INCOMMON

library(INCOMMON)
#> Warning: replacing previous import 'cli::num_ansi_colors' by
#> 'crayon::num_ansi_colors' when loading 'INCOMMON'

INCOMMON is a tool for the INference of COpy number and Mutation Multiplicity in ONcology. INCOMMON infers the copy number and multiplicity of somatic mutations from tumor-only read count data, and can be applied to tumour-only samples in an efficient and fast way. For each mutation in a sample, INCOMMON computes a probability distribution over all the possible combinations of total copy number and multiplicity, and choses the one with maximum probability.

In addition, INCOMMON offers a genome interpretation framework, in which the tumour genome is classified based on the mutant dosage of oncogenes and tumour suppressor genes. Clinical outcome analysis (e.g. survival analysis and organotropism) based on such complex genotypes can be performed with functions integrated in the package.

The INCOMMON model is designed to work with high coverage sequencing data such as targeted panels but, in principle, it can be used with any sequencing assay. INCOMMON is helpful also to analyse sequencing data from tumor-only assays, in paricular when alignment files (fastq, sam/bam, etc.) are not availble. However, if one can access higher-resolution whole-exome or whole-genome assays, specific deconvolution methodologies should be used.

The INCOMMON mutation copy-number caller

INCOMMON is a Bayesian model that can infer, for a tumour mutation, $(i)$ the total copy number at the mutant locus, and $(ii)$ the mutation multiplicity (i.e., the number of DNA copies that harbour the mutation). This information provides the allele-specific configuration of the mutant locus.

INCOMMON takes input read counts data for $n$ mutations $X=\{x_1, \ldots, x_n\}$ to develop the joint likelihood} $$\begin{equation} {p(X\mid \Theta) = \prod_{x_i \in X} p(x_i \mid \Theta) = \prod_{x_i\in X} p(d_i \mid \Theta)\, p(r_i \mid d_i, \Theta) \,.} \end{equation}$$

For every mutation $x_i=\langle r_i, d_i\rangle$, INCOMMON uses the number of reads $r_i$ with the alternative allele and the total reads $d_i$ (depth of sequencing). The model infers two sample-specific parameters, ($i$) the sample purity ($0 < \pi \leq 1$), and ($ii$) the rate of reads per chromosome copy ($\eta \in \mathbb{R}^+$), and two mutation-specific parameters, ($iii$) the tumour total copy number at the locus ($k_i\in \mathbb{Z}^+$) and ($iv$) the mutation multiplicty ($m_i\in \mathbb{Z}^+$, $m_i\leq k_i$).

The model adopted by INCOMMON relates copy number to coverage linearly, so the expected number of reads for $k$ chromosome copies is $k\eta$.

INCOMMON links the number of reads to the multiplicity and total copy number considering tumour/ normal admixing $$\begin{align} p(d_i\mid \pi, \eta, k_i) & = \text{Poisson}(d_i \mid \lambda) && \lambda = (1-\pi)2\eta + \pi \eta k_i \label{eq:likelihood_pois}\\ p(r_i\mid \pi, k_i, m_i) & = \text{Binomial}(r_i \mid d_i, \varphi) && \varphi = \frac{m_i\pi}{2\left(1-\pi\right)+k_i\pi}\label{eq:likelihood_binom} \end{align}$$

The sequencing depth follows a Poisson distribution with an expected number of reads $\lambda$ defined by combining tumour and normal readouts. Given the depth, the number of mutant reads follows a Binomial distribution with success probability $\varphi$ determined by the mixing of tumour and normal success rates.

INCOMMON uses Markov Chain Monte Carlo (implemented in stan) sampling to estimate a posterior distribution over $m$, $k$, $\eta$ and $\pi$. To leverage the massive amount of public WGS data of human tumours and gain precision with targeted assays, by default INCOMMON uses a biologically informed prior distribution for copy number and multiplicity configurations from the PCAWG and HMF cohorts.

To support orthogonal estimation of tumour purity (e.g. from histopathological evaluation) but resist potential error in the input estimate, INCOMMON centres a prior around the purity measurements provided with each sample. After posterior inference, INCOMMON uses posterior predictive checks to monitor the discrepancy between observed values and inferred posterior distributions on $k$, $m$, $\eta$ and $\pi$.

Input format

The input required for INCOMMMON classification consists of two data tables:

• genomic_data: a data table of annotated mutations with columns indicating, for each mutation, the sample name sample, mutant chromosome chr, start position from, end position to, reference allele ref, alternative allele alt, sequencing depth DP, number of reads with variant NV, variant allele frequency VAF, mutant gene name gene Hugo Symbol, and possibly the protein sequence of the variant in HGVS recommended format (preferably 1-letter amino-acid code HGVSp_Short).

• clinical_data: a data table of clinical data with matched sample names sample and purity purity (required), and clinical features like tumor type (ONCOTREE code) tumor_type (required for tumor specific priors), survival data such as OS_STATUS and time OS_MONTHS (required for survival analysis), metastasis data such as SAMPLE_TYPE (Primary or Metastasis), number of metastases MET_COUNT (required for metastatic propensity analysis) and metastatic site METASTATIC_SITE (required for metastatic tropism analysis), plus any other useful covariate.

• gene_roles: a data table reporting gene names gene and associated roles gene_role (“oncogene” or “TSG”). INCOMMON provides a set of gene roles extracted from the COSMIC Cancer Gene Census (v98) as default.

The input for downstream analysis is checked and cast in the expected format through the function init.

INCOMMON provides data from the publicly available MSK-MetTropism cohort in the correct format. The following example shows how this input is pre-processed by INCOMMON:

data(MSK_genomic_data)
data(MSK_clinical_data)
data(cancer_gene_census)

x = init(
  genomic_data = MSK_genomic_data,
  clinical_data = MSK_clinical_data,
  gene_roles = cancer_gene_census
)
#> ── INCOMMON - Inference of copy number and mutation multiplicity in oncology ───
#> 
#> ── Genomic data ──
#> 
#> ✔ Found 25659 samples, with 224939 mutations in 491 genes
#> ! No read counts found for 1393 mutations in 1393 samples
#> ! Gene name not provided for 1393 mutations
#> ! 201 genes could not be assigned a role (TSG or oncogene)
#> 
#> ── Clinical data ──
#> 
#> ℹ Provided clinical features:
#> ✔ sample (required for classification)
#> ✔ purity (required for classification)
#> ✔ tumor_type
#> ✔ OS_MONTHS
#> ✔ OS_STATUS
#> ✔ SAMPLE_TYPE
#> ✔ MET_COUNT
#> ✔ METASTATIC_SITE
#> ✔ MET_SITE_COUNT
#> ✔ PRIMARY_SITE
#> ✔ SUBTYPE_ABBREVIATION
#> ✔ GENE_PANEL
#> ✔ SEX
#> ✔ TMB_NONSYNONYMOUS
#> ✔ FGA
#> ✔ AGE_AT_SEQUENCING
#> ✔ RACE
#> ✔ Found 25257 matching samples
#> ✖ Found 513 unmatched samples

print(x)
#> ── [ INCOMMON ]  175054 PASS mutations across 24018 samples, with 290 mutant gen
#> ℹ Average sample purity: 0.4
#> ℹ Average sequencing depth: 649
#> # A tibble: 175,054 × 27
#>    sample    tumor_type purity chr     from     to ref   alt      DP    NV   VAF
#>    <chr>     <chr>       <dbl> <chr>  <dbl>  <dbl> <chr> <chr> <int> <int> <dbl>
#>  1 P-0028912 CHOL          0.3 chr17 7.58e6 7.58e6 G     A       837   133 0.159
#>  2 P-0028912 CHOL          0.3 chrX  5.32e7 5.32e7 G     A       832    85 0.102
#>  3 P-0003698 BLCA          0.2 chr17 7.58e6 7.58e6 C     A       437   109 0.249
#>  4 P-0003698 BLCA          0.2 chr5  1.49e8 1.49e8 C     T       360    36 0.1  
#>  5 P-0003698 BLCA          0.2 chr13 3.29e7 3.29e7 G     C      1027   162 0.158
#>  6 P-0003698 BLCA          0.2 chr13 3.29e7 3.29e7 G     C      1021   182 0.178
#>  7 P-0003698 BLCA          0.2 chr19 1.11e7 1.11e7 G     T       573    98 0.171
#>  8 P-0003698 BLCA          0.2 chr22 4.15e7 4.15e7 G     A       416    45 0.108
#>  9 P-0003698 BLCA          0.2 chrX  4.49e7 4.49e7 C     T       730   194 0.266
#> 10 P-0003823 BLCA          0.6 chr5  1.30e6 1.30e6 G     A       218   138 0.633
#> # ℹ 175,044 more rows
#> # ℹ 16 more variables: gene <chr>, gene_role <chr>, OS_MONTHS <dbl>,
#> #   OS_STATUS <dbl>, SAMPLE_TYPE <chr>, MET_COUNT <dbl>, METASTATIC_SITE <chr>,
#> #   MET_SITE_COUNT <dbl>, PRIMARY_SITE <chr>, SUBTYPE_ABBREVIATION <chr>,
#> #   GENE_PANEL <chr>, SEX <chr>, TMB_NONSYNONYMOUS <dbl>, FGA <dbl>,
#> #   AGE_AT_SEQUENCING <dbl>, RACE <chr>

Gene mutant dosage

Downstream of INCOMMON classification, the tumour genome can be interpreted in terms of the gene mutant dosage of tumor suppressor genes (typically through mutations with LOH) and enhanced activation of oncogenes (typically through mutations with mutant copy gain).

Mutant dosage classes (low, balanced and high) are derived from the fraction of alleles carrying the mutation (FAM)

$$\begin{equation}\label{e:expected_fma} \text{FAM}_{g,t}=\sum\limits_{k=1}^{k_{max}}\sum_{m=1}^k\frac{m}{k}p(m,k\:|\: X_{g,t})\, , \end{equation}$$

which is derived from the full posterior distribution of $p(m,k\:|\: X_{g,t})$, computed by INCOMMON.

Survival analysis

If patients’ survival status and time are provided as features in the clinical data table clinical_table, survival analysis can be performed. Downstream of classification, INCOMMON can stratify patients based on the mutant dosage of a TSG or an oncogene of interest.

INCOMMON provides the following functions, dedicated to fitting survival data:

• kaplan_meier_fit uses the Kaplan-Meier estimator to fit survival data from patients stratified with respect to the status of a specific tumor_type and gene (low, balanced or high mutant dosage)
• cox_fit uses a Cox proportional hazard ratio model to fit survival data. In addition to arguments tumor_type and gene, it accepts other covariates, given they are provided in the clinical_table.

Metastatic patterns

If information about metastatisation is provided, such as type of the sample (primary tumor or metastasis), whether patients are metastatic or not, and sites of metastatisation for primary tumors, in the clinical data table clinical_table, analysis of metastatic propensity and tropism based on INCOMMON classification and genome interpretation can be performed.

INCOMMON provides the following functions for analysis of metastases:

• met_propensity: uses a logistic regression test to compute the odds ratio (OR) of metastatisation between patients identified by the mutant dosage of a gene (low, balanced or high) for specific types of primary tumors tumor_type.

• met_tropism: uses a logistic regression test to compute the odds ratio (OR) to metastatise to a specific site METASTATIC_SITE between patients identified by the two mutational statuses of a gene (Mutant TSG with versus without LOH and Mutant oncogene with versus without amplification) for specific types of primary tumors tumor_type.