New method for modeling Wright-Fisher diffusion with selection

A new representation makes evaluating and simulating multi-allelic Wright-Fisher diffusion practically feasible.

Researchers obtained a new transition-function representation for multi-allelic Wright-Fisher diffusion. This model includes parent-independent mutation and genic selection.

The method uses a new gamma duality. This duality is obtained by augmenting classical polynomial duality with an exponential tilting. It yields a multivariate birth-death dual with explicit infinitesimal rates.

The Wright-Fisher transition law is expressed as an explicit mixture of standard Dirichlet components. The mixing weights come from the dual started from its entrance boundary.

Existing spectral approaches rely on truncation levels. These levels grow with selection strength and the number of alleles. This leads to rapidly increasing computational costs.

Other duality-based expansions lead to multi-type dual processes. Their transition probabilities involve intractable normalising constants. The new approach remains applicable for arbitrary numbers of alleles and selection coefficients.

Key facts

  • The new representation makes transition-law evaluation and simulation practically feasible.
  • The method uses a multivariate birth-death dual with explicit infinitesimal rates.
  • Numerical experiments show substantial runtime gains over specialised methods in some regimes.

All sections

Edition