NIMBus

NIMBus models regional mutation burden in cancer whole-genome sequences using negative binomial regression and a Gamma-Poisson mixture to identify frequently mutated coding and noncoding regions.

Key Features:

• Negative Binomial Regression Framework: Employs a negative binomial regression model to account for over-dispersion in mutation count statistics across genomic regions and individuals.
• Gamma-Poisson Mixture Model: Integrates a Gamma-Poisson mixture model to capture mutation-rate heterogeneity and adjust for individual-specific variation in mutation rates.
• Integration with Genomic Features: Regresses local mutation counts against functional genomics features from the Encyclopedia of DNA Elements (ENCODE), including replication timing and chromatin organization, to estimate regional background mutation rates.
• Application to Whole-Genome Sequencing Data: Applied to whole-genome cancer sequences from the PanCancer Analysis of Whole Genomes (PCAWG) project and other cohorts to identify mutated regions.
• Identification of Non-Coding Mutational Hotspots: Extends analysis to transcription factor binding sites within promoter regions that intersect DNase I hypersensitive sites (DHSs) to detect noncoding mutational hotspots.
• Comparative Analysis Capability: Provides functionality for comparing mutation burden analysis methods to evaluate performance relative to other approaches.

Scientific Applications:

• Coding Driver Discovery: Identifies frequently mutated coding genes such as TP53 in cancer whole-genome data.
• Noncoding Driver Detection: Detects recurrent noncoding mutations including hotspots like the TERT promoter.
• Regulatory Mutational Analysis: Pinpoints mutations in transcription factor binding sites within promoters that intersect DHSs to assess impacts on gene regulation.
• Method Benchmarking: Enables comparative evaluation of mutation burden methods across cohorts and genomic regions.

Methodology:

Uses negative binomial regression together with a Gamma-Poisson mixture model, regresses local mutation counts on ENCODE genomic features (e.g., replication timing, chromatin organization), analyzes transcription factor binding sites in promoters intersecting DNase I hypersensitive sites, and is applied to whole-genome sequences from PCAWG and other cohorts.