Protein embeddings improve phage-host interaction prediction

Abstract

AbstractWith the growing interest in using phages to combat antimicrobial resistance, computational methods for predicting phage-host interactions have been explored to help shortlist candidate phages. Most existing models consider entire proteomes and rely on manual feature engineering, which poses difficulty in selecting the most informative sequence properties to serve as input to the model. In this paper, we framed phage-host interaction prediction as a multiclass classification problem, which takes as input the embeddings of a phage’s receptor-binding proteins, which are known to be the key machinery for host recognition, and predicts the host genus. We explored different protein language models to automatically encode these protein sequences into dense embeddings without the need for additional alignment or structural information. We show that the use of embeddings of receptor-binding proteins presents improvements over handcrafted genomic and protein sequence features. The highest performance was obtained using the transformer-based protein language model ProtT5, resulting in a 3% to 4% increase of weighted F1 scores across different prediction confidence threshold,compared to using selected handcrafted sequence features.Author summaryAntimicrobial resistance is among the major global health issues at present. As alternatives to the usual antibiotics, drug formulations based on phages (bacteria-infecting viruses) have received increased interest, as phages are known to attack only a narrow range of bacterial hosts and antagonize the target pathogen with minimal side effects. The screening of candidate phages has recently been facilitated through the use of machine learning models for inferring phage-host pairs. The performance of these models relies heavily on the transformation of raw biological sequences into a collection of numerical features. However, since a wide array of potentially informative features can be extracted from sequences, selecting the most relevant ones is challenging. Our approach eliminates the need for this manual feature engineering by employing protein language models to automatically generate numerical representations for specific subsets of tail proteins known as receptor-binding proteins. These proteins are responsible for a phage’s initial contact with the host bacterium and are thus regarded as important determinants of host specificity. Our results show that this approach presents improvements over using handcrafted genomic and protein sequence features in predicting phage-host interaction.