The Metropolized Partial Importance Sampling MCMC mixes slowly on minimal reversal rearrangement paths
Markov chain Monte Carlo has been the standard technique for inferring the posterior distribution of genome rearrangement scenarios under a Bayesian approach. We present here a negative result on the rate of convergence of the generally used Markov chains. We prove that the relaxation time of the Markov chains walking on the optimal reversal sorting scenarios might grow exponentially with the size of the signed permutations, namely, with the number of syntheny blocks.
Keywords: Stochastic programming ; Markov processes ; analysis of algorithms and problem complexity ; biology and genetics ; Mitochondrial Genome Arrangements ; Bayesian Phylogenetic Inference ; Markov-Chains ; Signed Permutations ; Transpositions ; Algorithm ; Sequence ; Distance
Record created on 2009-05-01, modified on 2016-08-08