. E. de Vries, P. Nathan, R. S. N. Fehrmann, V. Bataille, G. A. P. Hospers, T. D. Spector, R. K. Weersma and N. Segata (2022). "Cross-cohort gut microbiome associations with immune checkpoint inhibitor response in advanced melanoma." Nat Med 28(3): 535-544. 6. Wu, H., X. Zheng, T. Pan, X. Yang, X. Chen, B. Zhang, L. Peng and C. Xie (2022).

[0484] "Dynamic microbiome and metabolome analyses reveal the interaction between gut microbiota and anti-PD-1 based immunotherapy in hepatocellular carcinoma." Int J Cancer 151(8): 1321- 1334.

[0485] 7. Derosa, L., V. lebba, C. A. C. Silva, G. Piccinno, G. Wu, L. Lordello, B. Routy, N. Zhao, C. Thelemaque, R. Birebent, F. Marmorino, M. Fidelle, M. Messaoudene, A. M. Thomas, G. Zalcman, S. Friard, J. Mazieres, C. Audigier-Valette, D. M. Sibilot, F. Goldwasser, A. Scherpereel, H. Pegliasco, F. Ghiringhelli, N. Bouchard, C. Sow, I. Darik, S. Zoppi, P. Ly, A. Reni, R. Daillere, E. Deutsch, K. A. Lee, L. A. Bolte, J. R. Bjork, R. K. Weersma, F. Barlesi, L. Padilha, A. Finzel, M. L. Isaksen, B. Escudier, L.Albiges, D. Planchard, F. Andre, C. Cremolini, S. Martinez, B. Besse, L. Zhao, N. Segata, J. Wojcik, G. Kroemer and L. Zitvogel (2024). "Custom scoring based on ecological topology of gut microbiota associated with cancer immunotherapy outcome." Cell 187(13): 3373-3389 e3316.

[0486] Yang, Y, X. Song, G. Wang, Y. Xia, Z. Xiong and L. Ai (2024). "Understanding Ligilactobacillus salivarius from Probiotic Properties to Omics Technology: A Review." Foods 13(6)

[0487] Hamada, K., J. Isobe, K. Hattori, M. Hosonuma, Y. Baba, M. Murayama, Y. Narikawa, H. Toyoda, E. Funayama, K. Tajima, M. Shida, Y. Hirasawa, T. Tsurui, H. Ariizumi, T. Ishiguro, R. Suzuki, R. Ohkuma, Y. Kubota, T. Sambe, M. Tsuji, S. Wada, Y. Kiuchi, S. Kobayashi, A. Kuramasu, A. Horiike, Y. G. Kim, T. Tsunoda and K. Yoshimura (2023). "Turicibacter and Acidaminococcus predict immune-related adverse events and efficacy of immune checkpoint inhibitor." Front Immunol 14: 1164724

[0488] Bidkhori, G., R. Benfeitas, M. Klevstig, C. Zhang, J. Nielsen, M. Uhlen, J. Boren and A. Mardinoglu (2018). "Metabolic network-based stratification of hepatocellular carcinoma reveals three distinct tumor subtypes." Proc Natl Acad Sci U S A 115(50): E11874-E11883

[0489] Salomon (emeritus), D. (2011). Linear Interpolation. In: The Computer Graphics Manual. Texts in Computer Science. Springer, London, https: / / doi.org / 10.1007 / 978-0-85729-886-7_9

[0490] Salton, G., 1959. The use of the central limit theorem for interpolating in tables of probability distribution functions. Mathematical Tables and Other Aids to Computation, 13, 213-216

[0491] Bishop, C.M. (2006) Pattern Recognition and Machine Learning. Springer, Berlin. https: / / link.springer.com / book / 9780387310732

[0492] Fluss, R., Faraggi, D., & Reiser, B. (2005). Estimation of the Youden Index and its associated cutoff point. Biometrical Journal, 47(4), 458-472.

[0493] doi: 10.1002 / bimj.200410135

Claims

CLAIMS1. A method of predicting the responsiveness to an Immune Checkpoint Inhibitor (ICQbased treatment in a cancer patient at baseline, said method comprising:assessing the relative abundance of at least 2 microbial taxa in a stool sample derived from the patient at baseline, wherein the at least 2 taxa are selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus; andstratifying the patient as a responder to the ICI-based treatment based on a probability score being higher or lower than a pre-determined threshold, wherein the probability score is determined using a look-up table generated from a non-linear model to match the relative abundance of the at least 2 microbial taxa to a probability score, wherein a cancer patient with a probability score above the pre-determined threshold is stratified as a responder, and a cancer patient with a probability score below the pre-determined threshold is stratified as a non-responder.

2. A method of claim 1 , wherein the look-up table is Table 3.

3. A method of claim 1 or 2, wherein the pre-determined threshold is a value selected within the range 0.4-0.6, preferably wherein the pre-determined threshold is 0.5.

4. A method of any one of claims 1 to 3, wherein assessing the relative abundance of the at least 2 microbial taxa comprises normalising and / or quantizing microbial count data, wherein normalising comprises:(i) receiving a plurality of microbial count data (features) for at least 2 microbial taxa associated with a stool sample,(ii) applying a centred log-ratio (CLR) transformation to the plurality of microbial count features,(iii) optionally selecting a subset of the transformed features, and(iv) scaling each feature in the subset to a bounded numerical range using feature-specific scaling parameters predetermined from a reference dataset, preferably wherein the bounded numerical range is [0,1], and / or preferably wherein feature-specific minimum and maximum values correspond to the CLR boundaries defined in Table 2;and wherein quantizing comprises:(v) performing a frugal quantization of the normalised abundance values, preferably using the key abundance values defined in Table 1.

5. A method of any one of the preceding claims, wherein determining the probability score using the look-up table is obtained by exact lookup, or, optionally, by N-linear interpolation.

6. A method of any one of claims 1 to 5, wherein the patient has hepatocellular carcinoma (HCC).

7. A method of anyone of claims 1 to 6, wherein the ICI-based treatment comprises at least one ICI.

8. A method of claim 7, wherein the at least one ICI is an anti-PD1 agent.

9. A method of claim 8, wherein the anti-PD1 agent is nivolumab or pembrolizumab.

10. A method of claim 7, wherein the at least one ICI is an anti-PDL1 agent.

11. A method of claim 7, wherein the at least one ICI is an anti-CTLA agent.

12. A method of claim 7, wherein the ICI-based treatment comprises a combination of 2 I Cl s such as an anti-PD1 agent and an anti-CTLA-4 agent.

13. A method of claim 7, wherein the ICI-based treatment is further combined with Molecular Targeted Therapy (MTT).

14. A method of claim 13, wherein the MTT comprises one or more therapies selected from a group consisting of: sorafenib, lenvatinib, apatinib and anlotinib.

15. A method of claim 7, wherein the ICI-based treatment is further combined with a Transarterial Chemoembolisation (TACE) therapy.

16. A method of claim 9, wherein the ICI-based treatment is used in combination with MTT and TACE therapy.

17. A method of claim 6, further comprising the step of treating the patients stratified as responders to the ICI-based treatment with an effective ICI-based treatment.

18. A method of claim 17, wherein the effective ICI-based treatment comprises an anti-PD1 agent.

19. A method of claim 18, wherein the anti-PD1 agent is nivolumab or pembrolizumab.

20. A method of claim 6, wherein stratifying the patient at baseline as a responder to the ICI-based treatment is based on lower relative abundance of Acidaminococcaceae family and lower relative abundance of Ligilactobacillus genus in the stool sample and stratifyingthe patient as a non-responder to the ICI-based treatment is based on higher relative abundance of Acidaminococcaceae family and higher relative abundance of Ligilactobacillus genus in the stool sample, wherein determining the relative abundance as high or low is made by the non-linear model in a context-dependent manner.

21. A method of treatment comprising, administering to a cancer patient an effective ICI-based treatment, wherein the patient has been stratified at baseline as a responder to the ICI-based treatment based on a probability score being higher or lower than a predetermined threshold, wherein the probability score is determined using a look-up table generated from a non-linear model to match relative abundance values of at least 2 microbial taxa in a stool sample derived from the patient to a probability score, the at least 2 taxa selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus;wherein a cancer patient with a probability score above the pre-determined threshold is stratified as a responder, and a cancer patient with a probability score below the predetermined threshold is stratified as a non-responder.

22. A method of claim 21, wherein the look-up table is Table 3.23 .A method of treatment of claim 21 or 22, wherein the pre-determined threshold is a value selected within the range 0.4-0.6, preferably wherein the pre-determined threshold is 0.5.

24. A method of any one of claims 21 to 23, wherein assessing the relative abundance of the at least 2 microbial taxa comprises normalising and / or quantizing microbial count data, wherein normalising comprises:(i) receiving a plurality of microbial count data (features) for at least 2 microbial taxa associated with a stool sample,(ii) applying a centred log-ratio (CLR) transformation to the plurality of microbial count features,(iii) optionally selecting a subset of the transformed features, and(iv) scaling each feature in the subset to a bounded numerical range using feature-specific scaling parameters predetermined from a reference dataset, preferably wherein the bounded numerical range is [0,1], and / or preferably wherein feature-specific minimum and maximum values correspond to the CLR boundaries defined in Table 2;115and wherein quantizing comprises:(v) performing a frugal quantization of the normalised abundance values, preferably using the key abundance values defined in Table 1.

25. A method of any one of the preceding claims, wherein determining the probability score using the look-up table is obtained by exact lookup, or, optionally, by N-linear interpolation.

26. A method of treatment according to any one of claims 21-25, wherein the patient has HCC.

27. A method according to any one of claim 21-26, wherein the ICI-based treatment comprises at least one ICI.

28. A method of, claim 27 wherein the at least 1 ICI is an anti-PD1 agent.29 . A method of claim 28, wherein the anti-PD1 agent is nivolumab or pembrolizumab.

30. A method of claim 27, wherein the at least one ICI is an anti-PDL1 agent.

31. A method of claim 27, wherein the at least one ICI is an anti-CTLA agent.

32. A method of claim 27, wherein the ICI-based treatment comprises a combination of 2 I Cl s such as an anti-PD1 agent and an anti-CTLA-4 agent.

33. A method of claim 27, wherein the ICI-based treatment further comprises Molecular Targeted Therapy (MTT).

34. A method of claim 33, wherein the MTT comprises one or more therapies selected from a group consisting of: sorafenib, lenvatinib, apatinib and anlotinib.

35. A method of claim 27, wherein the ICI-based treatment further comprises Transarterial Chemoembolisation (TACE) therapy.

36. A method of claim 27, wherein the ICI-based treatment further comprises a combination of MTT and TACE therapy.

37. The method of any one of the preceding claims, wherein the at least 2 microbial taxa comprise: (i) at least one family-level taxa selected from said Acidaminococcaceae family, Pasteurellaceae family, and Tannerellaceae family, and (ii) at least one genus-level taxa selected from said Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.11638. The method of claim 37, wherein the at least two family-level taxa comprise the three taxa Acidaminococcaceae family, Pasteurellaceae family and Tannerellaceae family, and wherein the at least two genus-level taxa comprise the five taxa Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

39. The method of any one of the preceding claims, wherein the at least two microbial taxa comprise exactly eight taxa.

40. The use of at least 1 ICI in the manufacture of a medicament for the therapeutic treatment of cancer, wherein the patient has been:stratified at baseline as a responder to the ICI-based treatment based on a probability score being higher or lower than a pre-determined threshold, wherein the probability score is determined by using a look-up table generated from a non-linear model to match relative abundance values of at least 2 microbial taxa in a stool sample derived from the patient to a probability score, the at least 2 taxa selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

41. A use of claim 40, wherein the look-up table is Table 3.42 .A use of claim 40 or 41 , wherein the pre-determined threshold is a value selected within the range 0.4-0.6, preferably wherein the pre-determined threshold is 0.5.

43. A use of any one of claims 40 to 42, wherein assessing the relative abundance of the at least 2 microbial taxa comprises normalising and / or quantizing microbial count data, wherein normalising comprises:(i) receiving a plurality of microbial count data (features) for at least 2 microbial taxa associated with a stool sample,(ii) applying a centred log-ratio (CLR) transformation to the plurality of microbial count features,(iii) optionally selecting a subset of the transformed features, and(iv) scaling each feature in the subset to a bounded numerical range using feature-specific scaling parameters predetermined from a reference dataset, preferably wherein the bounded numerical range is [0,1], and / or preferably wherein feature-specific minimum and maximum values correspond to the CLR boundaries defined in Table 2;and wherein quantizing comprises:117(v) performing a frugal quantization of the normalised abundance values, preferably using the key abundance values defined in Table 1.

44. A use according to any one of claims 40-43, wherein determining the probability score using the look-up table is obtained by exact lookup, or, optionally, by N-linear interpolation.

45. The use according to any one of claims 40-44, wherein the at least 1 ICI is an anti-PD1 agent and the patient has HCC.

46. The use according to any one of claims 40 -45 , wherein the at least 2 microbial taxa comprise: (i) at least one family-level taxa selected from said Acidaminococcaceae family, Pasteurellaceae family, and Tannerellaceae family, and (ii) at least one genus-level taxa selected from said Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

47. The use according to claim 46, wherein the at least two family-level taxa comprise the three taxa Acidaminococcaceae family, Pasteurellaceae family and Tannerellaceae family, and wherein the at least two genus-level taxa comprise the five taxa Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

48. An ICI for use in a method of treating cancer, the method comprising administering to a cancer patient the ICI, wherein the patient has been stratified at baseline as a responder to the ICI based on a probability score being higher or lower than a pre-determined threshold, wherein the probability score is determined by using a look-up table generated from a non-linear model to match relative abundance values of at least 2 microbial taxa in a stool sample derived from the patient to a probability score, the at least 2 taxa selected from a group consisting of: Acidaminococcaceae family, Pasteurellaceae family, Tannerellaceae family, Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.49 . The ICI for use according to claim 48, wherein the look-up table is Table 3. 50 . The ICI for use according to claim 48 or 49, wherein the pre-determined threshold is a value selected within the range 0.4-0.6, preferably wherein the predetermined threshold is 0.5.11851. The ICI for use according to any one of claims 48-50, wherein assessing the relative abundance of the at least 2 microbial taxa comprises normalising and / or quantizing microbial count data, wherein normalising comprises:(i) receiving a plurality of microbial count data (features) for at least 2 microbial taxa associated with a stool sample,(ii) applying a centred log-ratio (CLR) transformation to the plurality of microbial count features,(iii) optionally selecting a subset of the transformed features, and(iv) scaling each feature in the subset to a bounded numerical range using feature-specific scaling parameters predetermined from a reference dataset, preferably wherein the bounded numerical range is [0,1], and / or preferably wherein feature-specific minimum and maximum values correspond to the CLR boundaries defined in Table 2;and wherein quantizing comprises:(v) performing a frugal quantization of the normalised abundance values, preferably using the key abundance values defined in Table 1.

52. The ICI for use according to any one of claims 48 -51 , wherein determining the probability score using the look-up table is obtained by exact lookup, or, optionally, by bilinear interpolation.

53. The ICI for use according to any one of claims 48-52, wherein the patient has HCC.

54. The ICI for use according to any one of claims 48-53, wherein the ICI is an anti-PD1 agent.55 . The ICI for use according to claim 54, wherein the anti-PD1 agent is nivolumab or pembrolizumab.

56. The ICI for use according to any one of claims 48-53, wherein the ICI is an anti-PDL1 agent.

57. The ICI for use according to any one of claims 48-53, wherein the ICI is an anti-CTLA agent.

58. The ICI for use according to any one of claims 48-57, wherein the ICI is administered in combination with a second ICI agent, optionally wherein in the second ICI is an anti-PD1 agent or an anti-CTLA-4 agent.11959. The ICI for use according to any one of claims 48-58, wherein the treatment is used in combination with Molecular Targeted Therapy (MTT).

60. The ICI for use according to claim 59, wherein the MTT comprises one or more therapies selected from a group consisting of: sorafenib, lenvatinib, apatinib and anlotinib.

61. The ICI for use according to any one of claims 48-60, wherein the ICI is used in combination with Transarterial Chemoembolisation (TACE) therapy.

62. The ICI for use according to any one of claims 48-61, wherein the ICI is used in combination with MTT and TACE therapy.

63. The ICI for use according to any one of claims 48-62, wherein the at least 2 microbial taxa comprise: (i) at least one family-level taxa selected from said Acidaminococcaceae family, Pasteurellaceae family, and Tannerellaceae family, and (ii) at least one genus-level taxa selected from said Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

64. The ICI for use according to claim 63, wherein the at least two family-level taxa comprise the three taxa Acidaminococcaceae family, Pasteurellaceae family and Tannerellaceae family, and wherein the at least two genus-level taxa comprise the five taxa Ligilactobacillus genus, Dorea genus, Barnesiella genus, Ruminococcus genus and Atopobium genus.

65. The ICI for use according to any one of claims 48 to 64, wherein the at least two microbial taxa comprise exactly eight taxa.120

Images