Variational Bayesian Phylogenetic Inference

Variational Bayesian phylogenetic inference (VBPI) aims to efficiently reconstruct evolutionary relationships from molecular sequence data by transforming the complex Bayesian inference problem into an optimization problem, offering a faster alternative to traditional Markov Chain Monte Carlo methods. Current research focuses on improving the accuracy of VBPI, particularly in approximating the often multimodal posterior distributions over tree topologies and branch lengths, employing advanced techniques like graph neural networks, autoregressive models, and mixture models to achieve more flexible and accurate representations. These advancements enhance the scalability and accuracy of phylogenetic inference, leading to more reliable evolutionary tree reconstructions with implications for diverse fields including evolutionary biology, epidemiology, and comparative genomics.