mmsig

mmsig performs mutational signature fitting and statistical inference as an R package to characterize mutational processes in cancer genomes, with emphasis on hematological malignancies.

Key Features:

• Expectation–Maximization algorithm: Uses an expectation–maximization algorithm to fit known mutational signatures to observed tumor mutation catalogs.
• Dynamic error suppression: Implements a dynamic error-suppression procedure that applies cosine similarity thresholds during signature fitting to improve specificity.
• Bootstrapping-based confidence intervals: Generates confidence intervals for fitted signatures using bootstrapping to quantify uncertainty in signature estimates.
• Transcriptional strand bias assessment: Assesses transcriptional strand bias of mutations to provide insight into strand-specific mutational mechanisms.
• Detection of low-abundance signatures: Detects low-abundance signatures, including APOBEC cytidine deaminase signatures, consistent with refined definitions such as COSMIC v3.1.

Scientific Applications:

• Mutational process characterization: Characterizes biological mutational processes in cancer genomes to interpret underlying etiologies.
• Hematological malignancy analysis: Applies to analysis of mutational signatures in hematological malignancies and is adaptable to other cancers with well-characterized mutational landscapes.
• Prognostic and therapeutic biomarker identification: Supports identification of biologically and clinically relevant mutational signatures that may inform prognostication and potential therapeutic targeting.
• Detection of subtle mutational patterns: Enables detection of subtle or low-frequency mutational patterns that can inform mechanisms influencing disease progression or treatment response.

Methodology:

Signature fitting using an expectation–maximization algorithm, dynamic error suppression employing cosine similarities, bootstrapping to generate confidence intervals, and transcriptional strand bias assessment.