SigNet

Leveraging deep neural networks, SigNet refits mutational profiles under sparse-mutation regimes, surpassing non-negative least squares baselines. It quantifies uncertainty via error bars on signature weights and identifies out-of-distribution samples with its Detector module. Last update: 05/2024.

CBaSE

CBaSE uses somatic point mutations to estimate per-gene mutation probabilities without requiring external covariates. Unlike methods that assume a gamma distribution for gene-level counts, CBaSE considers a broader family of models. Outputs include q-values and dN/dS to quantify negative and positive selection for each gene. Extended context can be modeled up to pentamers. Last update: 07/2024.

MutPanning

Accounting for extended context (>5-mer), MutPanning implements two complementary positive-selection tests: enrichment of nonsynonymous mutations relative to neutrality and distributional deviation from the neutral expectation. These tests target driver genes and improve sensitivity when selection acts on only a subset of sites.

Mosaic and de novo variant datasets

We provide early and late mosaic variant files, collated from 13 published studies. Variants which do not fall in either of the two mosaic categories were annotated as de novo variants.

All code generated for the analyses presented in Cortés Guzmán et al. (2025) can be found in the GitHub repository:

All code generated for the analyses in Rodriguez-Galindo et al. (2020) can be found in the bitbucket archive:

Weghorn Lab GitHub