The Likelihood Prediction of Phylogenetic Trees based on Artificial Neural Network: a new perspective and preliminary attempt

摘要

Bayesian Evolutionary Analysis Sampling Trees (BEAST) is a widely spread phylogenetic inference tool using empirical evolution models and Bayesian statistics. However, the cost of calculating the likelihood function for massive sampled trees is very expensive, resulting in long execution time. For accelerating the process, this paper proposes a likelihood prediction model based on Artificial Neural Network (ANN) using the deep neighbor information between nodes from the topology representations of historical evolution trees. The experimental results indicate that the proposed method achieves 1.2-5.9x speedup factors on obtaining the likelihood probabilities in BEAST.