repgenHMM

repgenHMM models immune receptor generation from sequence data by using a Hidden Markov Model and dynamic programming to infer the probabilistic rules of gene rearrangements.

Key Features:

• Hidden Markov Model framework: Uses an HMM with dynamic programming to enumerate plausible scenarios that generate receptor sequences, including random choices of gene templates and base pair deletions and insertions occurring during T and B cell development.
• Baum-Welch parameter inference: Implements the Baum-Welch algorithm to iteratively infer probability distributions and estimate parameters for different rearrangement events.
• Application to T cell receptor chains: Validated on T cell receptor alpha and beta chain sequence data and on synthetic sequences generated by a known model.
• Synthetic sequence generation and probability/diversity estimation: Generates synthetic receptor sequences from inferred models, computes the probability of generation for any given sequence, and estimates theoretical repertoire diversity including for human T cells.
• Quantification of rearrangement distributions: Produces distributions of rearrangement events to support analysis of selection pressures and repertoire dynamics.

Scientific Applications:

• Modeling generative processes in immune repertoires: Characterizes the probabilistic rules of gene rearrangements underlying receptor generation in T and B cells.
• Analysis of T cell receptor repertoires: Estimates generation probabilities and theoretical diversity for TCR alpha and beta chains.
• Method validation with synthetic data: Provides synthetic sequences with known generative parameters for benchmarking inference methods.
• Investigation of selection and dynamics: Enables studies of selection pressures and repertoire dynamics by quantifying rearrangement distributions.

Methodology:

Applies dynamic programming within a Hidden Markov Model and uses the Baum-Welch algorithm to infer rearrangement parameters, generate synthetic sequences, compute sequence generation probabilities, and estimate repertoire diversity.