HyperHMM

HyperHMM infers evolutionary and progressive dynamics from the sequential acquisition of binary traits (e.g., genetic changes, symptoms, resistance traits) to model processes such as bacterial drug resistance and cancer progression.

Key Features:

• Hypercubic transition networks: Models dependencies between binary traits using hypercubic state-space representations within hidden Markov models (HMMs).
• Adapted Baum-Welch algorithm: Uses an adapted Baum-Welch expectation-maximization (EM) algorithm with resampling to quantify uncertainty and improve computational convergence compared with traditional Bayesian approaches.
• Trait interaction flexibility: Permits any combination of traits to exert arbitrary positive or negative effects on each other's acquisition.
• Scalability and efficiency: Provides orders-of-magnitude faster inference for large numbers of traits while maintaining robustness of inferred dynamics.

Scientific Applications:

• Bacterial drug resistance: Infers orders and interactions of resistance-associated traits to study evolution under antibiotic selection.
• Cancer progression: Analyzes the sequence and influence of genetic mutations or phenotypic changes in tumor evolution.
• General biological pathways: Explores complex, non-additive interactions among binary traits across evolutionary and disease-related systems.

Methodology:

HyperHMM performs hypercubic inference using hypercubic transition networks based on hidden Markov models and an adapted Baum-Welch (EM) algorithm with resampling to quantify uncertainty.