mhn: a Python package for analyzing cancer progression with Mutual Hazard Networks

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Cancers often progress undetected due to an observational gap until diagnosis.
• Mutual Hazard Networks (MHNs) are statistical models designed to reconstruct the dynamics of cancer progression.
• Existing MHN models face limitations in numerical efficiency, restricting analysis to fewer mutational events.

Purpose of the Study:

• To introduce mhn, a novel Python package for dynamic cancer progression analysis using MHNs.
• To overcome numerical efficiency challenges in training MHNs, enabling analysis of a larger number of mutational events.
• To provide researchers and clinicians with a fast, user-friendly framework for studying cancer dynamics.

Main Methods:

• The mhn package trains MHN models using tumor genotypes.
• It employs a state space restriction technique to enhance numerical efficiency during model training.
• The package supports the analysis of cancer progression involving more than 100 mutational events.

Main Results:

• The mhn package facilitates the reconstruction of the most likely evolutionary history of tumors.
• It enables the sampling of artificial tumor histories for further analysis.
• mhn provides visualization tools for genomic interactions and likely cancer progression trajectories.

Conclusions:

• The mhn Python package significantly extends previous MHN implementations by improving numerical efficiency.
• It allows for the analysis of complex cancer progression models with over 100 mutational events.
• mhn offers a powerful and accessible tool for advancing the study of cancer dynamics.