Researchers Introduce Phyloformer: A Fast and Accurate Deep Neural Network Method for Phylogenetic Reconstruction

Jun 24, 2024, 08:12 AM

Researchers have introduced Phyloformer, a new method for phylogenetic reconstruction that leverages deep neural networks. This method is described as fast, accurate, and versatile, outperforming traditional methods such as neighbor joining and maximum likelihood methods under complex models of sequence evolution. Phyloformer can infer evolutionary distances from a multiple sequence alignment and can be applied under various evolutionary models, including LG+GC, LG+GC with indels, CherryML co-evolution model, and SelReg with selection. Additionally, the method is described as a learnable function for reconstructing phylogenetic trees from homologous sequences, providing estimates of distances between all pairs of sequences. A preprint detailing Phyloformer has been released, highlighting its capabilities and applications. The researchers also propose RecA as a pan-bacterial genetic marker and apply a Random Forest machine learning model on RecA amino acid sequences for accurate and fast taxonomic annotation across the bacterial tree of life.