Cancer effect size is a metric that quantifies the intensity of natural selection on somatic mutations in cancer cell lineages. The output of CancereffectsizeR primarily includes quantified cancer effect sizes (selection intensities) of somatic variants, i.e., how much each mutation contributes to cancer cell proliferation and survival in a given tumor cohort (it produces site-specific mutation rates). Because of the variants being single-nucleotide variants (SNVs), users will see that the endpoint coordinate on the chromosome is one greater than the start position.
Each variant is shown with its cancer effect size (scaled selection coefficient, γ). For display and shading, γ is transformed as s = 10 · log10(γ), so neutral selection (γ = 1) corresponds to s = 0 and larger s means stronger positive selection. The inverse is γ = 10s/10.
You do not need to back-transform anything. Hovering over a variant shows its consequence, the cancer effect size γ, the transformed score s (labeled with an asterisk so it is not mistaken for γ), and the number of TCGA patients carrying the variant. Example from the breast cancer track: TP53_R175H (missense) | cancer effect size gamma = 10,000 | transformed score* = 40 | TCGA patients with variant: 19 | *transformed score s = 10*log10(gamma), not gamma.
Why scores are never negative. The bigBed format requires the standard BED score field to be a whole number from 0 to 1000. The transformed score is therefore rounded, and negative values (γ < 1, negative selection) would be stored as 0. Variants whose original estimate was recorded only as γ < 10 (8 variants across all tracks) are stored with a score of 0 and shown as “gamma = <10”. Because s is rounded, γ is approximate (within about ±12%).
Consistent with established cancereffectsizeR practice, this track includes only recurrently mutated single-nucleotide variants (i.e., variants observed in two or more tumors within this cohort). Reliable estimation of site-specific selection intensity requires sufficient observations at a site to distinguish selective signal from stochastic variation in the underlying neutral mutation process; non-recurrent (singleton) variants are therefore not represented in this track (Mandell et al., Cancer Research, 2023).
You can search for specific cancer effect size variants by typing the gene_variant names (e.g., TP53_R175G, DCUN1D5_D116A) in the UCSC Genome Browser search box. These names correspond to the mutation identifiers indexed in this track, enabling quick navigation to relevant variants.
Each variant is labeled with its consequence: missense, silent (synonymous, e.g. CLCNKA_I436I), nonsense, stop_lost or noncoding. The label appears on hover and on the item details page. To hide silent mutations, click the track name (or right-click the track and choose “Configure”), uncheck “silent” under Variant consequence, and click Submit. Silent variants are kept because cancereffectsizeR estimates their effect sizes in the same way as other variants; a high γ for a silent variant may reflect selection acting through splicing or gene regulation, or an underestimated local mutation rate, and should be interpreted with that in mind.
The track is presented in bigBed format. The file has the following fields:
The cancer effect size in CancereffectsizeR is calculated by estimating the neutral mutation rate at each variant site using tissue-specific gene mutation rates and mutational signatures, then comparing the observed mutation rate to this neutral rate. The resulting score, called the selection intensity or cancer effect size, reflects the relative proliferative advantage that a mutation confers to cancer cells.
Please feel free to contact Dr. Jeffrey P. Townsend (jeffrey.townsend@yale.edu) with any questions and/or concerns regarding this track.
Mandell JD, Cannataro VL, Townsend JP. Estimation of neutral mutation rates and quantification of somatic variant selection using cancereffectsizeR. Cancer Research. 2023 Feb 15; 83(4):500–505. doi:10.1158/0008-5472.CAN-22-1508