Identification of clonal neoantigens and their uses

Abstract

Methods are provided for determining whether a tumor-specific mutation in a subject is likely to be clonal. The method includes providing sequence data from one or more samples from a subject comprising tumor genetic material, the sequence data including for each of the one or more samples, and determining a likelihood that the tumor-specific mutation is clonal as a posterior probability that depends on a prior probability that the mutation is clonal and a probability that the sequence data would be observed if the tumor-specific mutation were (i) clonal, and (ii) non-clonal, given the tumor fraction and one or more candidate binding genotypes for each of the one or more samples. Related methods, systems, and articles of manufacture are also described.

Description

[Technical field]

[0001] Field of Disclosure The present disclosure relates to methods for determining whether a tumor-specific mutation is likely to be clonal and to methods for identifying clonal neo-antigens derived from tumor-specific mutations present in a tumor. The present disclosure also relates to methods and compositions for the treatment of cancer that utilize or target the identified clonal neo-antigens. [Background technology]

[0002] background Cancer cells are known to acquire mutations, some of which may confer an evolutionary advantage. As a result, tumors often contain multiple related populations (i.e., clones) with distinct genotypes. Characterizing the clonal composition of tumors is particularly important in the context of treatment. Indeed, the clinical benefit associated with targeting mutations present in only a subset of the tumor cell population (also called "subclonal" mutations) is limited because the therapy targets only a portion of the population, and unaffected clones may still be capable of proliferation, making them more likely to relapse or metastasize. Instead, there is a growing belief that effective tumor management may require targeting clonal neo-antigens (antigens expressed as a result of the presence of mutations present in every tumor cell) or combining multiple targeted therapies (McGranahan et al., 2015). In addition, the burden of clonal neoantigens is known to be associated with prognosis in at least some cancers and with sensitivity to checkpoint inhibitor treatment (McGranahan et al., 2016; Litchfield et al., 2021). Summary of the Invention [Means for solving the problem]

[0003] overview The present inventors have developed a novel method that addresses one or more of the problems of prior art approaches for determining whether tumor-specific mutations are likely to be clonal using sequence data from one or more samples containing tumor cells or genetic material derived therefrom. The method finds particular application in identifying clonal neoantigens, for example for cancer therapy or prognosis purposes. The method uses a rigorous statistical framework to classify individual mutations as clonal and provides probabilities reflecting the confidence of the assignment. The method is fast, flexible, robust, reproducible, and relies on interpretable assumptions, and can flexibly incorporate somatic copy number abnormality data and take into account the uncertainty of copy number calls due to multiple ploidy / purity solutions when providing its predictions.

[0004] Thus, according to a first aspect, there is provided a method for determining whether a tumor-specific mutation in a subject is likely to be clonal, comprising providing or obtaining sequence data from one or more samples from the subject comprising tumor genetic material, the sequence data comprising, for each of the one or more samples, a number of reads indicative of the tumor-specific mutation in the sample (d b and determining a likelihood that the tumor-specific mutation is clonal as a posterior probability that depends on the prior probability that the mutation is clonal and the probability that the sequencing data is observed when the tumor-specific mutation is (i) clonal, and (ii) non-clonal, given the tumor fraction for each of the one or more samples, and one or more candidate binding genotypes, each of which contains a genotype at the position of the tumor-specific mutation, for a normal population, a reference tumor population that does not contain the tumor-specific mutation, and a mutant tumor cell population that contains the tumor-specific mutation.

[0005] In this method, the probability that a mutation is clonal (P(Z=1)) is calculated as a posterior probability (p(Z=1|d b ,d,π,t,ρ). It can therefore be determined separately for any mutation simply using the sequence data pertaining to that mutation under investigation, producing an easily interpretable output that depends on explicit assumptions (i.e., the prior probability that the mutation is clonal, and the likelihood that the sequence data will be observed given an explicit model of population structure). In other words, such probability depends on both the data available from one or more samples containing tumor genetic material, and any prior knowledge available, through a Bayesian framework underscored by a rigorous set of assumptions that can be tested. This output can be used to compare, e.g., prioritize, any number of mutations, without it having to encompass the entire genome. It also has the ability to combine evidence from multiple samples containing tumor genetic material, but can equally well be determined using a single sample containing genetic material.

[0006] The method of this aspect may have one or more of the following features.

[0007] The method can be computer-implemented. Thus, the step of obtaining sequence data can be performed by a processor, and the step of determining the likelihood that a tumor-specific mutation is clonal can be performed by said processor. The step of obtaining sequence data can include receiving sequence data including sequence reads from one or more samples from a subject, and determining from the sequence reads a number of reads indicative of a tumor-specific mutation in a sample (d b), the number of reads in the sample that represent the corresponding germline allele, and the total number of reads at the location of the tumor-specific mutation (d). The step of determining at least the likelihood that the tumor-specific mutation is clonal may be computer-implemented. The step of determining the likelihood that the tumor-specific mutation is clonal may include a numerical integration step to determine the posterior probability. In particular, this step may include determining the posterior probability that the mutation is clonal by solving multiple one-dimensional integrals (e.g., a pair of integrals for each sample that represent the assumptions that the mutation is clonal and non-clonal, respectively) that integrate the probability that the sequence data is observed for all possible cancer cell fractions between 0 and 1, given the prior probability that the mutation is clonal and the probability that the sequence data is observed when the tumor-specific mutation is (i) clonal and (ii) non-clonal. These numerical integrals may be solved independently (e.g., in parallel) for each sample and for each mutation. The providing step may include one or more steps, all or some of which are computer-implemented.

[0008] The probability that a tumor-specific mutation is clonal may depend on the prior probability that the mutation is clonal (ρ) by the prior probability that the mutation is assigned to the clonal category given the prior probability that the mutation is clonal (P(Z=1|ρ)=ρ); and the prior probability that the mutation is assigned to the non-clonal category given the prior probability that the mutation is clonal (P(Z=0|ρ)=(1-ρ). The probability of observing sequence data when the tumor-specific mutation is clonal (given the tumor fraction and one or more candidate binding genotypes for each of the one or more samples) may be marginalized with respect to the cancer cell fraction. Similarly, the probability of observing sequence data when the tumor-specific mutation is not clonal, given the tumor fraction and one or more candidate binding genotypes for each of the one or more samples, may be marginalized with respect to the cancer cell fraction.

[0009] The probability that a tumor-specific mutation is clonal may depend on the prior probability that the mutation is assigned to the clonal category (P(Z=1|ρ)=ρ), given the prior probability that the mutation is clonal, multiplied by the probability that the sequence data of each sample would be observed if the mutation were clonal, given the tumor fraction and one or more candidate binding genotypes (this can be calculated as the likelihood of the sequence data of each sample, marginalized with respect to the cancer cell fraction, ψ1); and the prior probability that the mutation is assigned to the nonclonal category (P(Z=0|ρ)=(1-ρ), given the prior probability that the mutation is subclonal, multiplied by the probability that the sequence data of each sample would be observed if the mutation were nonclonal, given the tumor fraction and one or more candidate binding genotypes (this can be calculated as the likelihood of the sequence data of each sample, marginalized with respect to the cancer cell fraction, ψ0).

[0010] The probability that a tumor-specific mutation is clonal is calculated by (i) the prior probability that the mutation is assigned to the clonal category given the prior probability that the mutation is clonal, multiplied by the probability of observing sequence data for each sample given the tumor fraction and one or more candidate binding genotypes if the mutation is clonal (p(d b ,d,Z=1|π,t,ρ), which can be expressed as ρψ1), (ii)(i) (i.e., p(d b ,d,Z=1|π,t,ρ)) and the prior probability of assigning a mutation to the nonclonal category given the prior probability that the mutation is subclonal, multiplied by the probability of observing sequence data in each sample given the tumor fraction in each sample and one or more candidate binding genotypes if the mutation is nonclonal (p(d b ,d,Z=0|π,t,ρ, which can be expressed as (1-ρ)ψ0).

[0011] The probability that a mutation is clonal can be determined using equation (11a), where the term Pr(d b ,d|π,φ,t) may be given by any of equation (3), (4), (3a), (4a), (3b) or (4b). In equation (11), the terms p(φ|Z=0) and p(φ|Z=1) may be given by equation (6).

[0012] A clonal mutation may be a mutation that is present in all or essentially all tumor cells in one or more samples from a subject that contain tumor genetic material (or in all of the tumor genetic material in one or more samples). Such a mutation may be present, or may be assumed to be present, in all tumor cells of a subject (as being completely certain in this regard may involve sequencing of all tumor cells of a subject, but presence in essentially all cells in one or more samples may be used as an indication of this).

[0013] The probability of observing sequence data when considering the tumor fraction and one or more candidate binding genotypes for each sample can depend on the probability of observing sequence data when considering the tumor fraction, the cancer cell fraction, and one or more candidate binding genotypes (Pr(d,db|π,φ,t)). The probability of observing sequence data when considering the tumor fraction, the cancer cell fraction, and one or more candidate binding genotypes can be a weighted sum of the probabilities of observing sequence data when considering each of the tumor fraction, the cancer cell fraction, and one or more candidate binding genotypes.

[0014] Advantageously, the probability of observing sequence data (likelihood of sequence data) may be calculated for multiple candidate genotypes (e.g., calculated as a sum of probabilities including a term for each candidate genotype, see e.g., formula (3a), formula (3b)), and their contributions may be weighted, for example, to reflect prior knowledge of the relative probabilities of the candidate genotypes (e.g., any prior knowledge of whether some genotypes are more likely to appear than others). When such prior knowledge is not available or is undesirable, the probabilities of each of the candidate genotypes may be weighted equally. The weights of each candidate genotype are preferably considered to sum to 1, so that the total probability reflects the relative contribution of the different candidate binding genotypes considered. When a single candidate binding genotype is used, it may be assigned a weight of 1 (i.e., no summation may be performed).

[0015] Tumor fraction, cancer cell fraction, and specific candidate binding genotype (G i ) is the probability of observing sequence data when the probability of the sequence data being observed is considered (this is the likelihood of the sequence data for each sample, marginalized with respect to the proportion of cancer cells, ψ z (can be calculated as follows) is the parameter d b and ξ(G i Alternatively, the probability of observing sequence data given the tumor fraction, cancer cell fraction, and a particular candidate binding genotype can be calculated using a binomial distribution with the parameter d b , ξ(G i ,φ,t), and γ can be obtained using the beta binomial distribution. In both cases (i.e., whether the binomial or beta binomial distribution is used), ξ(G i ,φ,t) may represent the probability of extracting a read containing a mutant allele given a particular genotype Gi, cancer cell fraction φ, and tumor purity t. i,φ,t) can be calculated as a function of the total copy number for each of the normal, mutant and reference genotypes, the probability of extracting a read containing a variant from a population of genotype Gi given the proportion of alleles that are mutant in terms of genotype and sequencing error rate at that locus, the proportion of tumor in the sample and the proportion of cancer cells with that mutation.

[0016] The probability of observing the sequence data given the tumor fraction and one or more candidate binding genotypes for each sample can be determined as an integral over all possible values ​​of the cancer cell fraction for each sample, where the cancer cell fraction is the proportion of tumor cells that contain the tumor-specific mutation. Thus, determining the likelihood that the tumor-specific mutation is clonal can include numerically integrating the integral using a processor.

[0017] The cancer cell fraction (φ) can take values ​​between 0 and 1. In other words, the probability of observing the sequence data when the mutation is clonal or nonclonal, given the tumor fraction of each sample and one or more candidate binding genotypes, can be determined by integrating a value that depends on the cancer cell fraction over all possible values ​​of the cancer cell fraction (i.e., marginalizing with respect to the cancer cell fraction). The value that depends on the cancer cell fraction is Pr(d b ,d|π,φ,t)p(φ|Z=z), where the first term is the probability of observing sequence data given the tumor fraction, cancer cell fraction, and one or more candidate binding genotypes, and the second term is the prior probability of the cancer cell fraction when the mutation is classified as clonal or nonclonal (Z=1 or Z=0, respectively) (i.e., the probability based on assumptions of how the cancer cell fraction should behave for clonal / nonclonal mutations). Thus, the probability of observing sequence data given the tumor fraction and one or more candidate binding genotypes for each sample is:

number

[0018] The prior probability of a particular cancer cell fraction when a mutation is classified as clonal can be defined as a beta distribution with parameter α (set to a value greater than 1, e.g., 99, although any other value may be used) and β=1 (beta(φ|α,1)). The prior probability of a particular cancer cell fraction when a mutation is classified as non-clonal can be defined as a beta distribution with parameter α=1 and β=1 (beta(φ|1,1)).

[0019] Sequencing data may be obtained from multiple samples, and the probability of observing the sequencing data for each of the multiple samples given the tumor fraction and one or more candidate binding genotypes may be determined as the product of the probabilities of observing the sequencing data for each sample given the tumor fraction and one or more candidate binding genotypes for each sample.

[0020] Advantageously, the method is able to seamlessly integrate evidence supporting / refuting the clonality of mutations obtained from multiple samples when they are available. This is particularly advantageous since methods for inferring the clonal structure of tumor cell populations that can utilize multi-region sequencing data have been shown to perform particularly well in benchmark studies (Farahani et al., 2017).

[0021] The method may further comprise determining or providing at least one estimate of tumor fraction and at least one corresponding set of one or more candidate binding genotypes for each sample. The tumor fraction estimate may be determined using a method for determining an allele-specific copy number profile in a sample comprising a mixture of tumor cells and normal cells. Methods for doing this using sequencing or array data are known in the art, for example, by expressing the allele-specific data as a function of parameters including allele-specific copy number, tumor aneuploidy and tumor cell fraction, and identifying the values ​​of those parameters that best fit all the data. Examples of such methods include, among others, ASCAT (Van Loo et al., 2010). Alternatively, the tumor fraction estimate may be determined experimentally. Thus, the method may further comprise determining a tumor fraction estimate for each of one or more samples. In particular, the method may include determining, by a processor, for each sample, at least one estimate of tumor fraction, the processor using the sequencing data to determine estimates of tumor fraction and allele-specific copy number, and determining, by the processor, a set of one or more candidate binding genotypes associated with the allele-specific copy number.

[0022] The set of one or more candidate genotypes can be determined using allele-specific copy numbers for tumor cells in the mixed sample or variables derived therefrom (or conversely, such allele-specific copy numbers may be derived therefrom, such as B allele fraction and log R). The allele-specific copy numbers for tumor cells in the mixed sample can be determined using a method for determining an allele-specific copy number profile in a sample comprising a mixture of tumor and normal cells, such as, for example, ASCAT (Van Loo et al., 2010), or ascatNgs (Raine et al., 2016), among others.

[0023] Thus, the method may further include determining, for each of the one or more samples, estimates for at least two of the copy number of the major allele in the tumor cells in the sample, the copy number of the minor allele in the tumor cells in the sample, and the total copy number at the position of the tumor-specific mutation in the tumor cells in the sample. The estimates of the copy number in the tumor cells in the sample may represent a summarized (e.g., average) estimate for the entire population of tumor cells in the sample.

[0024] The set of one or more candidate combination genotypes is determined to be a set of genotypes in which only the normal population contains one or more normal alleles A (i.e., G H = AA or A, e.g., if the locus is on a sex chromosome; the reference population does not contain the mutant allele B (i.e., G R =(A) * n); and the mutant population contains at least one copy of the mutant allele B (i.e., G V =(A) * m(B) * l) can be obtained as a candidate binding genotype that meets the assumption.

[0025] Advantageously, the set of candidate genotypes may include candidate binding genotypes that further fit either of the assumptions: (i) the reference population genotype matches the normal population genotype, and the mutant population has a copy number equal to the total copy number at that position, up to the major copy number of the mutant allele; or (ii) the reference population has a copy number equal to the total copy number at that position, and the mutant population has one mutant allele and a copy number equal to the total copy number at that position ("major copy number prior"). This approach advantageously strikes a good balance between taking into account the uncertainty of the population's genotypes, while avoiding taking into account too many conditions.

[0026] Alternatively, or in addition, the set of one or more candidate combination genotypes may be a set of genotypes in which each mutation is diploid and heterozygous (i.e., G V =A-B, G RAlternatively, or in addition, the set of one or more candidate combination genotypes may include any of the candidate combination genotypes that meet the assumption that each mutation is diploid and homozygous (i.e., G V =BB,G R Alternatively, or in addition, the set of one or more candidate binding genotypes may include any of the candidate binding genotypes that meet the assumption that the genotypes in the mutant population have the expected total copy number in the region of the mutation, with exactly one being the mutant allele (i.e., G V =(A) * Alternatively, or in addition, the set of one or more candidate binding genotypes may include any of the candidate binding genotypes that meet the assumption that the mutant population has the expected total copy number in the region of the mutation, with at least one being the mutant allele, and the reference population is either AA or a genotype with a copy number equal to the expected total copy number and no mutant allele (i.e., G R =(A) * n, where n is the total copy number, G V =(A) * m(B) * The set of one or more candidate binding genotypes may include any of the candidate binding genotypes that meet the assumption that the genotypes in the mutant population have a number of mutant alleles that corresponds to either the major or minor copy number (the "parental mode").

[0027] The probability of observing the sequence data may be a combination of multiple probabilities of observing the sequence data given the respective tumor fractions and corresponding sets of one or more candidate binding genotypes for at least one of the one or more samples, and optionally the method includes determining multiple estimates of tumor fractions and corresponding sets of one or more candidate binding genotypes for at least one of the one or more samples. Thus, the method may include determining multiple estimates of tumor fractions for at least one sample. This may include determining, by a processor, multiple estimates of tumor fractions and corresponding multiple allele-specific copy numbers that match the sequence data, and determining, by a processor, multiple sets of one or more candidate binding genotypes associated with the multiple allele-specific copy numbers.

[0028] The method is advantageously capable of determining the probability that a mutation is clonal, taking into account multiple possible tumor fractions and corresponding candidate binding genotype sets. In other words, the method is capable of determining the probability that a mutation is clonal, integrating over multiple copy number solutions from which the tumor fraction and candidate binding genotype can be determined. In contrast, prior art approaches typically rely on a single estimate of tumor purity and allele-specific copy number (from which the candidate binding genotype can be determined), which is often manually selected based on an expert-defined optimality criterion. The step of selecting the copy number solution considered optimal is a highly error-prone step, and the output of a method that relies on a single solution can vary significantly depending on that solution.

[0029] Thus, advantageously, the probability of observing sequence data (likelihood of sequence data) may be calculated for multiple sets of candidate genotypes and corresponding tumor fraction estimates (e.g., as a sum of probabilities including a term for each copy number solution, see formula (3b), formula (4b)), and the contributions may be weighted to reflect, for example, the confidence in the copy number solutions from which the tumor fraction estimates and candidate genotype sets were derived. The weights of the copy number solution contributions are preferably considered to sum to 1, resulting in a total probability reflecting the relative contributions of the different copy number solutions considered. When a single copy number solution is used, it can be assigned a weight of 1 (i.e., no summation may be performed).

[0030] The prior probability that a mutation is clonal may be set to a neutral prior value or a value derived from prior data and / or expert knowledge. Advantageously, the method described herein is able to take into account prior knowledge about the mutation in the Bayesian framework used to determine the probability that the mutation is clonal, if such knowledge is available. However, the method is also able to provide a likelihood that a mutation is clonal even in the absence of such knowledge.

[0031] The value of the prior probability that a mutation is clonal may depend on the subject, the tumor, the mutation, or a combination thereof. For example, the value may be determined using previously collected data on a relevant patient cohort, such as patients suffering from the same type or subtype of cancer. For example, knowledge of the proportion of clonal / subclonal mutations in such a cohort may be used to set the prior probability used herein. Alternatively, the value may be set arbitrarily based on prior knowledge of the cancer type or mutation. For example, a particular mutation that is found in multiple cancer samples and is often identified as clonal in those samples may be assigned a probability higher than 0.5.

[0032] The method may further include repeating the method for a plurality of tumor-specific mutations identified in the subject. The method may further include ranking or otherwise prioritizing the plurality of tumor-specific mutations based at least in part on their determined likelihood to be clonal in the subject.

[0033] The method may further include identifying one or more tumor-specific mutations of the subject. Identifying one or more tumor-specific mutations of the subject may be performed, such as by using sequence data from one or more samples from the subject that contain tumor genetic material and sequence data from one or more germline samples from the subject, and comparing the sequence data. Identifying one or more tumor-specific mutations of the subject may include aligning sequence data from at least one sample that contains tumor genetic material with a reference sequence, and identifying positions where the sequence of the sample differs from the reference sequence. The method may further include aligning sequence data from at least one germline sample with a reference sequence, and identifying positions where the sequence of the sample that contains tumor genetic material differs from the germline sample.

[0034] The step of providing sequence data from one or more samples from a subject may include or consist of receiving sequence data from a user (e.g., via a user interface), from one or more computing devices, or from one or more data stores or databases.

[0035] The step of providing sequence data may further include sequencing one or more samples from the subject that contain the tumor genetic material (or otherwise determining the sequence composition of the genomic material present in the samples).

[0036] The method may further include sequencing one or more germline samples from the subject (or otherwise determining the sequence composition of the genomic material present in the samples).

[0037] The method may further include obtaining one or more samples containing tumor genetic material, and optionally one or more germline samples, from the subject.

[0038] The method may further include providing the probability that the determined tumor-specific mutation is clonal and / or a value derived therefrom or related thereto to a user, e.g., via a user interface. For example, the method may include providing a "clonal status" flag or value based on the probability that the determined tumor-specific mutation is clonal. As another example, the method may include providing information identifying the mutation, such as the sequence of the mutation and its genomic location.

[0039] According to a further aspect, there is provided a method of identifying one or more clonal neo-antigens in a subject, comprising: identifying a plurality of tumor-specific mutations in the subject; determining whether one or more of the tumor-specific mutations are likely to be clonal in the subject using a method of any embodiment of the previous aspect; and determining whether one or more of the tumor-specific mutations are likely to give rise to a neo-antigen, wherein a clonal neo-antigen is a tumor-specific mutation that satisfies one or more predefined criteria as to whether the tumor-specific mutation is likely to be clonal and one or more criteria as to whether the tumor-specific mutation is likely to give rise to a neo-antigen. Also described according to this aspect is a method for identifying one or more clonal neo-antigens in a subject, the method comprising: identifying, by a processor, a plurality of tumor-specific mutations in the subject using sequence data from one or more samples from the subject; determining, by the processor, whether one or more of the tumor-specific mutations are likely to be clonal in the subject using the method of any of the preceding claims; and selecting, by the processor, one or more of the tumor-specific mutations as clonal neo-antigen candidates, wherein the clonal neo-antigen candidates are tumor-specific mutations that satisfy at least one or more predetermined criteria as to whether the tumor-specific mutation is likely to be clonal and optionally one or more criteria as to whether the tumor-specific mutation is likely to give rise to a neo-antigen.

[0040] The method of this aspect may have any one or more of the following features.

[0041] A clonal neo-antigen may be a tumor-specific mutation that meets at least a criterion selected from: having a probability of being clonal above a predefined threshold, having a probability of being clonal above a threshold adaptively set to select a predefined number of tumor-specific mutations that have the highest probability of being clonal among the tumor-specific mutations for which the probability was determined, and having a probability of being clonal above a threshold adaptively set to select a predefined top percentile of tumor-specific mutations among the tumor-specific mutations for which the probability was determined. Thus, the one or more predefined criteria for whether a tumor-specific mutation may be clonal may be selected from: the mutation having a likelihood of being clonal above a predefined threshold, having a likelihood of being clonal above a threshold adaptively set to select a predefined number of tumor-specific mutations that have the highest likelihood of being clonal among the tumor-specific mutations for which the likelihood was determined, and having a likelihood of being clonal above a threshold adaptively set to select a predefined top percentile of tumor-specific mutations among the tumor-specific mutations for which the likelihood was determined.

[0042] A clonal neo-antigen may be a tumor-specific mutation that meets at least the following criteria: associated with an expression product expressed in tumor cells, predicted to give rise to a protein or peptide not expressed in normal cells of the subject, predicted to give rise to at least one peptide that may be presented by an MHC molecule, predicted to give rise to at least one peptide that may be presented by an MHC allele known to be present in the subject, and predicted to give rise to an immunogenic protein or peptide. For example, a clonal neo-antigen may be a tumor-specific mutation that meets the criteria of predicted to give rise to a change in the sequence of a protein (e.g., because it is a coding sequence, because it affects a splice site, because it gives rise to a truncated peptide, etc.), and thus to a protein or peptide that may not be expressed in normal cells of the subject. Whether this is the case can be further confirmed, for example, by comparison with the predicted normal proteome of the subject. Thus, one or more criteria for whether a tumor-specific mutation is likely to give rise to a neoantigen can be selected from: the mutation is associated with an expression product expressed in tumor cells; the mutation is predicted to result in a protein or peptide that is not expressed in normal cells of the subject; the mutation is predicted to result in at least one peptide that can be presented by an MHC molecule; the mutation is predicted to result in at least one peptide that can be presented by an MHC allele known to be present in the subject; and the mutation is predicted to result in an immunogenic protein or peptide.

[0043] The method may further include identifying one or more peptides associated with one or more clonal neo-antigens (i.e., one or more peptide sequences predicted to be present in tumor cells as a result of the presence of a tumor-specific mutation), where the tumor-specific mutation meets one or more criteria (related to the likelihood of clonality and the likelihood of giving rise to a clonal neo-antigen) as described above.

[0044] As one skilled in the art will appreciate, the complexity of the operations described herein (due at least to the complexity of determining the posterior probabilities that require numerical integration as described herein, and the amount of data typically generated by sequencing genomic DNA) is beyond the scope of mental activity. Thus, unless the context dictates otherwise (e.g., when sample preparation or collection steps are described), all steps of the methods described herein are computer-implemented.

[0045] According to a further aspect, there is provided a method of providing a prognosis to a subject diagnosed with cancer, the method comprising identifying a plurality of tumor-specific mutations in one or more samples from the subject, and determining the likelihood that each of the tumor-specific mutations is clonal using the method of any embodiment of the first aspect.

[0046] The method may further include classifying subjects as having a high or low burden of clonal neo-antigens depending at least in part on the proportion of tumor-specific mutations that have a probability of being clonal above a predetermined threshold, where subjects with a high burden of clonal neo-antigens have an improved prognosis compared to subjects with a low burden of clonal neo-antigens.

[0047] According to a further aspect, there is provided a method of providing immunotherapy to a subject diagnosed with cancer, comprising identifying one or more clonal neo-antigens using a method as described herein, such as the method according to any embodiment of the second aspect; and designing an immunotherapy that targets one or more of the identified clonal neo-antigens.

[0048] The method may have any one or more of the following features.

[0049] Immunotherapies targeting one or more of the clonal neo-antigens can be compositions comprising immunogenic compositions, immune cells, or therapeutic antibodies. Immunogenic compositions can include one or more of the clonal neo-antigens identified (e.g., neo-antigen peptides or proteins, or cells presenting the neo-antigens), or sufficient material for the expression of one or more of the identified clonal neo-antigens (e.g., DNA or RNA molecules encoding the neo-antigens). Compositions comprising immune cells can include T cells, B cells, and / or dendritic cells. Compositions comprising therapeutic antibodies can include one or more antibodies that recognize at least one of one or more of the identified clonal neo-antigens. The antibodies can be monoclonal antibodies.

[0050] In any embodiment of any aspect, the cancer may be selected from bladder cancer, gastric cancer, esophageal cancer, breast cancer, colorectal cancer, cervical cancer, ovarian cancer, endometrial cancer, renal cancer (renal cell), lung cancer (small cell, non-small cell, and mesothelioma), brain cancer (glioma, astrocytoma, glioblastoma), melanoma, lymphoma, small intestine cancer (duodenal and jejunal cancer), leukemia, pancreatic cancer, hepatobiliary tumors, germ cell cancer, prostate cancer, head and neck cancer, thyroid cancer, and sarcoma. The cancer may be lung cancer. The cancer may be melanoma. The cancer may be bladder cancer. The cancer may be head and neck cancer.

[0051] In any embodiment of any aspect, the subject may be a human.

[0052] Designing an immunotherapy targeting one or more of the identified clonal neo-antigens can include designing one or more candidate peptides for each of the one or more clonal neo-antigens being targeted, each peptide comprising at least a portion of the clonal neo-antigen being targeted.

[0053] The method may further include obtaining one or more candidate peptides. The method may further include testing the one or more candidate peptides for one or more properties. Testing may be performed in vitro or in silico. For example, the one or more peptides may be tested for immunogenicity, propensity to be presented by MHC molecules (optionally by specific MHC molecule alleles, which may be selected depending on the MHC alleles expressed by the subject), ability to induce proliferation of immune cell populations, etc.

[0054] The method may further include producing the immunotherapy. The method may further include obtaining a population of dendritic cells pulsed with one or more of the candidate peptides. The immunotherapy may be a composition comprising T cells that recognize at least one of the one or more of the identified clonal neo-antigens. The composition may be enriched for T cells that target at least one of the one or more of the identified clonal neo-antigens. The method may include obtaining a T cell population and expanding the T cell population to increase the number or relative proportion of T cells that target at least one of the one or more of the identified clonal neo-antigens.

[0055] The method may further include obtaining a T cell population. The T cell population may be isolated from a subject, for example, from one or more tumor samples obtained from the subject, or from a peripheral blood sample or other tissue sample from the subject. The T cell population may include tumor-infiltrating lymphocytes. The T cells may be isolated using methods well known in the art. For example, the T cells may be purified from a single cell suspension made from the sample based on the expression of CD3, CD4, or CD8. The T cells may be enriched from the sample by passing through a Ficoll-paque gradient.

[0056] The method may further comprise expanding the T cell population. For example, the T cells may be expanded by ex vivo culture in conditions known to provide mitogenic stimulation to the T cells. By way of example, the T cells may be cultured with cytokines such as IL-2 or with mitogenic antibodies such as anti-CD3 and / or CD28. The T cells may be co-cultured with antigen presenting cells (APCs), which may be irradiated. The APCs may be dendritic cells or B cells. The dendritic cells may be pulsed with peptides containing one or more of the identified neo-antigens as single stimuli or as a pool of stimulatory neo-antigen peptides. The expansion of the T cells may be performed using methods known in the art, including, for example, the use of artificial antigen presenting cells (aAPCs) that provide additional co-stimulatory signals, and autologous PBMCs presenting the appropriate peptides. The autologous PBMCs may be pulsed with peptides containing the neo-antigens as discussed herein as single stimuli or alternatively as a pool of stimulatory neo-antigens.

[0057] According to a further aspect, there is provided a method of expanding a T cell population for use in treating cancer in a subject, the method comprising identifying one or more clonal neo-antigens using a method as described herein, such as a method according to any embodiment of the second aspect; obtaining a T cell population comprising T cells capable of specifically recognizing one of the identified clonal neo-antigens; and co-culturing the T cell population with a composition comprising the identified clonal neo-antigen.

[0058] The method may have one or more of the following features.

[0059] The T cell population obtained may be assumed to comprise T cells capable of specifically recognizing one of the identified clonal neo-antigens. The method preferably comprises identifying a plurality of clonal neo-antigens. The T cell population may comprise a plurality of T cells, each of which has the ability to specifically recognize one of the identified clonal neo-antigens, and co-culturing the T cell population with a composition comprising the identified clonal neo-antigens. The co-culturing may result in the expansion of the T cell population specifically recognizing one or more neo-antigens. The expansion may be performed by co-culturing the T cells with the neo-antigen and antigen-presenting cells. The antigen-presenting cells may be dendritic cells. The expansion may thus be a selective expansion of the neo-antigen-specific T cells. The expansion may further comprise one or more non-selective expansion steps.

[0060] According to a further aspect, there is provided a composition comprising a population of T cells obtained or obtainable by a method according to any of the embodiments of the preceding aspects.

[0061] According to a further aspect, there is provided a composition comprising a neoantigen, a neoantigen-specific immune cell, or an antibody recognizing a neoantigen, for use in treating or preventing cancer in a subject, wherein said neoantigen has been identified as a clonal neoantigen using the methods described herein.

[0062] According to a further aspect, there is provided a composition comprising a neo-antigen, a neo-antigen-specific immune cell, or an antibody recognizing a neo-antigen, wherein said neo-antigen has been identified as a clonal neo-antigen using the methods described herein.

[0063] According to a further aspect, there is provided a cell or population of cells expressing a neo-antigen on its surface, wherein said neo-antigen has been identified as a clonal neo-antigen using the methods described herein.

[0064] According to a further aspect, there is provided a neoantigen, an immune cell recognizing a neoantigen, or an antibody recognizing a neoantigen, for use in treating or preventing cancer in a subject, wherein said neoantigen has been identified as a clonal neoantigen using the methods described herein.

[0065] According to a further aspect, there is provided the use of a neoantigen, an immune cell recognizing a neoantigen, or an antibody recognizing a neoantigen, in the manufacture of a medicament for use in treating or preventing cancer in a subject, wherein said neoantigen has been identified as a clonal neoantigen using the methods described herein.

[0066] According to a further aspect, there is provided a method of treating a subject diagnosed with cancer comprising administering an immunotherapy provided using the methods described herein, or a composition as described herein.

[0067] According to a further aspect, there is provided a system comprising: a processor; and a computer readable medium comprising instructions which, when executed by the processor, cause the processor to perform the steps of any method described herein, such as a method according to any embodiment of the first, second, third or fourth aspect above.

[0068] According to a further aspect, there is provided a non-transitory or more computer readable medium comprising instructions that, when executed by one or more processors, cause one or more processors to perform the steps of any method described herein, such as a method according to any embodiment of the first, second, third or fourth aspect above.

[0069] According to a further aspect, there is provided a computer program comprising code which, when executed on a computer, causes the computer to carry out the steps of any method described herein, such as the method according to any embodiment of the first, second, third or fourth aspect above. [Brief description of the drawings]

[0070] BRIEF DESCRIPTION OF THE DRAWINGS [Figure 1] FIG. 1 is a flow chart outlining a method for determining whether a tumor-specific mutation is likely to be clonal, and its use in identifying clonal neoantigens. [Diagram 2] 1 is a flow chart generally illustrating a method of providing immunotherapy. [Diagram 3]

[0023] An embodiment of a system for determining whether a tumor-specific mutation is likely to be clonal and / or for identifying clonal neo-antigens and / or for providing immunotherapy is shown. [Figure 4] 1 illustrates generally the structured populations postulated by the methods disclosed herein. [Diagram 5] 1 shows the results of the method for identifying clonal mutations described herein using synthetic data. Each plot shows the ROC curve (Receiver Operating Characteristic, True Positive Rate (TPR=TP / (TP+FN), where TP=true positives, FN=true negatives; sensitivity) as a function of false positive rate (FPR=FP / (FP+TN), where FP=false positives, TN=true negatives; 1-specificity) for different maximum copy numbers and sample numbers, with three separate curves each showing a different simulated read depth. A. Using a binomial emission model for simulation and identification. B. Using a beta-binomial emission model for simulation and a binomial model for identification. C. Using a beta-binomial emission model for simulation and identification. [Figure 6] The results of the method described herein for identifying clonal mutations are shown using data from a cell line mixing experiment with diploid (A) and aneuploid (B) cell lines. The title of each plot indicates which samples were included in the analysis. Each plot shows a ROC curve using the method as described herein (labeled "ACE"), as well as a single point showing the TPR and FPR obtained when running the same data with PyClone (Roth et al., 2014). [Figure 7] 1 shows the results of the method for identifying clonal mutations described herein using data from the TRACERx project. A. Summary of the analyzed data. Per sample, the number of clonal and subclonal SNVs identified using the TRACERx pipeline (manually curated PyClone results) is shown. B. Distribution of the probability of ubiquity predicted from the method as described herein. (Left) Results when considering all ASCAT (Van Loo et al., 2010) ploidy / purity solutions weighted by probability. (Right) Results when selecting the most likely solution from ASCAT. C. ROC curves illustrating the predictive performance of the method as described herein (taking into account all possible solutions from ASCAT weighted by probability and not) using manually curated PyClone (Roth et al., 2014) calls as ground truth. [Figure 8] The results of a run-time comparison between the method as described herein and a control method are shown. The control method (PyClone-VI) has been heavily optimized for computational efficiency but is not parallelizable. The method described herein (ACE) has not been subjected to extensive computational efficiency optimization but is more efficient by design and is parallelizable (implementations with 2, 3, and 4 CPUs are shown). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0071] Detailed Description Given its importance in cancer therapy and prognosis, the present method has obvious clinical applications in identifying clonal mutations and neoantigens in cancer patients. Several methods have been proposed that attempt to reconstruct the clonal structure of tumor samples (Schwartz and Schaeffer, 2017). However, all of these approaches suffer from various shortcomings that limit their clinical utility. Thus, the inventors have determined that there is still a need for improved methods to determine whether tumor-specific mutations are likely to be clonal. This problem of identifying clonal mutations in cancer (mutations present in all cells of the cancer) is far from trivial. In fact, primary samples contain a mixture of cells (including normal and cancer cells), and cancer genomes often have extensive heterogeneous copy number variations. Thus, the allele prevalence of a mutation is the result of a combination of factors, including the proportion of normal cells, the proportion of tumor cells that carry the mutation, and the allele copy number of the mutation in each of those tumor cells. This is compounded by the technical noise associated with any experimental protocol (e.g., sequencing errors), the limitations of those protocols (e.g., uncertainties in short-read alignments), and the inherent limitation that primary samples represent a partial snapshot of the cancer state. Roth et al. (2014) attempted to address this challenge by proposing a Bayesian clustering method (called "PyClone") that classifies a collection of deep-sequenced somatic mutations into clusters that share the same evolutionary history. This is done by estimating their cellular prevalence and taking into account allelic imbalance introduced by segment-wise copy number changes and contamination of normal cells.

[0072] Although this approach is advantageous in that it highlights the clarity of its statistical framework, the inventors have found that it is less suitable for use in non-research environments for several reasons. First, it is relatively slow and operates at the level of a set of mutations (such as a large set of deep sequenced mutations, or more recently, a complete genome) rather than individual mutations. Both of these drawbacks are a consequence of the fact that the method infers cluster structure for the entire set of mutations. Therefore, this approach solves a complex problem that requires the calculation of the posterior density for model parameters including the cellular prevalence of each mutation in the set of all mutations identified in multiple samples from cancer, as well as the cluster structure for this set of mutations. The inventors have found that this is undesirable for at least two reasons: it is unnecessarily slow when cluster structure for the entire set of mutations is not the objective, and it makes it impossible to track any mutation through the clinical pipeline independently of other mutations that are analyzed at the same time. In fact, the identification of any mutation as potentially clonal is inherently tied to other mutations that are analyzed together. Thus, any change in the set of mutations to be analyzed requires a rerun of the pipeline, potentially producing different results. In contrast, our proposed approach analyzes each mutation independently and does not infer any structure that depends on the population of mutations. This results in a clear improvement in computer technology, since our algorithm has a lower computational complexity than the prior art (PyClone), and therefore requires less computational resources (processor, memory and / or network resources) overall and for each mutation. Furthermore, since there are no dependencies between mutations, the processing of the set of mutations can be fully parallelized. This results in a further reduction in complexity, increased processing speed, and improved tractability, since a problem with a single mutation in one run does not affect the runs of other mutations.

[0073] In addition, the prior art methods require manual intervention by experts to convert the output into an index of clonality by analyzing the cluster structure, selecting clusters of mutations assumed to be clonal, and prioritizing the mutations within the selected clusters. Finally, the methods also require manual intervention by experts to select appropriate copy number estimates, with incorrect copy number solutions having a large impact on the clustering results, and lack an automated method to set copy numbers with the required level of reliability. Such multiple levels of manual intervention are problematic, not least because they require the presence of (human) experts in a task that could be performed more efficiently by a machine. In fact, the process that experts apply to select copy number solutions, selecting clusters of mutations assumed to be clonal, and prioritizing the mutations within those clusters, simply cannot be reliably automated. This process requires expertise and relies on experts using their judgment to make choices that are ultimately at least somewhat subjective. It cannot be automated, and it cannot be reliably reproduced due to the subjectivity involved in the process. In contrast, according to the method of the present invention, none of these manual interventions are required. Multiple copy number solutions and their confidence levels can be seamlessly integrated (thus eliminating the need for any selection, whether manual or automatic), and clonality likelihood is predicted for each mutation individually (thus eliminating the need to pick out clusters of mutations or prioritize within them, and allowing a natural, predictable, and reproducible prioritization to be defined based on predicted likelihoods). In this way, clonal mutations are identified with a higher degree of confidence than was previously possible, enabling uses beyond pure research into the field of therapeutic drug development.The potential use of such clonal mutations in cancer treatment and cancer monitoring and management has been envisioned, but their real-world use has been hampered in practice by the lack of a means to reliably identify these mutations for incorporation into clinical pipelines, so the availability of reliable estimates of the likelihood of clonality for individual mutations represents a distinct improvement for these fields.

[0074] In this disclosure, the following terms will be used, and are intended to be defined as indicated below.

[0075] A "sample" as used herein may be a cell or tissue sample, biological fluid, extract (e.g., a DNA extract obtained from a subject) from which genomic material can be obtained for genomic analysis, such as genomic sequencing (e.g., whole genome sequencing, whole exome sequencing). A sample may be a cell, tissue, or biological fluid sample (e.g., a biopsy) obtained from a subject. Such a sample may be referred to as a "subject sample". In particular, a sample may be a blood sample, or a tumor sample, or a sample derived therefrom. A sample may be freshly obtained from a subject, or may have been processed and / or stored (e.g., frozen, fixed, or subjected to one or more purification, concentration, or extraction steps) prior to genomic analysis. A sample may be a cell or tissue culture sample. Thus, a sample as described herein may refer to any type of sample that includes cells or genomic material derived therefrom, whether from a biological sample obtained from a subject, or from a sample obtained, for example, from a cell line. In an embodiment, a sample is a sample obtained from a subject, such as a human subject. The sample is preferably from a mammal (e.g., a mammalian cell sample or a sample from a mammalian subject such as a cat, dog, horse, donkey, sheep, pig, goat, cow, mouse, rat, rabbit or guinea pig), preferably from a human (e.g., a human cell sample or a sample from a human subject). Furthermore, the sample may be transported and / or stored, collection may occur at a location remote from the genomic sequence data collection (e.g., sequencing) site, and / or any computer-implemented method steps described herein may occur at a location remote from the sample collection site and / or remote from the genomic data collection (e.g., sequencing) site (e.g., computer-implemented method steps may be performed using networked computers, such as using a "cloud" provider).

[0076] A "mixed sample" refers to a sample that is hypothesized to contain multiple cell types or genetic material from multiple cell types. In the context of the present disclosure, a mixed sample is typically one that contains or is hypothesized (expected) to contain tumor cells or genetic material from tumor cells. A sample obtained from a subject, such as a tumor sample, is typically a mixed sample (unless it has been subjected to one or more purification and / or separation steps). Typically, the sample contains tumor cells and at least one other cell type (and / or genetic material derived therefrom). For example, a mixed sample may be a tumor sample. A "tumor sample" refers to a sample that is derived from or obtained from a tumor. Such a sample may contain tumor cells and normal (non-tumor) cells. The normal cells may include immune cells (such as lymphocytes) and / or other normal (non-tumor) cells. The lymphocytes in such a mixed sample may be referred to as "tumor infiltrating lymphocytes" (TILs). The tumor may be a solid tumor or a non-solid or hematological tumor. The tumor sample may be a primary tumor sample from a subject, a tumor-associated lymph node sample, or a sample from a metastatic site. The sample containing tumor cells or genetic material derived from tumor cells may be a bodily fluid sample. Thus, the genetic material derived from tumor cells may be circulating tumor DNA or tumor DNA in exosomes. Alternatively, or in addition, the sample may contain circulating tumor cells. The mixed sample may be a sample of cells, tissues, or bodily fluids that have been subjected to processing to extract genetic material. Methods for extracting genetic material from biological samples are known in the art. The mixed sample may have been subjected to one or more processing steps that may change the ratio of multiple cell types or genetic material derived from multiple cell types present in the sample. For example, a mixed sample containing tumor cells may have been subjected to processing to enrich the sample with respect to tumor cells. Thus, a sample of purified tumor cells may be referred to as a "mixed sample" based on the fact that, even though for certain purposes, the sample may be assumed to be pure (i.e., have a tumor fraction of 1 or 100%), small amounts of other types of cells may be present.

[0077] The term "tumor fraction" (also sometimes referred to as "tumor purity" or simply "purity" or aberrant cell fraction (ACF)) refers to the proportion of DNA-containing cells in a mixed sample that are tumor cells, or an equivalent proportion assumed to give rise to a particular mixture of genetic material from tumor and non-tumor cells in the sample. Methods for determining tumor fraction in a sample are known in the art. For example, in the context of a cell or tissue sample, tumor fraction may be estimated by analyzing pathology slides (e.g., hematoxylin and eosin (H&E) stained slides or other histochemistry or immunohistochemistry slides, by counting tumor cells in one or more representative areas of the sample), or by using high-throughput assays such as flow cytometry. In the context of samples containing genetic material, tumor fraction may be estimated using sequence analysis processes that attempt to deconvolute the tumor and germline genomes, such as ASCAT (Van Loo et al., 2010), ABSOLUTE (Carter et al., 2012), or ichorCNA (Adalsteinsson et al., 2017).

[0078] A "normal sample" or "germline sample" refers to a sample that is assumed to be free of tumor cells or genetic material derived from tumor cells. A germline sample may be a blood sample, a tissue sample, or a purified sample, such as a sample of peripheral blood mononuclear cells from a subject. Similarly, the terms "normal", "germline" or "wild type", when referring to a sequence or genotype, refer to the sequence / genotype of cells other than tumor cells. A germline sample may contain a small percentage of tumor cells or genetic material derived therefrom, but may nevertheless be assumed to be practically free of said cells or genetic material. In other words, all cells or genetic material may be assumed to be normal, and / or sequence data that does not fit this assumption may be disregarded.

[0079] The term "sequence data" refers to information that indicates the presence and preferably also the amount of genomic material with a particular sequence in a sample. Such information can be obtained using sequencing technologies, such as next generation sequencing (NGS), such as whole exome sequencing (WES), whole genome sequencing (WGS), or sequencing of captured genomic loci (targeted or panel sequencing), or using array technologies, such as copy number variation arrays, or other molecular counting assays. When NGS technologies are used, the sequence data can include a count of the number of sequencing reads with a particular sequence. When non-digital technologies, such as array technologies, are used, the sequence data can include a signal (e.g., intensity value) that indicates the number of sequences with a particular sequence in the sample, for example, by comparison with an appropriate control. The sequence data can be mapped to a reference sequence, such as a reference genome, using methods known in the art, such as, for example, Bowtie (Langmead et al., 2009). In this way, the count of the sequencing reads or an equivalent non-digital signal can be linked to a specific genomic location (where "genomic location" refers to the location in the reference genome to which the sequence data was originally mapped). Furthermore, the genomic location may include a mutation, in which case the count of the sequencing reads or an equivalent non-digital signal can be linked to each of the possible variants (also referred to as "alleles") at a specific genomic location. The process of identifying the presence of a mutation at a specific location in a sample is referred to as "variant calling" and can be performed using methods known in the art (e.g., GATK HaplotypeCaller, https: / / gatk.broadinstitute.org / hc / en-us / articles / 360037225632-HaplotypeCaller, etc.).For example, sequence data may include a count of the number of reads (or an equivalent non-digital signal) that match a germline (also sometimes referred to as a "reference") allele at a particular genomic location, and a count of the number of reads (or an equivalent non-digital signal) that match a mutated (also sometimes referred to as an "alternate") allele at a genomic location.

[0080] Furthermore, the sequence data can be used to infer copy number profiles along the genome using methods known in the art. The copy number profile may be allele-specific. In the context of the present disclosure, the copy number profile is preferably allele-specific and tumor / normal sample specific. In other words, the copy number profile used in the present disclosure is preferably obtained using a method designed to analyze a sample containing a mixture of tumor cells and normal cells to generate allele-specific copy number profiles for the tumor cells and normal cells in the sample. The allele-specific copy number profile of the mixed sample can be obtained from the sequence data (e.g., using read counts as described above), for example, using ASCAT (Van Loo et al., 2010). Other methods are known and equally suitable. Preferably, in the context of the present disclosure, the method used to obtain the allele-specific copy number profile is one that reports multiple possible copy number solutions and associated quality / confidence metrics. For example, ASCAT outputs a goodness of fit metric for each combination of ploidy (not segment specific, but total tumor sample ploidy) values ​​and the purity when evaluating the corresponding allele-specific copy number profile. Note that the tumor-specific copy number profile generated by such methods represents an average or summary of the entire tumor cell population (i.e., it does not take into account the heterogeneity within the tumor population, which is the subject of the novel developments described herein).

[0081] The term "total copy number" refers to the total number of copies of a genomic region in a sample. The term "major copy number" refers to the number of copies of the most prevalent allele in a sample. Conversely, the term "minor copy number" refers to the number of copies of an allele other than the most prevalent allele in a sample. Unless otherwise indicated, these terms refer to the inferred major and minor copy numbers (and total copy numbers) for an inferred tumor copy number profile. The term "normal copy number" or "normal total copy number" refers to the number of copies of a genomic region of a normal cell in a sample. A normal cell typically has two copies of each chromosome (unless the cell is genetically male and the chromosome is a sex chromosome), therefore the normal copy number may be assumed to be equal to 2 in embodiments (unless the genomic region is on the X or Y chromosome and the sample under analysis is from a male subject, in which case the normal copy number may be assumed to be equal to 1). Alternatively, the normal copy number of a particular genomic region may be determined using a normal sample.

[0082] The term "log R value" (sometimes referred to as "logR", "logRR", "LLR") refers to a measure of normalized total signal intensity that quantifies the total copy number at a genomic locus. In the context of the present disclosure, the term typically refers to the log R value for a sample that contains tumor genetic material, and normalization is typically performed relative to a normal sample (preferably a matched normal sample, but may also be a process-matched normal sample or other suitable normal reference sample). For example, when NGS is used, log R can be determined as a logarithmic transformation after read depth normalization (log(read depth tumor / read depth normal)). The term "mean B allele frequency" (MBAF, sometimes referred to as "B allele frequency" (BAF)) is a measure of normalized allele intensity ratio at a genomic location. In the context of the present disclosure, this term typically refers to the BAF value for a sample containing tumor genetic material, and normalization is typically performed relative to a normal sample (which is preferably a matched normal sample, but may also be a process-matched normal sample or other suitable normal reference sample). For example, the BAF can be determined as the ratio of allele frequencies for tumor alleles versus normal alleles. Copy number profiles typically include copy number estimates for genomic regions called "segments." Thus, the BAF and logR associated with a genomic location may refer to the BAF and logR of a segment that overlaps a particular genomic location (such as the genomic location of a mutation). Furthermore, the BAF and logR can be used to determine the corresponding major and minor copy numbers. In embodiments, copy number metric values ​​can be provided for both tumor and normal copy number profile estimates, even if only tumor copy number profile values ​​are used.

[0083] The terms "tumor-specific mutation", "somatic mutation" or simply "mutation" are used interchangeably and refer to a difference in nucleotide sequence (e.g. DNA or RNA) in tumor cells compared to healthy cells from the same subject. This difference in nucleotide sequence may result in the expression of a protein that is not expressed in healthy cells from the same subject. For example, the mutation may be a single nucleotide variation (SNV), a multi-nucleotide variation (MNV), a deletion mutation, an insertion mutation, a translocation, a missense mutation, a translocation, a fusion, a splice site mutation, or any other change in the genetic material of the tumor cell. The mutation may result in the expression of a protein or peptide that is not present in healthy cells from the same subject. Mutations may be identified by exome sequencing, RNA sequencing, whole genome sequencing and / or targeted gene panel sequencing and / or routine single gene Sanger sequencing, followed by sequence alignment and comparison of DNA and / or RNA sequences from tumor samples to DNA and / or RNA from a reference sample or reference sequence (e.g. germline DNA and / or RNA sequences, or reference sequences from a database). Suitable methods are known in the art.

[0084] "Indel mutation" refers to the insertion and / or deletion of bases in the nucleotide sequence (e.g., DNA or RNA) of an organism. Typically, indel mutations appear in the DNA, preferably genomic DNA, of an organism. In embodiments, the indel can be 1-100 bases, such as 1-90, 1-50, 1-23 or 1-10 bases. The indel mutation may be a frameshift indel mutation. A frameshift indel mutation is a change in the reading frame of a nucleotide sequence caused by the insertion or deletion of one or more nucleotides. Such frameshift indel mutations may generate a novel open reading frame, which is typically highly distinct from the polypeptide encoded by the non-mutated DNA / RNA in the corresponding healthy cells of the subject.

[0085] "Neo-antigens" (or "neo-antigens") are antigens that arise as a result of mutations in cancer cells. Thus, neo-antigens are not expressed (or expressed at significantly lower levels) by normal (i.e., non-tumor) cells. When neo-antigens are processed, distinct peptides may be generated that can be recognized by T cells when presented in the context of MHC molecules. As described herein, neo-antigens can be used as the basis of cancer immunotherapy. When referring to "neo-antigens" herein, it is intended to include peptides derived from neo-antigens as well. The term "neo-antigens" as used herein is intended to encompass any portion of neo-antigens that is immunogenic. An "antigenic" molecule, as referred to herein, is a molecule that has the ability to stimulate an immune response, either by itself or a portion thereof, when presented to the immune system or immune cells in an appropriate manner. Binding of neo-antigens to specific MHC molecules (encoded by specific HLA alleles) can be predicted using methods known in the art. Examples of methods for predicting MHC binding include those described by Lundegaard et al., O'Donnel et al., and Bullik-Sullivan et al. For example, MHC binding of neoantigens can be predicted using the netMHC-3 (Lundegaard et al.) and netMHCpan4 (Jurtz et al.) algorithms. Neoantigens predicted to bind to specific MHC molecules are then predicted to be presented by said MHC molecules on the cell surface.

[0086] A "clonal neo-antigen" (also sometimes referred to as a "truncal neo-antigen") is a neo-antigen that arises by a mutation that is present in essentially all of the tumor cells in one or more samples from a subject (or that can be assumed to be present in essentially all of the original tumor cells from which the tumor genetic material in one or more samples is derived). Similarly, a "clonal mutation" (also sometimes referred to as a "truncal mutation") is a mutation that is present in essentially all of the tumor cells in one or more samples from a subject (or that can be assumed to be present in essentially all of the original tumor cells from which the tumor genetic material in one or more samples is derived). Thus, a clonal mutation can be a mutation that is present in all tumor cells in one or more samples from a subject. A "subclonal" neo-antigen is a neo-antigen that arises by a mutation that is present in a subset or percentage of cells in one or more tumor samples from a subject (or that can be assumed to be present in a subset of tumor cells from which the tumor genetic material in one or more samples is derived). Similarly, a "subclonal" mutation is one that is present in a subset or proportion of cells in one or more tumor samples from a subject (or that can be hypothesized to be present in a subset of tumor cells from which the tumor genetic material in one or more samples is derived). A neoantigen or mutation may be clonal in the context of one or more samples from a subject, but may not be truly clonal in the context of the entire population of tumor cells that may be present in the subject (e.g., including all areas of the primary tumor and metastases). Thus, a clonal mutation may be "truly clonal" in the sense that it is a mutation that is present in essentially all of the subject's tumor cells (i.e., in every tumor cell), since one or more samples may not be representative of each and every subset of cells present in the subject.Thus, within the context of the present disclosure, a "clonal neoantigen" or "clonal mutation" may also be referred to as a "ubiquitous neoantigen" or "ubiquitous mutation," indicating that the neoantigen is present in essentially every tumor cell analyzed, but may not be present in every tumor cell that may be present in a subject. The terms "clonal" and "ubiquitous" are used synonymously, unless the context indicates that a reference to "true clonality" was intended. The expression "essentially all tumor cells," in the context of one or more samples or subjects, may refer to at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 91%, at least 92%, at least 93%, at least 94%, at least 95%, at least 96%, at least 97%, at least 98%, or at least 99% of the tumor cells in one or more samples or subjects.

[0087] Nevertheless, neoantigens / mutations identified as likely to be clonal (or "ubiquitous") as described herein are likely to be truly clonal, or at least more likely to be truly clonal than neoantigens / mutations identified as not likely to be clonal. Furthermore, when one or more samples used to identify a clonal neoantigen / mutation capture a more complex genetic diversity picture of the tumor (e.g., by including multiple samples from a subject, such as samples from different regions of the tumor, and / or by including samples that uniquely capture the diversity of tumor cells, such as ctDNA samples), the confidence in the likelihood that a clonal neoantigen / mutation identified herein is truly clonal increases. Conversely, for neoantigens / mutations identified as not likely to be clonal as described herein, the identification of a neoantigen / mutation as not likely to be clonal, even with the limited view provided by the sample extraction process, indicates that there is evidence that the neoantigen / mutation is not present in all tumor cells, making it less likely to be truly clonal. In this way, the process of identifying clonal neoantigens / mutations can be viewed as prioritizing which candidate neoantigens / mutations are most likely to be clonal based on a limited view of the clonal architecture of a subject's tumor available from one or more samples.

[0088] The term "cancer cell fraction" (or "CCF") refers to the proportion of tumor cells that contain mutations, such as mutations that give rise to a particular neo-antigen. In the context of the present disclosure, the cancer cell fraction may be estimated based on one or more samples and may therefore not be equal to the true cancer cell fraction in a subject (as explained above). Nevertheless, a cancer cell fraction estimated based on one or more samples may provide a useful indication of the likely true cancer cell fraction. Furthermore, as explained above, the accuracy of such an estimate may be increased when the sample or samples used to estimate the cancer cell fraction capture a more complete genetic diversity picture of the tumor. Additional noise sources and confounding factors in genomic data mean that the cancer cell fraction determined from one or more samples represents an estimate. Thus, while truly clonal mutations / neo-antigens should have a CCF=1, in practice mutations / neo-antigens that are more likely to be clonal are expected to be associated with higher CCF estimates (which may not be equal to 1) than mutations that are less likely to be clonal, which are expected to be associated with lower CCF estimates.

[0089] For example, the cancer cell fraction estimate can be obtained by integrating the mutant allele frequency with the copy number and purity estimates, as described by Landau et al. (2013). Such CCF estimates can also be used to identify mutations that may be clonal. For example, a clonal mutation can be defined as a mutation having an estimated cancer cell fraction (CCF) ≥ 0.75, such as CCF ≥ 0.80, 0.85, 0.90, 0.95, or 1.0. A subclonal mutation can be defined as a mutation having a CCF < 0.95, 0.90, 0.85, 0.80, or 0.75. Further, the CCF estimate may be related to (e.g., derived from) a distribution that associates a probability with each of a plurality of possible values of CCF between 0 and 1, from which a statistical estimate of confidence can be obtained. For example, a mutation can be defined as potentially clonal if the 95% CCF confidence interval is ≥ 0.75, i.e., the upper limit of the 95% confidence interval of the estimated CCF is 0.75 or greater. In other words, a mutation can be defined as potentially clonal if the interval of CCF with a lower limit L and an upper limit H is such that P(L < CCF < H) = 95% and H ≥ 0.75. Alternatively, a mutation can be identified as clonal if the chance or probability that its cancer cell fraction (CCF) reaches or exceeds the required value as defined above, e.g., 0.75 or 0.95, is higher than 50%, such as 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or more. In other words, a mutation can be identified as clonal if P(CCF > 0.75) ≥ 0.5. For example, a mutation may be classified as potentially clonal or subclonal based on whether the posterior probability that its CCF exceeds 0.95 (or 0.75, or any other selected threshold) is greater than or less than 0.5, respectively.

[0090] According to the method of the present disclosure, the likelihood that a mutation is clonal is determined, as further explained below. This is equivalent to P(CCF=1). In this regard, as further explained below, a mutation may be identified as likely to be clonal if P(CCF=1) exceeds a threshold. The threshold may be fixed. For example, a mutation may be identified as likely to be clonal if P(CCF=1)>0.05. Alternatively, the threshold may be determined for a particular set of mutations under investigation. In embodiments, the threshold may be set to reach a predetermined precision and / or recall based on a benchmark dataset that includes known clonal / non-clonal status. The benchmark dataset may be determined using synthetic data and / or using a dataset obtained from a population with known clonality structure (e.g., cell line mixture data). For example, a mutation may be identified as likely to be clonal if P(CCF=1)>t, where t is the maximum value such that 95% (or any other value, e.g., 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, etc.) of the truly clonal mutations in the benchmark dataset are identified (i.e., at most a 5% false negative rate). As another example, a mutation may be identified as likely to be clonal if P(CCF=1)>t, where t is the minimum value such that at least 50% (or any other value, e.g., 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, etc.) of the mutations above the threshold in the benchmark dataset are truly clonal mutations (i.e., at least a 50% true positive rate). Alternatively, a threshold may be set such that any mutation (or a certain % of mutations) associated with an estimated CCF with a confidence interval that meets the criteria described above (e.g., that the upper limit of the 95% confidence interval of the estimated CCF is 0.75 or greater) is selected as potentially clonal.Alternatively, the threshold may be set such that any mutation (or a certain % of mutations) associated with a putative CCF that has a posterior probability distribution that meets the criteria described above (e.g., a posterior probability that the CCF exceeds 0.95 (or 0.75, or any other chosen threshold) is greater than 0.5) is selected as potentially clonal.

[0091] Cancer immunotherapy (or simply "immunotherapy") refers to a therapeutic approach that includes the administration to a subject of an immunogenic composition (e.g., a vaccine), a composition comprising immune cells, or an immunoactive drug, such as a therapeutic antibody. The term "immunotherapy" may also refer to the therapeutic composition itself. In the context of the present disclosure, immunotherapy typically targets a neo-antigen. For example, an immunogenic composition or vaccine may include a neo-antigen, a neo-antigen-presenting cell, or materials necessary for the expression of the neo-antigen. As another example, a composition comprising immune cells may include T and / or B cells that recognize the neo-antigen. Immune cells may be isolated from a tumor or other tissue (including, but not limited to, lymph nodes, blood, or ascites), expanded ex vivo or in vitro, and readministered to the subject (a process referred to as "adoptive cell therapy"). Alternatively, or in addition, T cells may be isolated from a subject and engineered to target the neo-antigen (e.g., by insertion of a chimeric antigen receptor that binds the neo-antigen), and readministered to the subject. As another example, a therapeutic antibody may be an antibody that recognizes the neo-antigen. One skilled in the art will understand that when the neo-antigen is a cell surface antigen, the antibody as referred to herein will recognize the neo-antigen. When the neo-antigen is an intracellular antigen, the antibody will recognize a neo-antigen peptide-MHC complex. As referred to herein, an antibody that "recognizes" a neo-antigen encompasses both of these possibilities. Furthermore, immunotherapy may target multiple neo-antigens. For example, an immunogenic composition may include multiple neo-antigens, cells that present multiple neo-antigens, or materials necessary for the expression of multiple neo-antigens. As another example, a composition may include immune cells that recognize multiple neo-antigens. Similarly, a composition may include multiple immune cells that recognize the same neo-antigen. As another example, a composition may include multiple therapeutic antibodies that recognize multiple neo-antigens. Similarly, a composition may include multiple therapeutic antibodies that recognize the same neo-antigen.

[0092] The composition as described herein may be a pharmaceutical composition that further comprises a pharma- ceutically acceptable carrier, diluent or excipient.The pharmaceutical composition may optionally comprise one or more additional pharma- ceutically active polypeptides and / or compounds.Such a formulation may be, for example, in a form suitable for intravenous infusion.

[0093] When referring to "immune cells," it is intended to encompass cells of the immune system, such as T cells, NK cells, NKT cells, B cells, and dendritic cells. In a preferred embodiment, the immune cells are T cells. The neo-antigen-recognizing immune cells may be engineered T cells. The neo-antigen-specific T cells may express a chimeric antigen receptor (CAR) or T cell receptor (TCR) that specifically binds to the neo-antigen or neo-antigenic peptide, or an affinity-enhanced T cell receptor (TCR) that specifically binds to the neo-antigen or neo-antigenic peptide (as discussed further herein below). For example, the T cells may express a chimeric antigen receptor (CAR) or T cell receptor (TCR) that specifically binds to the neo-antigen or neo-antigenic peptide (e.g., an affinity-enhanced T cell receptor (TCR) that specifically binds to the neo-antigen or neo-antigenic peptide). Alternatively, the population of neo-antigen-recognizing immune cells may be a population of T cells isolated from a subject with a tumor. For example, the T cell population may be generated from T cells that are in a sample isolated from the subject, such as, for example, a tumor sample, a peripheral blood sample, or a sample from another tissue of the subject. The T cell population may be generated from a sample from a tumor in which a neo-antigen is identified. In other words, the T cell population may be isolated from a sample derived from a tumor of a patient to be treated, where a neo-antigen is also identified from a sample from said tumor. The T cell population may include tumor infiltrating lymphocytes (TILs).

[0094] The term "antibody" (Ab) includes monoclonal antibodies, polyclonal antibodies, multispecific antibodies (e.g., bispecific antibodies), and antibody fragments that exhibit the desired biological activity. The term "immunoglobulin" (Ig) may be used synonymously with "antibody." After suitable neo-antigens have been isolated, for example by the methods of the present disclosure, antibodies can be made using methods known in the art.

[0095] An "immunogenic composition" is a composition capable of inducing an immune response in a subject. The term is used synonymously with the term "vaccine". The immunogenic composition or vaccine described herein can lead to the generation of an immune response in a subject. The "immune response" that can be generated can be humoral and / or cell-mediated immunity, e.g., stimulation of antibody production, or stimulation of cytotoxic or killer cells that can recognize and destroy (or otherwise eliminate) cells expressing an antigen on their surface that corresponds to the antigen in the vaccine. The immunogenic composition can include one or more neo-antigens, or materials necessary for the expression of one or more neo-antigens. In addition, the neo-antigens can be delivered in the form of cells, such as antigen-presenting cells, e.g., dendritic cells. Antigen-presenting cells, such as dendritic cells, can be pulsed or loaded with neo-antigens or neo-antigen peptides, or can be genetically modified (by DNA or RNA transfer) to express one, two or more neo-antigens or neo-antigen peptides, e.g., 2, 3, 4, 5, 6, 7, 8, 9 or 10 neo-antigens or neo-antigen peptides. Methods for preparing dendritic cell immunogenic compositions or vaccines are known in the art.

[0096] Neo-antigenic peptides may be synthesized using methods known in the art. The term "peptide" is used in its ordinary sense to mean a series of residues, typically L-amino acids, connected to one another by peptide bonds, typically between the a-amino group and the carboxyl group of adjacent amino acids. The term includes modified peptides and synthetic peptide analogs. Neo-antigenic peptides may contain cancer cell specific mutations (e.g., non-silent amino acid substitutions encoded by single nucleotide variants (SNVs)) at any residue position within the peptide. By way of example, peptides capable of binding to MHC class I molecules are typically 7-13 amino acids in length. Thus, amino acid substitutions may occur at positions 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 or 13 in a peptide containing 13 amino acids. In embodiments, longer peptides, such as 21-31 mers, may be used and may have mutations at any position, such as in the middle of the peptide, e.g., at positions 10, 11, 12, 13, 14, 15 or 16. Such peptides can also be used to stimulate both CD4 and CD8 cells to recognize neo-antigens.

[0097] As used herein, "treatment" refers to reducing, alleviating, or eliminating one or more symptoms of the disease under treatment compared to the symptoms prior to treatment. "Prevention" (or prophylaxis) refers to delaying or preventing the onset of disease symptoms. Prevention may be absolute (preventing the disease from occurring) or may be effective only in some individuals or for a limited period of time.

[0098] As used herein, the term "computer system" includes hardware, software, and data storage devices for implementing the system or performing the method according to the above-described embodiments. For example, a computer system may include a central processing unit (CPU), input means, output means, and data storage, which may be embodied as one or more connected computing devices. Preferably, the computer system has a display or includes a computing device with a display to provide a visual output display (e.g., in designing a business process). Data storage may include RAM, disk drives, or other non-transitory computer-readable media. A computer system may include multiple computing devices connected by a network and capable of communicating with each other over the network. It is expressly envisioned that a computer system may consist of or include a cloud computer.

[0099] As used herein, the term "computer-readable medium" includes, without limitation, any non-transitory medium or media that can be directly read and accessed by a computer or computer system. Media may include, but are not limited to, magnetic storage media, such as floppy disks, hard disk storage media, and magnetic tape; optical storage media, such as optical disks or CD-ROMs; electrical storage media, such as memory, including RAM, ROM, and flash memory; and hybrids and combinations of the above, such as magnetic / optical storage media.

[0100] Identification of clonal mutations The present disclosure provides a method for determining whether a tumor-specific mutation is likely to be clonal using sequence data from one or more samples containing tumor cells or genetic material derived therefrom. The present disclosure also provides a method for identifying clonal neo-antigens, comprising determining whether one or more tumor-specific mutations are likely to be clonal. An exemplary method will be described by reference to FIG. 1. In optional step 10, a sample containing genomic material from a tumor may be obtained from a subject. The sample is typically a sample containing mixed genomic material from multiple cell types, including tumor cells. Preferably, a corresponding sample may be obtained or previously obtained that does not contain genomic material from tumor cells or from which genomic material from normal cells may be extracted. A corresponding sample is a sample obtained from the same subject as the tumor sample. The use of a corresponding normal sample improves the accuracy of calling somatic (tumor-specific) mutations, since any variant position identified in the tumor sample can be compared to variant positions in the corresponding normal sample to exclude germline variants. The same corresponding normal sample may be used to analyze multiple tumor samples from a subject. Additionally, the corresponding sample and one or more tumor samples may be obtained at different time points. For example, the first tumor sample and corresponding sample can be obtained at the time of diagnosis or tumor resection, and additional tumor samples can be obtained and analyzed together with the first corresponding sample at a later time.When corresponding samples are not available, reference samples or genomes containing common somatic variants can be used.Alternatively, processed corresponding normal samples can be used, which do not have to be obtained from the same subject, or can be obtained from a pool of subjects.

[0101] In optional step 12, the sequence content of one or more mixed samples and optionally corresponding samples may be determined, for example, by sequencing the genomic material in the samples using one of whole exome sequencing or whole genome sequencing. Alternative methods, such as allele-specific copy number arrays, may be used, but sequencing methods are preferred because they generate a digital output representative of the number of each specific sequence in the sample. In optional step 14, the sequence data may be analyzed to identify one or more mutations that are present in the tumor cells but may not be present in the non-cancerous cells. These represent tumor-specific mutations and can be used as candidate neo-antigens. This may include aligning sequences from one or more samples (i.e., one or more mixed samples and one or more germline samples, if available) and identifying genomic locations where the tumor sequence differs from the germline sequence or may be assumed to differ from the germline sequence (e.g., if no germline sequence is available for the subject).

[0102] In step 16, sequence data of the mixed sample at the genomic location of the candidate tumor-specific mutation is obtained, including the count of reads supporting the mutant allele (also referred to as the "non-reference allele"), the count of reads supporting one or more germline alleles at that genomic location (A, collectively referred to as the "germline alleles" and also referred to as the "reference", "wild type" or "normal" allele if the locus is heterozygous in the germline population), and / or the total read count at the genomic location of the candidate tumor-specific mutation. Only two of these metrics need to be obtained, with the third being inferred from any two of these. The sequence data may alternatively or additionally include read data or intensity data from which the counts can be obtained. In optional step 18, information on at least one copy number solution matching each sample containing tumor genetic material may be obtained. This information may include allele-specific copy number metrics for the tumor fraction of the sample selected from major copy number, minor copy number, total copy number, average B allele frequency, log R value and tumor ploidy, and normal copy number, or information derived from these metrics, such as a set of candidate binding genotypes that fit these allele-specific copy number metrics. Not all such allele-specific copy number metrics are necessary, as some may contain overlapping information and / or be associated with suitable default values. For example, normal copy number may be associated with suitable default values ​​as explained above. Furthermore, only two of major copy number, total copy number and minor copy number are sufficient to infer the third. Similarly, these three values ​​can be inferred from MBAF and logR values ​​(and vice versa). Optionally, the copy number solution may be associated with a corresponding confidence metric. When such metrics are not available, each copy number solution may be assumed to be equally likely. Each candidate binding genotype includes genotypes at the position of the tumor-specific mutation for a normal population, a reference tumor population that does not contain the tumor-specific mutation, and a mutant tumor cell population that contains the tumor-specific mutation.

[0103] In step 20, the probability that the tumor-specific mutation is clonal is determined as a posterior probability that depends on the prior probability that the mutation is clonal and the probability that the sequence data is observed when the tumor-specific mutation is (i) clonal and (ii) non-clonal, given the tumor fraction and one or more candidate binding genotypes for each of the one or more samples. The prior probability is a probability that corresponds to a belief about the amount before any evidence is taken into account. In this context, the prior probability that the mutation is clonal may correspond to the probability that the mutation is clonal in the tumor, based on prior knowledge or assumptions, without taking into account the sequence data from the mixed sample. In step 22, it is determined whether the tumor-specific mutation is likely to give rise to a neo-antigen. For example, it may be determined whether the mutation is likely to give rise to a peptide or protein that is not expressed from germline cells (these genomes do not carry the mutation). This step can be performed at any time after step 14, and in particular does not have to be performed after steps 16-20. For example, candidate tumor-specific mutations may be filtered according to whether they are likely to give rise to neo-antigens before determining whether the tumor-specific mutation is likely to be clonal. In step 24, tumor-specific mutations that satisfy one or more criteria applied to the results of step 20 and one or more criteria applied to the results of step 22 may be identified. These may be considered to represent clonal neo-antigen candidates. In optional step 26, the results of any of the preceding steps (and in particular steps 20-24) may be provided to a user, for example via a user interface. These results may be used, for example, to provide an immunotherapy or prognosis to the subject, as further described below.

[0104] Applicable The above method finds application in the context of cancer diagnosis, prognosis and treatment procedures. In particular, the above method may be used to provide immunotherapy targeting clonal neo-antigens. Thus, also described herein is a method of providing immunotherapy to a subject, comprising identifying one or more clonal neo-antigens from one or more samples from the subject.

[0105] FIG. 2 illustrates in schematic form a method of providing an exemplary immunotherapy. In optional step 210, one or more samples containing tumor genetic material and one or more germline samples are obtained from a subject. The subject may be a subject diagnosed with cancer, and may (but need not be) the same subject to whom the immunotherapy is provided. In step 212, a list of clonal neo-antigen candidates is obtained using the methods described herein, for example, by reference to FIG. 1. The list may include a single neo-antigen, or multiple neo-antigens. Preferably, the list includes multiple neo-antigens. In step 214, an immunotherapy is designed that targets at least one (and optionally multiple) candidate neo-antigens. Designing such an immunotherapy may include identifying one or more candidate peptides for each of the clonal neo-antigen candidates (step 214A). For example, for at least one of the clonal neo-antigen candidates, multiple peptides may be designed that differ in their length and / or the location of sequence mutations that characterize the neo-antigen compared to the corresponding germline peptide. In step 214B, one or more of the identified peptides may be tested in vitro and / or in silico to evaluate one or more properties, such as their immunogenicity, their potential to be presented by MHC molecules, etc. In optional step 214C, one or more of the peptides may be selected, e.g., based on the results of step 214B.

[0106] In step 216, a selected peptide may be obtained. A peptide with a selected sequence may be obtained using any method known in the art, for example, using an expression system or by direct synthesis. In step 218, one or more candidate peptides may be used to generate an immunotherapy. The immunotherapy may include one or more candidate peptides or sufficient material for their expression (e.g., in the case of an immunogenic composition or vaccine), or may include molecules or cells obtained using the candidate peptides (e.g., in the case of a therapeutic antibody that selectively binds to a candidate peptide, or an immune cell that specifically recognizes a candidate peptide). In optional step 220, the immunotherapy may be administered to a subject, preferably the subject from which the sample used to identify the clonal neo-antigen was obtained. An example of generating an immunotherapy that includes a T cell population enriched for T cells that recognize one or more clonal neo-antigens is described. In step 218A, a T cell population may be obtained. The T cells may be obtained from the subject to be treated, but this is not required. The T cells may be obtained from a tumor sample, a blood sample, or any other tissue sample. In step 218B, a population of dendritic cells may be obtained. For example, the population of dendritic cells may be derived from mononuclear cells (e.g., peripheral blood mononuclear cells, PBMCs) from the subject to be treated. In step 218C, the population of dendritic cells may be pulsed with a candidate peptide. In step 218D, the population of pulsed dendritic cells may be used to selectively expand the T cell population. Additional expansion factors, such as cytokines or stimulatory antibodies, may be used.

[0107] Thus, the present disclosure also provides a T cell composition comprising a T cell population selectively enriched for T cells that recognize one or more clonal neo-antigens, where the one or more clonal neo-antigens are identified using any of the methods described herein.

[0108] In a T cell composition as described herein, the expanded population of neoantigen-responsive T cells may have higher activity than a non-expanded T cell population as measured by the response of the T cell population to restimulation with a neoantigen peptide. Activity may be measured by cytokine production, where higher activity is a 5-10 fold or greater increase in activity.

[0109] A plurality of clonal neo-antigens may refer to a plurality of peptides or proteins each comprising a different tumor-specific mutation giving rise to the neo-antigen. The plurality may be 2-250, 3-200, 4-150, or 5-100 tumor-specific mutations, e.g., 5-75 or 10-50 tumor-specific mutations. Each tumor-specific mutation may be represented by one or more clonal neo-antigen peptides. In other words, a plurality of clonal neo-antigens may comprise a plurality of different peptides, some of which comprise sequences that contain the same tumor-specific mutation (e.g., at different positions within the sequence of the peptide, or within peptides of various lengths).

[0110] T cell populations generated according to the present disclosure will exhibit an increased number or proportion of T cells targeting one or more neo-antigens predicted to be clonal. That is, the composition of the T cell population will differ from a "native" T cell population (i.e., a population that has not been subjected to the expansion step discussed herein) in that it will have an increased proportion or proportion of T cells targeting neo-antigens predicted to be clonal. T cell populations according to the present disclosure may have at least about 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, or 100% T cells targeting neo-antigens predicted to be clonal.

[0111] The immunotherapeutics described herein may be used to treat cancer. Accordingly, the present disclosure also provides a method of treating cancer in a subject, comprising administering to the subject an immunotherapeutic composition as described herein.

[0112] In addition, the presence of clonal neo-antigens has been shown to be associated with improved cancer prognosis. Accordingly, also described herein is a method of providing a prognosis to a subject diagnosed with cancer, comprising determining the proportion and / or number of clonal neo-antigens in one or more tumor samples from the subject.

[0113] Suitably, in any embodiment of any aspect described herein, the cancer may be ovarian cancer, breast cancer, endometrial cancer, renal cancer (renal cell), lung cancer (small cell, non-small cell and mesothelioma), brain cancer (glioma, astrocytoma, glioblastoma), melanoma, Merkel cell carcinoma, clear cell renal cell carcinoma (ccRCC), lymphoma, small intestine cancer (duodenal and jejunal cancer), leukemia, pancreatic cancer, hepatobiliary tumors, germ cell carcinoma, prostate cancer, head and neck cancer, thyroid cancer and sarcoma. For example, the cancer may be lung cancer, such as lung adenocarcinoma or lung squamous cell carcinoma. As another example, the cancer may be melanoma. In an embodiment, the cancer may be selected from melanoma, Merkel cell carcinoma, renal cancer, non-small cell lung cancer (NSCLC), bladder urothelial carcinoma (BLAC) and head and neck squamous cell carcinoma (HNSC) and microsatellite instability (MSI) high cancer. In some embodiments, the cancer is non-small cell lung cancer (NSCLC). In other embodiments, the cancer is melanoma.

[0114] Treatment using the compositions and methods of the present disclosure may also include targeting circulating tumor cells and / or metastases derived from the tumor. Treatment according to the present disclosure targeting one or more neo-antigens may help prevent the evolution of anti-therapeutic tumor cells that may occur with standard approaches such as chemotherapy, radiation therapy, or non-specific immunotherapy. The methods and uses for targeting cancer described herein may be performed in combination with additional cancer therapies. In particular, the T cell compositions described herein may be administered in combination with immune checkpoint intervention, costimulatory antibodies, chemotherapy and / or radiation therapy, targeted therapy, or monoclonal antibody therapy. "In combination" may refer to administration of the additional therapy before, simultaneously, or after administration of the T cell composition as described herein.

[0115] The disclosure also provides a method of making an immunotherapeutic composition, the method comprising identifying a neo-antigen as potentially clonal, and making an immunotherapeutic composition that targets the neo-antigen.

[0116] Also described herein is a method of treating a subject diagnosed with cancer, comprising: identifying a plurality of tumor-specific mutations in the subject; determining whether one or more of the tumor-specific mutations are likely to be clonal in the subject; selecting one or more of the tumor-specific mutations as candidate clonal neo-antigens, the candidate clonal neo-antigens being tumor-specific mutations that meet at least one or more predetermined criteria as to whether the tumor-specific mutation is likely to be clonal; and identifying the one or more clonal neo-antigens by treating the subject with an immunotherapy that targets one or more of the selected candidate clonal neo-antigens; wherein the determination of whether the tumor-specific mutation in the subject is likely to be clonal is performed using the methods described herein. In particular, the determination of whether the tumor-specific mutation in the subject is likely to be clonal includes obtaining, by a processor, sequence data from one or more samples from the subject that contain tumor genetic material, the sequence data comprising, for each of the one or more samples, a number of reads that indicate the tumor-specific mutation in the sample (d b ), the number of reads in the sample that represent the corresponding germline allele, and the total number of reads at the position of the tumor-specific mutation (d), and determining, by the processor, a likelihood that the tumor-specific mutation is clonal as a posterior probability that depends on the prior probability that the mutation is clonal and the probability that the sequencing data is observed when the tumor-specific mutation is (i) clonal, and (ii) non-clonal, given the tumor fraction for each of the one or more samples, and one or more candidate binding genotypes, each of which includes a genotype at the position of the tumor-specific mutation, for a normal population, a reference tumor population that does not contain the tumor-specific mutation, and a mutant tumor cell population that includes the tumor-specific mutation.

[0117] The clonal neo-antigen candidates may be selected as tumor-specific mutations that further satisfy at least one or more predetermined criteria as to whether the tumor-specific mutations are likely to give rise to neo-antigens. The step of selecting one or more of the tumor-specific mutations as clonal neo-antigen candidates by the processor may include determining whether the one or more tumor-specific mutations satisfy one or more criteria as to whether the tumor-specific mutations are likely to give rise to neo-antigens selected from the following: the mutation is associated with an expression product expressed in tumor cells, the mutation is predicted to result in a protein or peptide that is not expressed in normal cells of the subject, the mutation is predicted to result in at least one peptide that may be presented by an MHC molecule, the mutation is predicted to result in at least one peptide that may be presented by an MHC allele known to be present in the subject, and the mutation is predicted to result in an immunogenic protein or peptide. The step of selecting, by the processor, one or more of the tumor-specific mutations as clonal neo-antigen candidates may include determining, by the processor, whether one or more tumor-specific mutations satisfy one or more predefined criteria for whether a tumor-specific mutation is likely to be clonal selected from: the mutation having a likelihood of being clonal above a predefined threshold; the mutation having a likelihood of being clonal above a threshold adaptively set to select a predefined number of tumor-specific mutations with the highest likelihood of being clonal among the tumor-specific mutations for which the likelihood was determined; and the mutation having a likelihood of being clonal above a threshold adaptively set to select a predefined top percentile of tumor-specific mutations among the tumor-specific mutations for which the likelihood was determined.

[0118] The immunotherapy targeting one or more of the selected clonal neo-antigens may be an immunogenic composition, a composition comprising immune cells, or a therapeutic antibody. The immunotherapy may be a composition comprising T cells that recognize at least one of the one or more of the identified selected clonal neo-antigens. The composition may be enriched for T cells that target at least one of the one or more of the identified selected clonal neo-antigens. The method may include obtaining a population of T cells and expanding the T cell population to increase the number or relative proportion of T cells that target at least one of the one or more of the identified selected clonal neo-antigens.

[0119] system FIG. 3 illustrates an embodiment of a system for determining whether a tumor-specific mutation is likely to be clonal and / or for identifying a clonal neo-antigen and / or for providing a prognosis or for providing an immunotherapy based at least in part on the identified clonal neo-antigen according to the present disclosure. The system includes a computing device 1, which includes a processor 101 and a computer-readable memory 102. In the illustrated embodiment, the computing device 1 also includes a user interface 103, which is illustrated as a screen, but may include any other means of conveying information to a user, such as by an audible or visual signal. The computing device 1 is communicatively connected, such as via a network 6, to a sequence data collection means 3, such as a sequencing machine, and / or one or more databases 2 that store sequence data. The one or more databases may additionally store other types of information that the computing device 1 may use, such as, for example, reference sequences, population parameters, etc. The computing device may be a smartphone, tablet, personal computer, or other computing device. The computing device is configured to implement a method for determining whether a tumor-specific mutation is likely to be clonal as described herein. In an alternative embodiment, the computing device 1 is configured to communicate with a remote computing device (not shown), which is itself configured to implement the method of determining whether a tumor-specific mutation is likely to be clonal, as described herein. In such a case, the remote computing device may also be configured to transmit the results of the method to the computing device. The communication between the computing device 1 and the remote computing device may be through a wired or wireless connection, and may take place over a local or public network, such as via the public Internet or via Wi-Fi®.

[0120] The sequence data collection 3 means may be wired to the computing device 1 or may be capable of communicating through a wireless connection, such as through a network 6, as shown. The connection between the computing device 1 and the sequence data collection means 3 may be direct or indirect (such as through a remote computer). The sequence data collection means 3 is configured to obtain sequence data from a nucleic acid sample, such as a genomic DNA sample, extracted from a cell and / or tissue sample. In some embodiments, the sample may have been subjected to one or more pre-processing steps, such as DNA purification, fragmentation, library preparation, target sequence capture (such as exon capture and / or panel sequence capture). Preferably, the sample has been subjected to amplification, or when it has been subjected to amplification, it has been performed in the presence of an amplification bias control means, such as a unique molecular identifier. Within the context of the present disclosure, any sample preparation process (whether whole genome or sequence specific) suitable for use in determining genome copy number profiles may be used. The sequence data collection means 3 is preferably a next generation sequencer. The sequence data collection means 3 may be directly or indirectly connected to one or more databases 2 in which sequence data (raw or partially processed) may be stored.

[0121] The following are presented as examples and should not be construed as limiting the scope of the claims. EXAMPLES

[0122] Working Example These examples describe the method for identifying clonal mutations according to the present disclosure and demonstrate its use using simulated data and multiple types of experimental data.

[0123] method Mutation genotype model The data for this model are allele counts from N mutations (n=1,...N) from S samples (s=1,...,S). For simplicity, and because the method can analyze a single sample and mutation, the notation used in this section will not explicitly include the subscripts n for mutation and s for sample. As illustrated in Figure 4, this model assumes that the set of sequenced cells is separated by each mutation into three subpopulations: (i) a normal cell population consisting of cells with a healthy germline genome (which may be diploid in the region of the mutation); (ii) a reference cell population consisting of cancer cells without the mutation in question (which may be aneuploid in the region of the mutation in question); and (iii) a mutant cell population consisting of cancer cells with the mutation in question (which may be aneuploid in the region of the mutation in question, which may not have the same copy number in said region as the reference population). The term "mutation" is intended in its broad sense herein to refer to any genetic alteration detectable in sequence data, particularly genomic sequence data. This specifically includes single nucleotide variants (SNVs), multiple nucleotide variants (MNVs), indels, and the like.

[0124] Let G = (A, B, AA, AB, AAA, AABB, ...) be the set of all genotypes (where A and B represent the reference and mutant alleles, respectively). For example, AB would represent a heterozygous mutant (containing one reference / normal allele A and one mutant allele B) with a total copy number of 2. Based on this notation, in FIG. 4, the normal population has genotype AA (where both As can be the same or different, i.e., the normal population can be homozygous or heterozygous, but both alleles are normal), the reference population has genotype AAA (where the A allele is selected from the A alleles of the normal population), and the mutant population has genotype AABB (where the A allele is selected from the A alleles of the normal population, and the B allele is any non-reference allele). We assume that the genotype of all cells in each subpopulation is constant (i.e., with reference to FIG. 4, all cells in the normal population have genotype AA, all cells in the reference population have genotype AAA, and all cells in the mutant population have genotype AABB). G=(G H ;G R ;G V )∈G 3 Let be vectors whose components are the genotypes of the normal (healthy), reference and mutant populations, respectively (each of these individual genotypes will be referred to collectively below as "G"). Let t be the proportion of cancer cells in the sample. This is often referred to as the tumor content, tumor purity or cellularity of the sample. Let φ be the proportion of cancer cells in the sample that harbor the mutation, i.e., the relative proportion of cancer cells in the mutant population. This is often referred to as the cancer cell fraction (CCF) or cellular prevalence of the mutation. Let ε be the assumed sequencing error rate. The following function is defined:

number

number

number

number

[0125] Let ξ(G,φ,t) be the probability of extracting a read containing a mutant allele. Assuming that the initial population of cells extracted during sequencing is infinite, the probability of extracting a read containing a mutant allele is roughly proportional to the number of copies of the mutant allele in the input pool of DNA. More formally, taking into account sequencing errors, the probability of extracting a mutant allele (given a set of genotypes G, tumor content t, and cancer cell fraction φ) is given by the following formula (Equation (1)):

number

[0126] The variable ξ(G,φ,t) captures the sum of the copy numbers of the variant allele originating from each genotype, multiplied by the probability of extracting a read containing the mutation from that genotype, normalized by the sum of the total copy numbers of both alleles originating from each genotype.

[0127] The variable d is the total number of reads covering the mutation in the sample, of which d b Thus, the number of reads d, d b The probability of observing (P(d,d b |G,φ,t)) is the parameter d band ξ(G,φ,t) (Equation (3)). This means that the sum of m Bernoulli random variables with parameter p is 2 Instead, for example, if the data has more variance than can be explained by the binomial model, a beta-binomial model (equation (4)) with mean ξ(G,φ,t) and precision (the inverse of the variance) γ: P(d,d b |G,φ,t)=binomial(d b |d,ξ(G,φ,t)) (3) P(d,d b |G,φ,t,γ)=beta binomial(d b |d,ξ(G,φ,t),γ) (4) can be used.

[0128] The parameter γ is set to 200 in the following example, but other values ​​are possible. So far, we assume that the genotypes of the subpopulations are known. In general, this may be true for the healthy population (e.g., from matched germline samples), but not for the reference and mutant populations. Instead, for regions overlapping the mutation, allele-specific copy number estimates are typically observed. This information can be used to derive priors for the set of plausible genotypes. We explain how to do this in the next section. For now, let us assume that we have prior probabilities for a vector π, where π i is the i-th plausible combined genotype of the population, G i The probability of the observed data being marginal for all plausible genotypes is given by (Eq. (3a) and (4a)): P(d,d b |π,φ,t)=Σ i π i Binomial (d b |d,ξ(G,φ,t)) (3a) P(d,d b |π,φ,t,γ)=Σ i π i Beta binomial (d b |d,ξ(G,φ,t),γ) (4a) can be written as follows.

[0129] In later sections, we use the notation Pr(d,d b |π,φ,t) will be used to refer equally to the expressions in Eq. (3a) and Eq. (4a). Since φ and t relate to individual samples, the above notation can be used to denote φ, t, t, t, respectively. s and t s Note that this is an abbreviation for

[0130] Derive mutant genotype priors The above model uses either known joint genotypes, or a prior probability π, where π i is the i-th plausible combined genotype of the population, G i is the prior probability of (i.e., G i (where x is one possible combination of genotypes for the healthy, mutant and reference populations). Various methods can be used to set the likely genotype priors.

[0131] For example, one possible method can be referred to as the "major copy number" method. major and c minor means the major and minor allele copy numbers for the region overlapping the mutation in the tumor sample. In this method, the "major copy number method", two cases are considered: (a) In the first case, the mutation occurs before the copy number event. In this case, the reference population genotype matches the normal population. major Consider all possible mutant genotypes for the mutant population up to a chromosome carrying a mutant of (b) In the second case, the mutation occurs after a copy number event. In this case, the reference population is minor +c minor The mutant population has a mutant allele of 1 and a reference allele of c major +c minor Has a reference allele of -1.

[0132] We set the prior weights to be equal for all possible mutant genotypes. For example, c minor =2 and c minor = 1 and normal copy number is 2. Possible genotypes are: G1=(AA, AA, AAB) G2=(AA, AA, ABB) G3=(AA, AAA, AAB) , each with a prior probability of 1 / 3. If allele-specific copy numbers are not available, then c major Set to the total number of copies, and c minor Note that one can set α to 0. This approach assumes that the mutation occurs only once, so if there are more than one copy of the mutant allele in a mutant population, this occurred because the mutation preceded a copy number change at the locus and was subsequently amplified. This approach provides a good compromise between accounting for uncertainty in the genotypes of a population, while not considering too many states.

[0133] Alternative approaches may be used to set the mutant genotype priors. Another possible approach is to simply assume that each mutation is diploid and heterozygous (i.e., a mutant in a mutant population appears on only one of the two chromosomes, G = (G H =AA,G R =AA,G V =AB). This can be referred to as the "AB prior". Yet another simplistic approach is to assume that each mutation is diploid and homozygous (i.e., the mutations in the mutant population appear on both chromosomes, G = (G H =AA,G R =AA,G V =BB). This can be referred to as the "BB prior". Yet another possible simple approach is to assume that the genotypes of the mutant population have the expected total copy number in the region of the mutation, with exactly one being the mutant allele (i.e., assume the total copy number is 3, G = (G H =AA,G R =AA,G V=AAB), i.e., this results in only considering G1 in the "major copy number" method described above. This can be referred to as a "no zygosity prior." These approaches can often be oversimplified, since they essentially consider a single possible genotype.

[0134] Another possible approach is to assume that the genotypes of the mutant population have the expected total copy number at the region of the mutation, at least one of which is the mutant allele, and that the reference population is either AA or a genotype with a copy number equal to the expected total copy number, and no mutant allele (equally probable). This can be called a "total copy number prior", and intuitively means that the genotypes of the mutant population at the locus have the expected total copy number and can have any number of copies (>0) of the mutant allele (i.e., if we assume the total copy number is 3, then the possible genotypes, with equal probability, are G = (G H =AA,G R =AA,G V =AAB), G2=(G H =AA,G R =AA,G V =ABB), G3=(G H =AA,G R =AA,G V =BBB), G4=(G H =AA,G R =AAA, G V= AAB), i.e. it essentially ignores the major and minor copy numbers and considers all possible genotypes with n copies - leading to additional genotypes being considered compared to the "major copy number" method above). Yet another approach that can be used is to "trust" the major and minor allele numbers predicted from the copy number caller, and thus only consider genotypes with a number of mutant alleles corresponding to either the major or minor copy number. This can be referred to as the "parent" mode. For example, if major copy number = 3 and minor copy number = 1, this approach would consider the following possible genotypes with equal probability: G1 = (AA, AA, AAAB), G2 = (AA, AA, ABBB), G3 = (AA, AAAA, AAAB) (i.e. either 1 or 3 mutant alleles in the mutant population). In contrast, the "major copy number" approach "trusts" the range of possible major copies by considering all values ​​between 1 and the predicted major copy number, but not its absolute value. In the above example, major copy number = 3, minor copy number = 1, this may lead to considering one more genotype compared to the "parental" mode, i.e. G1 = (AA, AA, AAAB), G2 = (AA, AA, AABB), G3 = (AA, AA, ABBB), G4 = (AA, AAAA, AAAB). Thus, the "major copy number" approach strikes a good balance between accounting for additional uncertainty from the copy number calls (compared to the "parental" approach) without overthinking the uncertainty (compared to the "total copy number" approach).

[0135] Clonality estimation model This section outlines a hierarchical Bayesian model for identifying ubiquitous mutations. Let Z be a Bernoulli variable that is 1 when the mutation is ubiquitous (assumed to be clonal) and 0 otherwise. Let ρ be the prior probability that the mutation is ubiquitous, which is set to 0.5 in the following examples. As above, φ is the proportion of cancer cells in the sample that harbor the mutation. The model is therefore Z|ρ~Bernoulli(Z|ρ) (5) φ|Z~beta (φ|α=1,β=1), for Z=0; For beta(φ|α,β=1), Z=1 (6) d b ,d|π,φ,t~Pr(d,d b |π,φ,t) (7) (where α is the parameter >1 in the distribution for φ|Z=1), which is set to α=99 in the following examples. A beta distribution with parameters α=99 and β=1 is skewed towards 1, capturing the assumption that clonal mutations should be enriched for higher cancer cell fractions φ. Other values ​​of the parameters are possible, but values ​​that capture this assumption are preferred. As mentioned above, the probability in Equation (7) is given by Equation (3) / Equation (3a) or Equation (4) / Equation (4a).

[0136] The joint distribution can be expressed as follows (Equation (8)): For one sample, p(d b ,d,φ,Z=z|π,t,ρ)=p(Z=z|ρ)Pr(d b ,d|π,φ,t)p(φ|Z=z) (8) Or for multiple samples:

number

[0137] The proportion of cancer cells harboring the mutation (φ) is unknown. However, for a given sample,

number

number

[0138] amount

number

number

[0139] Finally, the quantity we wish to estimate is the observed lead (d b ,d), the genotype prior (π), the tumor fraction estimate (t), and the prior probability that the mutation is clonal (ρ, i.e., p(Z=1|d b ,d,π,t,ρ) we try to estimate. Taking the above into account, this can be expressed as:

number

number

[0140] Thus, estimating equation (11) for z=1 (i.e., equation (11a)) gives the probability that the mutation is ubiquitous (i.e., assumed to be clonal when considering one or more available samples). This requires the evaluation of S one-dimensional integrals (one for each sample, in equations (9), (10)), which can be done efficiently using known numerical integration. Any numerical integration algorithm known in the art can be used for this purpose. For example, a lattice approximation may be used. This is advantageously simple, given that there is a single parameter (φ) to integrate over.

[0141] It provides an estimate of the probability that a mutation is clonal given the available data; it can be computed efficiently; it is easily interpretable (given a rigorous statistical model that utilizes explicit, well-defined assumptions); it can be determined for any mutation without manual input; it is independent of any other mutations analyzed; it can rigorously include prior knowledge about the mutation; and it can be used to objectively and automatically prioritize a list of mutations (including their associated probabilities) for testing and / or use.

[0142] Accounting for uncertainty in copy number predictions The models described above already provide numerous advantages, but they can be further enhanced by taking into account the uncertainty in the prediction of the copy number estimates used in the models. Indeed, the models described above assume that the copy numbers (e.g., major / minor / total / copy numbers used to derive genotype priors) are accurately predicted. In reality, there may be some uncertainty in those values. Indeed, the problem of allele-specific copy number analysis of tumors is complex, and many solutions have been proposed for it. One commonly used approach is ASCAT (allele-specific copy number analysis of tumors, Van Loo et al., 2010), which takes into account both tumor cell aneuploidy and non-abnormal cell infiltration in interpreting bulk copy number profiles, and outputs estimated allele-specific copy number profiles and associated tumor purity estimates. In brief, ASCAT evaluates multiple possible combinations of tumor ploidy and tumor fraction based on the assumption that the relevant allele-specific copy number calls should be as close as possible to non-negative integers for germline heterozygous single nucleotide polymorphisms (SNPs). The solution deemed optimal (estimated tumor ploidy, tumor purity and allele-specific copy number calls for the tumor and normal portions of the sample) is then reported along with its goodness of fit (based on the assumptions above).

[0143] The model provided above can be adapted and adjusted for multiple copy number solutions and their uncertainties by amending π to include components for genotypes from each predicted copy number state (e.g., each proposed solution including major and minor copy states) and weighted by the probability associated with that state. In addition, since tumor purity estimates can be estimated together with their copy number states (e.g., as is the case when an approach such as ASCAT is used), the associated tumor purity estimates can also be taken into account. Note that this may not be necessary, for example, when tumor purity is estimated or measured separately and is not inherently related to the copy number state estimate. Nevertheless, for generalization, a set of C possible copy number / tumor content states (e.g., c major , c minor , and C possible sets of estimates for t). C Let be a vector whose components are the probabilities for each such possible set of estimates. For each state C, let be a vector of possible genotypes as described above, π CG It is possible to calculate π CG to C The final genotype vector can be obtained by multiplying the components for state C of P(d,d b |π,φ)=Σ i π i Binomial (d b |d,ξ(G i ,φ,t i )) (3b) P(d,d b |π,φ,γ)=Σ i π i Beta binomial (d b |d,ξ(G i ,φ,t i ),γ) (4b) Here, the tumor content t i now depends on the particular state (and π i is π CGThese new densities can be substituted into the relevant equations above. In particular, the problem to be solved may then be expressed as solution equation (11a), where Pr(d b ,d|π,φ,t) is given by equation (3b) or equation (4b). i , c major , c minor The value of (and hence the appropriate π according to the model used) CG ) and π C is provided as an output of many methods of performing allele-specific copy number analysis of tumors, including but not limited to ASCAT, as described above. For the avoidance of any doubt, any technique that generates an allele-specific copy number state estimate (typically associated with a tumor purity estimate) along with a confidence or other metric that can be used to weight multiple solutions relative to one another can be used for this purpose.

[0144] implementation The method described herein can be implemented using any programming language known in the art. In the following example, a Python script was used to implement the above method. It took as input, for each mutation, the mutation identifier, the sample identifier, the number of reads that match the reference allele at the mutation position, the number of reads that match the other allele at the mutation position, and for each of one or more copy number solutions: the major copy number (for tumor) that overlaps the mutation for the specified copy number solution, the minor copy number (for tumor) that overlaps the mutation for the specified copy number solution, the copy number of normal cells at the mutation (may be set to default=2 for autosomes, or 1 for sex chromosomes of male subjects), and the tumor purity value for the specified copy number solution (which can also be obtained as the output of ASCAT, for example, or can be obtained separately). For a given copy number solution, the major and minor copy numbers overlapping the mutation for the tumor population can be determined directly from ASCAT (e.g., using ascatNgs, Raine et al., 2016) or can be derived from the output of ASCAT, for example, using the average B allele frequency of the copy number segments overlapping the mutation, the log R value of the copy number segments overlapping the mutation, and the ploidy of the solution. For example, the allele-specific copy number estimate at position i for the tumor

number

number

[0145] In the following examples, whenever copy number solutions were estimated, they were done using ASCAT (Van Loo et al., 2010).

[0146] result Synthetic Data The approach described above was first tested using synthetic data simulated using the population model described above by reference to FIG. 4 as captured by equation (3) (binomial model) and equation (4) (beta-binomial model). A total of 1000 mutations were simulated for all synthetic datasets. Coverage depth was simulated from a Poisson distribution with mean values ​​of 50, 100, and 1000, corresponding to realistic coverage regimes. Copy number profiles were randomly simulated with a maximum total copy of 2 or 6. These settings correspond to a genomically stable genome (maximum total copy number of 2) and a highly aneuploid genome (maximum total copy number of 6). Datasets containing one, two, or four samples were simulated. In all cases, tumor content was fixed at 1.0.

[0147] Figure 5A shows the results when the data was simulated from a binomial model and the method described above was run using the binomial model. For the copy number 2 runs, performance is nearly perfect regardless of depth or number of samples (area under the curve, AUC = 0.997-1.0, see Table 1 below). For the copy number 6 runs, performance improves considerably with increasing depth and number of samples (AUC = 0.853-1.0, see Table 1 below).

[0148] [Table 1]

[0149] Figure 5B shows the same analysis done by setting the accuracy to 100 and simulating from a beta-binomial distribution using a binomial emission model for the model run above. This simulates the case where the observed data is more variable than expected. The results are still nearly perfect for the simulation of copy number 2 (AUC 0.997-1.0, Table 2). However, for the simulation of copy number 6, the results (although still good) are not as good as the previous analysis (AUC 0.846-0.996, Table 2). Increasing the coverage depth does not improve the performance as much as before. In contrast, increasing the number of samples has a substantial effect.

[0150] Figure 5C shows the results when data is simulated using a beta binomial distribution with precision 100 and analyzed by the above method using a beta binomial emission model with precision 200. Performance is similar to the previous analysis using the above method with binomial emission (Table 2). Using the beta binomial model for excess variation data shows a slightly better performance, especially at low sequencing depth and / or high copy number, compared to using a binomial model for the same excess variation data. However, in this example, the performance improvement is not significant, and therefore it would be advantageous to use a simpler binomial model in each case (which is simpler to parameterize because it does not need to set high-precision hyperparameters).

[0151] [Table 2]

[0152] Cell line mixture data We next tested this method using data generated by mixing cell lines (Farahani et al., 2017). Data were obtained from two experiments, where 1) diploid cell lines were mixed, and 2) aneuploid cell lines were mixed. Multiple mixtures were formed in various ratios as listed in Table 3. In each experiment, there are three classes of mutations: 1) mutations unique to cell line A; 2) mutations unique to cell line B, and 3) mutations shared by both cell lines. For the purposes of testing the method described herein, shared mutations were treated as ubiquitous mutations and cell line specific mutations were treated as subclonal mutations. Allele count data and copy number information were extracted from Supplementary Tables S1A and S1B of Farahani et al., downloaded from the publisher's website. Note that it was not possible to account for copy number uncertainty with this data, as only predicted major / minor copies are given for each mutation.

[0153] The method described above, which is a run with a binomial emission model, was compared to a PyClone (Roth et al., 2014) version 0.13.1 run for 10,000 iterations, with 1,000 iterations discarded as burnout. To identify ubiquitous mutations with PyClone, we first identified the cluster with the highest CCF across all samples. All mutations in this cluster were treated as ubiquitous and all other mutations were treated as subclonal. Note that this strategy does not assign a confidence value to predict ubiquity, and therefore PyClone results are plotted as a single point on a ROC curve. In contrast, the method described herein assigns a probability of ubiquity to each mutation, which allows the curve to be plotted. Thus, the threshold used to consider mutations as clonal / subclonal can also be adjusted depending on the desired balance of precision and recall.

[0154] [Table 3]

[0155] FIG. 6A shows the results for a diploid cell line mixture. The plot on the left of FIG. 6A is from a run using only a single sample. They were worst using mixture S3, which was 90% cell line A and 10% cell line B, for both methods. This makes sense, since mutations from cell line A would have a CCF close to the shared mutations. As the mixture becomes more balanced and the CCF of mutations in the cell lines moves away from 1, performance improves. The right side of FIG. 6A shows the results from a run using multiple samples. FIG. 6B shows the results from a mixture of aneuploid cell lines. These results are generally consistent with the diploid case, but as expected, performance is uniformly worse for both methods.

[0156] This data indicates that, as expected, adding more samples may improve performance. In the particular setup of simulation data and artificial samples analyzed here, the performance of the method described herein and PyClone is similar for some samples, but in some multi-sample examples, the method described herein outperforms PyClone.

[0157] Informal timing tests of each method revealed that the methods described herein took approximately 10 seconds per analysis (wherein the analysis involves determining the likelihood of clonality for all mutations identified in the patient). PyClone took significantly longer for the same analysis, on the order of 300-3000 seconds depending on the number of MCMC iterations performed. Thus, even when all mutations identified in a patient are evaluated in the same analysis (which is not a requirement of the method since these evaluations are completely independent), the method is at least 30 times faster than the prior art methods. This difference can be even greater if the evaluation of mutations is performed in parallel rather than sequentially (which is not possible using the prior art methods).

[0158] Thus, this data demonstrates that the methods described herein are at least comparable to and superior to comparable state-of-the-art methods in the case of multiple samples. Collecting multiple samples when possible may increase the likelihood of identifying truly clonal variants when using methods that can take advantage of it, such as those described herein, and thus has great clinical and practical implications. In contrast to the methods described herein, PyClone does not have the ability to provide mutation-specific predictions, provide confidence estimates for each such prediction, or take into account the uncertainty of the copy number solution, and requires manual input and much longer computation time to analyze the same data.

[0159] Thus, the method described herein has some advantages that are particularly meaningful in practical clinical contexts, in addition to special cases (such as the simulation and artificial samples described above) where the prior art methods may be used with similar performance. For example, the need for extensive manual input for curation and selection of at least a single copy number solution means that in the absence of such expert input, the method described herein will perform much better than the prior art methods, even in a single sample situation. Moreover, even if expert manual input is available, in a real situation (i.e., using real tumor samples from patients), where identifying the correct copy number solution is much more error-prone than the artificial controlled situation studied above, the method described herein will perform much better than the prior art methods.

[0160] TRACERx data The methods described herein were further tested on data from the CRUK TRACERx study (http: / / tracerx.co.uk / ). This data had previously been analyzed using the TRACERx pipeline (see McGranahan et al., 2016), and clonal calls based on the PyClone method were available. In addition, ASCAT ploidy / purity solution data were available for all samples. A summary of the data is shown in Figure 7A (based on available clonal calls obtained using PyClone). Thus, it was possible to perform the method taking into account uncertainties regarding the ploidy / purity solutions of choice used to derive copy numbers.

[0161] FIG. 7B shows the distribution of predicted ubiquity probabilities from the method described herein when 1) all ASCAT purity / ploidy solutions are considered, and 2) the maximum possible ASCAT solution is considered. It is observed that in our method, higher probabilities were systematically assigned to ubiquitous mutations when purity / ploidy uncertainty was considered. FIG. 7C shows the ROC curve comparing the two approaches. When considering ploidy and purity uncertainty, the performance of our method improved significantly, from 0.76 to 0.91 AUC. One caveat of this analysis is that the curated PyClone results were assumed to be true (as opposed to the simulation data and artificial samples analyzed above, ground truth cannot be obtained). The previous section suggests that the method described herein outperforms PyClone for real tumor samples and multiple sample situations, so the actual performance of the method described herein may be better than reported here.

[0162] Runtime comparison with prior art methods The method described herein was officially timed side-by-side with the latest version of PyClone (PyClone-VI, Gills & Roth, BMC Bioinformatics volume 21, Article number: 571 (2020)). The results of this analysis are shown in Figure 8 (where "ACE" refers to the method as described herein). This version of PyClone has been significantly optimized and is therefore faster than the original version used in the informal timing test above. However, it necessarily analyzes all tumor-specific mutations identified in a cancer together, which results in a significant increase in run time as the number of mutations increases, which cannot be countered by parallelization. In contrast, the method disclosed herein analyzes each mutation independently and is therefore fully parallelizable. As a result, the run time associated with the method is significantly shorter in both cases, and here the analysis for a single cancer can be run on multiple CPUs (see Figure 8, which shows results for 2, 3 or 4 CPUs). As an example, for a single patient with 200-1000 mutations (which is a clinically realistic range), the time to identify the probability of clonality for each mutation using the present method can be as short as 12.8-36.5 seconds (for 200 and 1000 mutations, respectively). In contrast, the time to run PyClone-VI is 16.8-99.1 seconds (for 200 and 1000 mutations, respectively). In addition, this time estimate does not include the time to manually analyze the cluster structure produced by PyClone to select clusters of mutations assumed to be clonal, and to prioritize mutations within the selected clusters. In contrast, the method described herein produces a directly interpretable independent probability that each mutation is clonal, which can be used directly (i.e., without a further evaluation step for clonality) to prioritize mutations. This represents a significant labor saving as well as an increase in the tractability of the process.

[0163] Consideration These examples demonstrate the development and evaluation of a novel model for identifying mutations that are ubiquitously present throughout a patient's cancer cell population, i.e., clonal mutations. There are several innovations in this work that significantly improve performance compared to previous solutions. In particular, we rewrite the problem as a classification problem and provide probabilities for predictions. This provides the ability to rank mutations instead of providing a method that returns a binary answer. Furthermore, we show how the uncertainty regarding the number of overlapping copies can be taken into account. In real-world data, this leads to a significant increase in performance in terms of the accuracy of identifying clonal mutations. In practice, this means that clonal mutations identified by the methods described herein are much more likely to be truly clonal than mutations identified using prior art methods, leading to higher validation rates and ultimately higher clinical success rates. Other benefits of the methods described herein include the ability to add new mutations without changing the scores of other mutations. Furthermore, the method is orders of magnitude faster than PyClone and other similar clustering-based methods.

[0164] References Adalsteinsson, VA, Ha, G., Freeman, SS et al. Scalable whole-exome sequencing of cell-free DNA reveals high concordance with metastatic tumors. Nat Commun 8, 1324 (2017). Bulik-Sullivan B, Busby J, Palmer CD, Davis MJ, Murphy T, Clark A, Busby M, Duke F, Yang A, Young L, Ojo NC, Caldwell K, Abhyankar J, Boucher T, Hart MG, Makarov V, Montpreville VT, Mercier O, Chan TA, Scagliotti G, Bironzo P, Novello S, Karachaliou N, Rosell R, Anderson I, Gabrail N, Hrom J, Limvarapuss C, Choquette K, Spira A, Rousseau R, Voong C, Rizvi NA, Fadel E, Frattini M, Jooss K, Skoberne M, Francis J, Yelensky R. Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification. Nat Biotechnol. 2018 Dec 17. Carter SL, Cibulskis K, Helman E, McKenna A, Shen H, Zack T, Laird PW, Onofrio RC, Winckler W, Weir BA, Beroukhim R, Pellman D, Levine DA, Lander ES, Meyerson M, Getz G. Absolute quantification of somatic DNA alterations in human cancer. Nat Biotechnol. 2012 May;30(5):413-21. Hossein Farahani, Camila PE de Souza, Raewyn Billings, Damian Yap, Karey Shumansky, Adrian Wan, Daniel Lai, Anne-Marie Mes-Masson, Samuel Aparicio, and Sohrab P Shah. Engineered in-vitro cell line mixtures and robust evaluation of computational methods for clonal decomposition and longitudinal dynamics in cancer. Scientific Reports, 7(1): 13467, 2017. Vanessa Jurtz, Sinu Paul, Massimo Andreatta, Paolo Marcatili, Bjoern Peters and Morten Nielsen. NetMHCpan-4.0: Improved Peptide-MHC Class I Interaction Predictions Integrating Eluted Ligand and Peptide Binding Affinity Data. J Immunol November 1, 2017, 199 (9) 3360-3368. Langmead, B., Trapnell, C., Pop, M. et al. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol 10, R25 (2009). Lundegaard C, Lamberth K, Harndahl M, Buus S, Lund O, Nielsen M. NetMHC-3.0: accurate web accessible predictions of human, mouse and monkey MHC class I affinities for peptides of length 8-11. Nucleic Acids Res. 2008 Jul 1;36(Web Server issue):W509-12. Nicholas McGranahan, Francesco Favero, Elza C de Bruin, Nicolai Juul Birkbak, Zoltan Szallasi, and Charles Swanton. Clonal status of actionable driver events and the timing of mutational processes in cancer evolution. Science translational medicine, 7(283):283ra54- 283ra54, 2015. McGranahan, N., Furness, A. J., Rosenthal, R., Ramskov, S., Lyngaa, R., Saini, S. K., Jamal- Hanjani, M., Wilson, G. A., Birkbak, N. J., Hiley, C. T., Watkins, T. B., Shafi, S., Murugaesu, N., Mitter, R., Akarca, A. U., Linares, J., Marafioti, T., Henry, J. Y., Van Allen, E. M., Miao, D., ... Swanton, C. (2016). Clonal neoantigens elicit T cell immunoreactivity and sensitivity to immune checkpoint blockade. Science (New York, N.Y.), 351(6280), 1463-1469. Timothy J. O’Donnell, Alex Rubinsteyn, Maria Bonsack, Angelika B. Riemer, Uri Laserson, Jeff Hammerbacher. MHCflurry: Open-Source Class I MHC Binding Affinity Prediction. Cell Systems Vol. 7, Issue 1, 129-132, July 25, 2018. Andrew Roth, Jaswinder Khattra, Damian Yap, Adrian Wan, Emma Laks, Justina Biele, Gavin Ha, Samuel Aparicio, Alexandre Bouchard-Cote, and Sohrab P Shah. PyClone: statistical inference of clonal population structure in cancer. Nature methods, 11(4):396, 2014. Russell Schwartz and Alexander A Schaffer. The evolution of tumor phylogenetics: principles and practice. Nature Reviews Genetics, 18(4):213, 2017. Van Loo P, Nordgard SH, Lingjaerde OC, Russnes HG, Rye IH, Sun W, Weigman VJ, Marynen P, Zetterberg A, Naume B, Perou CM, Borresen-Dale AL, Kristensen VN. Allele-specific copy number analysis of tumors. Proc Natl Acad Sci USA. 2010 Sep 28;107(39):16910-5. Landau DA, Carter SL, Stojanov P, McKenna A, Stevenson K, Lawrence MS, Sougnez C, Stewart C, Sivachenko A, Wang L, Wan Y, Zhang W, Shukla SA, Vartanov A, Fernandes SM, Saksena G, Cibulskis K, Tesar B, Gabriel S, Hacohen N, Meyerson M, Lander ES, Neuberg D, Brown JR. [ PubMed ] Getz G, Wu CJ. Evolution and impact of subclonal mutations in chronic lymphocytic leukemia. Cell. 2013 Feb 14;152(4):714-26. doi: 10.1016 / j.cell.2013.01.019. Raine KM, Van Loo P, Wedge DC, Jones D, Menzies A, Butler AP, Teague JW, Tarpey P, Nik-Zainal S, Campbell PJ. ascatNgs: Identifying Somatically Acquired Copy-Number Alterations from Whole-Genome Sequencing Data. Curr Protoc Bioinformatics. 2016 Dec 8;56:15.9.1-15.9.17. doi: 10.1002 / cpbi.17. Litchfield K, Reading JL, Puttick C, Thakkar K, Abbosh C, Bentham R, Watkins TBK, Rosenthal R, Biswas D, Rowan A, Lim E, Al Bakir M, Turati V, Guerra-Assungao JA, Conde L, Furness AJS, Saini SK, Hadrup SR, Herrero J, Lee SH, Van Loo P, Enver T, Larkin J, Hellmann MD, Turajlic S, Quezada SA, McGranahan N, Swanton C. Meta-analysis of tumor- and T cell-intrinsic mechanisms of sensitization to checkpoint inhibition. Cell. 2021 Feb 4;184(3):596-614.e14.

[0165] All references cited in this specification are incorporated by reference in their entirety and for all purposes to the same extent as if each individual publication or patent or patent application was specifically and individually indicated to be incorporated by reference in its entirety.

[0166] The specific embodiments described herein are provided by way of example, not by way of limitation. Those skilled in the art will appreciate that various improvements and modifications of the described compositions, methods, and uses of the technology will be apparent without departing from the scope and spirit of the technology as described. Any subheadings herein are included for convenience only and should not be construed as limiting the disclosure in any way.

[0167] The methods of any embodiment described herein may be provided as a computer program configured, when executed on a computer, to perform one or more of the methods described above, or as a computer program product or computer readable medium carrying the computer program.

[0168] Unless the context dictates otherwise, the feature descriptions and definitions set forth above are not limited to any particular aspect or embodiment of the present invention, but apply equally to all aspects and embodiments described.

[0169] Throughout the specification and claims, the following terms shall be construed as expressly associated herein, unless the context clearly dictates otherwise. The phrase "in one embodiment" as used herein does not necessarily refer to the same embodiment, but may. Additionally, the phrase "in another embodiment" as used herein does not necessarily refer to different embodiments, but may. Thus, as described below, various embodiments of the invention can be readily combined without departing from the scope or spirit of the invention.

[0170] It must be noted that, as used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from "about" one particular value and / or to "about" another particular value. When such a range is expressed, another embodiment includes from the one particular value and / or to the other particular value. Similarly, when values ​​are expressed as approximations by use of the antecedent word "about," it will be understood that the particular value forms another embodiment. The term "about" in connection with numerical values ​​is optional and means, for example, ±10%.

[0171] Throughout this specification, including the claims which follow, unless the context otherwise requires, the words "comprise" and "include", as well as variations such as "comprises", "comprising", and "including", will be understood to imply the inclusion of a stated whole or step or group of whole or steps, but not the exclusion of any other whole or step or group of whole or steps.

[0172] Other aspects and embodiments of the present invention provide those aspects and embodiments described above, where the term "comprising" is replaced by the term "consisting of" or "consisting essentially of", unless the context indicates otherwise.

[0173] "And / or," as used herein, should be construed as a specific disclosure of each of the two specified features or components with or without the other. For example, "A and / or B" should be construed as a specific disclosure of (i) A, (ii) B, and (iii) each of A and B, as if each were individually set forth herein.

[0174] Where appropriate, the features disclosed in the foregoing description, or in the following claims, or the accompanying drawings, either expressed in their specific form, or expressed in terms of means for performing a disclosed function, or in terms of a method or process for achieving a disclosed result, may be utilized separately or in any combination of such features, to realize the invention in various of its forms.

Claims

1. 1. A method for determining whether a tumor-specific mutation in a subject is likely to be clonal, comprising: and a processor configured to receive sequence data from one or more samples from the subject that contain tumor genetic material, for each of the one or more samples, a number of reads that are indicative of the tumor-specific mutation in the sample (d ｂ ), a number of reads representing the corresponding germline allele in the sample, and a total number of reads at the position of the tumor-specific mutation (d); and The processor determines the likelihood that the tumor-specific mutation is clonal by: a prior probability that the mutation is clonal; the probability that the sequence data is observed when the tumor-specific mutation is (i) clonal, and (ii) non-clonal, given a tumor fraction for each of the one or more samples, and one or more candidate binding genotypes, each including a genotype at the position of the tumor-specific mutation, for a normal population, a reference tumor population that does not contain the tumor-specific mutation, and a mutant tumor cell population that includes the tumor-specific mutation; The posterior probability depends on The method includes:

2. 2. The method of claim 1, wherein the probability of observing the sequence data given the tumor fraction and one or more candidate binding genotypes for each sample depends on the probability of observing the sequence data given the tumor fraction, the cancer cell fraction, and the one or more candidate binding genotypes (Pr(d,db|π,φ,t)), and optionally the probability of observing the sequence data given the tumor fraction, the cancer cell fraction, and the one or more candidate binding genotypes is a weighted sum of the probabilities of observing the sequence data given each of the tumor fraction, the cancer cell fraction, and the one or more candidate binding genotypes.

3. 3. The method of claim 1 or 2, wherein the probability of observing the sequence data given the tumor fraction and one or more candidate binding genotypes for each sample is determined as an integral over all possible values ​​of the cancer cell fraction for each sample, the cancer cell fraction being the proportion of tumor cells containing the tumor-specific mutation.

4. 3. The method of claim 1 or 2, wherein sequence data from a plurality of samples is obtained, and the probability of observing the sequence data given a tumor fraction and one or more candidate binding genotypes for each of the plurality of samples is determined as the product of the probability of observing the sequence data for each sample given the tumor fraction in each sample and the one or more candidate binding genotypes.

5. 3. The method of claim 1 or 2, further comprising determining, by the processor, for each sample, at least one estimate of the tumor fraction and at least one corresponding set of one or more candidate binding genotypes.

6. 3. The method of claim 1 or 2, wherein the probability of observing the sequence data combines probabilities of observing a plurality of the sequence data given a respective tumor fraction and a corresponding set of one or more candidate binding genotypes for at least one of the one or more samples, and optionally the method includes obtaining, by the processor, a plurality of estimates of the tumor fraction and a plurality of corresponding sets of one or more candidate binding genotypes for at least one of the one or more samples.

7. The method of claim 1 or 2, wherein the prior probability that the mutation is clonal is set to a neutral prior value or a value derived from prior data and / or expert knowledge.

8. 3. The method of claim 1 or 2, further comprising repeating the method for a plurality of tumor-specific mutations identified in the subject, and optionally further comprising ranking or otherwise prioritizing the plurality of tumor-specific mutations based at least in part on their determined likelihood to be clonal in the subject.

9. 3. The method of claim 1 or 2, further comprising: optionally using sequencing data from one or more samples from the subject containing tumor genetic material and one or more germline samples from the subject to identify one or more tumor-specific mutations in the subject.

10. 3. The method of claim 1 or 2, further comprising providing the determined probability that the tumor-specific mutation is clonal and / or a value derived therefrom or related thereto to a user, for example via a user interface.

11. 1. A method for identifying one or more clonal neo-antigens in a subject, comprising: identifying a plurality of tumor-specific mutations in said subject; Determining whether one or more of the tumor-specific mutations are likely to be clonal in the subject using the method of claim 1 or 2; and determining whether said one or more of said tumor-specific mutations are likely to give rise to neoantigens; In a method comprising: A clonal neo-antigen is a tumor-specific mutation that satisfies one or more predetermined criteria as to whether said tumor-specific mutation is likely to be clonal and one or more criteria as to whether said tumor-specific mutation is likely to give rise to a neo-antigen.

12. 12. The method of claim 11, wherein the clonal neo-antigen is a tumor-specific mutation that meets at least a criterion selected from: having a probability of being clonal above a predefined threshold; having a probability of being clonal above a threshold that is adaptively set to select a predefined number of tumor-specific mutations with the highest probability of being clonal among the tumor-specific mutations for which the probability was determined; and having a probability of being clonal above a threshold that is adaptively set to select a predefined top percentile of tumor-specific mutations among the tumor-specific mutations for which the probability was determined.

13. The method of claim 11, wherein the clonal neo-antigen is a tumor-specific mutation that meets at least the criteria selected from: being associated with an expression product expressed in tumor cells; being predicted to give rise to a protein or peptide that is not expressed in normal cells of the subject; being predicted to give rise to at least one peptide that can be presented by an MHC molecule, preferably an MHC allele known to be present in the subject; and being predicted to give rise to an immunogenic protein or peptide.

14. 11. A method of providing a prognosis to a subject diagnosed with cancer, comprising identifying a plurality of tumor-specific mutations in one or more samples from the subject, and determining the likelihood that each of the tumor-specific mutations is clonal using the method of claim 1 or 2.

15. 1. A method of providing immunotherapy to a subject diagnosed with cancer, comprising: Identifying one or more clonal neo-antigens using the method of claim 11; and Designing an immunotherapy that targets the one or more clonal neo-antigens identified. The method includes:

16. 16. The method of claim 15, wherein the immunotherapy targeting the one or more clonal neo-antigens is an immunogenic composition, a composition comprising immune cells, or a therapeutic antibody.

17. 17. The method of claim 16, wherein the immunotherapy is a composition comprising T cells that recognize at least one of the one or more identified clonal neo-antigens.

18. 18. The method of claim 17, wherein the composition is enriched for T cells that target at least one of the one or more identified clonal neo-antigens, and optionally the method comprises obtaining a T cell population and expanding the T cell population to increase the number or relative proportion of T cells that target at least one of the one or more identified clonal neo-antigens.

19. 20. A composition comprising a population of T cells obtained or obtainable by the method of claim 18.

20. 20. Use of the composition of claim 19 for the manufacture of a medicament for treating a subject diagnosed with cancer.

21. A processor; and A computer readable medium comprising instructions which, when executed by said processor, cause said processor to perform the steps of the method according to claim 1 or 2. A system including:

22. One or more non-transitory computer-readable media comprising instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the method of claim 1 or 2.

Images