Skip to content

gpvolve-v2

Markov-chain evolutionary dynamics on genotype-phenotype maps.

Given a genotype-phenotype map (gpmap-v2) and a graph over it (gpgraph-v2), gpvolve-v2 builds a row-stochastic transition matrix using a fixation model (SSWM, Moran, McCandlish, Bloom DMS, weak-mutation) and analyzes the resulting Markov state model:

  • stationary distribution and relaxation timescales
  • transition path theory: forward and backward committors, reactive flux
  • evolutionary pathways: shortest, greedy, dominant by flux, stochastic ensembles
  • PCCA+ metastable clustering
  • fitness peak and valley detection
  • Wright-Fisher and Gillespie simulation backends

Two hot loops live in Rust via PyO3 + rayon: the stochastic walker sampler (~500x faster at 2^14 states) and the BiCGSTAB committor solver (~1700x faster at 2^14 vs scipy.sparse.linalg.spsolve). Transition-matrix assembly, stationary distributions, TPT setup and PCCA+ stay in vectorized numpy/scipy where they already run under a second on 2^14 maps.

Where to start