Regression Models for Localised Mutations

Regression Models for Localised Mutations (RM2) is a tool for evaluating differential mutation rates and processes across classes of functional sites. RM2 uses negative binomial regression to assess whether sites of interest show an enrichment or depletion of mutations compared to flanking control regions of the same length. Further, RM2 was designed to test whether subclasses of mutations, like those of specific signatures, strandedness, transcription direction, and other features, show differential patterns. This allows for quick and systematic characterization of mutational processes and can be easily extended to site-based studies like pathway analysis.

As input, RM2 requires:

1. Mutation file with optional annotation columns
2. Set of genomic regions containing the chr, start and end position of the sites

Preprint

Functional and genetic determinants of mutation rate variability in regulatory elements of cancer genomes

Installation

devtools

Using the R package devtools, run: devtools::install_github('https://github.com/reimandlab/RM2')

From source

Clone the repository: https://github.com/reimandlab/RM2.git
Open R in the directory you cloned the package in and run install.packages("RM2", repos=NULL, type="source")

Usage

Required packages
Biostrings
BSgenome
BSgenome.Hsapiens.UCSC.hg19
GenomeInfoDb
GenomicRanges
ggplot2
MASS
parallel
reshape2
S4Vectors

Load the RM2 package

library(RM2)

Load sites

data("ctcf_chr3_4")
head(ctcf_chr3_4, 2)
#    chr  start    end
# 1 chr4  29087  29584
# 2 chr4 236026 236407

Load mutations

data("mutations_chr3_4")
head(mutations_chr3_4, 2)
#         chr     start       end ref alt
# 253901 chr4 128796154 128796154   G   T
# 253910 chr4 132966670 132966670   C   A

Add mutation annotations

muts = cbind(mutations_chr3_4, get_mut_trinuc_strand(mutations_chr3_4))
head(muts, 2)
#         chr     start       end ref alt mut_trinuc mut_strand ref_alt
# 253901 chr4 128796154 128796154   G   T      ACA_A          c     C_A
# 253910 chr4 132966670 132966670   C   A      ACA_A          w     C_A

Run regression for total mutations

window_size = 25
results = RM2(maf = muts,
              sites = ctcf_chr3_4,
              window_size = window_size)
results
#                 mut_type           pp this_coef obs_mut exp_mut exp_mut_lo exp_mut_hi       fc pp_cofac this_coef_cofac n_sites_tested
# 1 total_muts__total_muts 1.288732e-10 0.5575766     273     197        170    226.025 1.380711       NA              NA           7194

Visualize results

n_patients = 150
dfr = get_mutations_in_flanked_sites(muts, ctcf_chr3_4, window_size, n_patients)
plot_mutations_in_flanked_sites(dfr, window_size)

Run regression with additional mutation subclasses

mut_class_columns = c(NA, "mut_strand", "ref_alt")
results = RM2(maf = muts,
              sites = ctcf_chr3_4,
              window_size = window_size,
              mut_class_columns = mut_class_columns)

Run regression with mutation co-factor

muts$cofactor_col = sample(c(0,1), nrow(muts), replace=T)
results = RM2(maf = muts,
              sites = ctcf_chr3_4,
              window_size = window_size,
              cofactor_column = "cofactor_col")

Run downsampled RM2

n_iterations = 10
n_sites_sampled = 5000
results = RM2_downsample(maf = muts,
                         sites = ctcf_chr3_4,
                         window_size = window_size,
                         n_sites_sampled = n_sites_sampled,
                         n_iterations = n_iterations)