musicatk

musicatk performs discovery, prediction, and analysis of mutational signatures in cancer genomics to characterize somatic mutation patterns arising from carcinogenic exposures and aberrant cellular processes.

Key Features:

• Flexible Mutation Counting: Supports multiple schemas for counting mutation types and combining count tables from various schemas for comparative analyses.
• Signature Deconvolution and Prediction: Implements multiple methods for deconvoluting mutational signatures and predicting sample-specific exposures using pre-existing signature sets.
• Exploratory Analysis Tools: Enables comparison of identified signatures to the Catalogue Of Somatic Mutations In Cancer (COSMIC) and embeds tumors into two dimensions using uniform manifold approximation and projection (UMAP).
• Tumor Clustering and Subgroup Analysis: Clusters tumors by exposure frequencies and identifies differentially active exposures between tumor subgroups.
• Visualization Capabilities: Plots exposure distributions across annotations such as tumor type to support interpretation of results.

Scientific Applications:

• Mutational signature discovery: Identification of distinct somatic mutational signatures in cancer cohorts.
• Exposure prediction: Estimation of signature exposures in individual tumor samples using pre-defined signatures.
• Signature annotation and comparison: Matching and comparing discovered signatures to COSMIC reference signatures.
• Tumor heterogeneity analysis: Characterizing tumor subgroups and differential exposures to study heterogeneity across cohorts.
• Visualization of exposure patterns: Examining exposure distributions across tumor types and other annotations.

Methodology:

Performs preprocessing and flexible mutation-counting with combinable count tables, applies signature deconvolution and exposure prediction methods, compares signatures to COSMIC, uses UMAP for two-dimensional embedding, performs clustering based on exposures and differential exposure analysis, and generates exposure distribution plots.