inst/markdown/methods/germline-rarevariants.md

In CRI-iAtlas/iatlas-app: What the Package Does (One Line, Title Case)

Title: Rare Variants

Descriptions: For rare variant analysis, we focused on well-annotated, germline pathogenic or likely pathogenic cancer predisposition variants as previously defined (allele frequency in 1000 Genomes and ExAC (release r0.3.1) < 0.05%) (Huang et al., 2018). Exome files related to samples for which all the covariates (age, imputed sex, PC 1-7, and cancer type) and at least one immune trait was available were retained (N = 9,138). There were 832 pathogenic/likely pathogenic SNPs/Indels events with at least one copy of rare allele in the whole exome sequencing data, corresponding to 586 distinct pathogenic SNPs/Indels mapping to 99 genes.

We performed a pathway burden analysis using selected pre-defined biological pathways such as DNA damage repair and onco-genic processes, pan-cancer and per cancer (Bailey et al., 2018; Huang et al., 2018; Knijnenburg et al., 2018). These pathways were manually curated to generate a list of mutually exclusive pathways. The only genes that were not collapsed into pathways were BRCA1 and BRCA2 for which a sufficient number of events across cancers exist. Overall, 21 genotypic variables were used for analyses (Figure S7A). In the pan-cancer analysis, we only included genes (BRCA1/BRCA2) or pathways with a number of events (mutations) greater than 4 across cancers, including a total of 90 genes. For each pathway, variants that fall within its selected set of genes were collapsed based on the presence or absence of any rare variant (i.e., 0 if no rare variant was present and 1 if there is at least one variant). We conducted regression analyses (linear or logistic, as done for GWAS) to assess the association between the pathways’ burden of rare variants and immune traits. Traits assessed in these analyses were the same as the ones used for heritability analyses, with the addition of the immune subtypes (C1, C2, etc.), DNA-alteration related metrics such as the mutational load, the neoantigen load, the degree of copy number alterations (Thorsson et al., 2018) and the MANTIS score (threshold = 0.4, Middha et al., 2017). All pan-cancer regression models included the following covariates: age, sex, cancer type, and PC1-7.

In the pan-cancer analysis, we used a Benjamini-Hochberg FDR (Benjamini and Hochberg, 1995) to correct for multiple hypothesis testing, accounting for all 21 genes and pathways tested and 154 phenotypes (139 immune traits, 9 DNA related metrics, and 6 immune subtypes). We used a cutoff of FDR p < 0.1 to identify significant gene/pathway-immune trait associations and a threshold of nominal p < 0.005 (FDR <= 0.25) to identify suggestive associations. We used a more permissive cut-off in these analyses than the ones used in the heritability and GWAS to reduce type II error due to the low number of events (germline mutations).

Reference Listing

Contributors: Mohamad Saad, Jessica Roelands, Davide Bedognetti

CRI-iAtlas/iatlas-app documentation built on Feb. 7, 2025, 9:02 p.m.