Predicting Phage Behavior with PhaTYP

Bacteriophages, tiny viruses that attack bacteria, live in two ways: as virulent or temperate. Knowing which lifestyle a phage has helps understand how it interacts with its bacterial host, which is important for using phages in treatments like phage therapy. Since keeping up with the increasing number of sequenced phages is tough, scientists are turning to computers to predict phages' lifestyles. However, this is challenging due to limited data and the vast amount of phage sequences from metagenomic data. Plus, current tools struggle with short sequences. Enter PhaTYP, a new tool designed to improve lifestyle prediction for short phage sequences. It uses two training tasks: self-supervised learning and fine-tuning. PhaTYP was rigorously tested and compared with four top methods: DeePhage, PHACTS, PhagePred, and BACPHLIP. Results show PhaTYP performs better and is more reliable with short sequences. Moreover, PhaTYP was used to analyze phage lifestyles in human neonates' gut data, proving its usefulness in studying phages in metagenomic data. This helps expand our understanding of microbial communities.