HiLDA

HiLDA applies a hierarchical latent Dirichlet allocation model within a Bayesian framework to statistically test whether the mutational exposures of Shiraishi-model mutational signatures differ between two groups of tumor samples.

Key Features:

• Statistical Testing: Performs statistical tests to compare mutational signature exposures between two groups.
• Inference: Provides inference on mutational patterns and their relative frequencies across tumor mutational catalogs.
• Visualization: Includes visualization capabilities for results.

Scientific Applications:

• Cancer Mutational Analysis: Characterizes somatic mutation data to investigate differences in mutational signatures between tumor sets.
• Temporal Mutation Studies: Identifies associations between mutational pattern frequencies and time of mutation occurrence, such as before or after tumor initiation in colon cancer.
• Clinical and Epidemiological Stratification: Detects significant associations between mutational signature exposures and variables including sex, age, smoking status, and tumor site in esophageal cancer.

Methodology:

HiLDA employs hierarchical latent Dirichlet allocation within a Bayesian framework to infer mutational patterns and their relative frequencies from tumor mutational catalogs and compare estimated frequencies between tumor sets.