Prediction of virus-host associations using protein language models and multiple instance learning

Abstract

Predicting virus-host association is essential to understand how viruses interact with host species, and discovering new therapeutics for viral diseases across humans and animals. Currently, the host of the majority of viruses is unknown. Here, we introduce EvoMIL, a deep learning method that predicts virus-host association at the species level from viral sequence only. The method combines a pre-trained large protein language model and attention-based multiple instance learning (MIL) to allow protein-orientated predictions. Our results show that protein embeddings capture stronger predictive signals than traditional handcrafted features, including amino acids and DNA k-mers, and physio-chemical properties. EvoMIL binary classifiers achieve AUC values of over 0.95 for all prokaryotic and nearly 0.8 for almost all eukaryotic hosts. In multi-host prediction tasks, EvoMIL achieved median performance improvements of 8.6% in prokaryotic hosts and 1.8% in eukaryotic hosts. Furthermore, EvoMIL estimates the importance of single proteins in the prediction and maps them to an embedding landscape of all viral proteins, where proteins with similar functions are distinctly clustered together.

Hidden Challenges in Evaluating Spillover Risk of Zoonotic Viruses using Machine Learning Models

“…One explanation is infectivity changes driven by limited mutations, mainly in the spike protein 23,24 . Utilizing protein language models, known for extracting feature vectors that reflect the structural and functional properties of proteins 11 , could enhance model performance by detecting key changes at the amino acid level relevant to viral infectivity 25 .…”

Predictive Modeling of Proteins Encoded by a Plant Virus Sheds a New Light on Their Structure and Inherent Multifunctionality

“…Predictive modeling is increasingly relevant in virus-host interaction research with regard to functional predictions of virus proteins and their interactions with host proteins [1][2][3][4]. The information gained from prediction models can be used to better understand the nature of virus-virus, virus-host, and host-host protein interactions during viral infection.…”

“…The information gained from prediction models can be used to better understand the nature of virus-virus, virus-host, and host-host protein interactions during viral infection. For example, powerful computational methods can predict protein structure and protein functions from viral amino acid sequences [2][3][4]. In cases where the structure of a similar protein to a candidate viral protein is experimentally determined, algorithms based on modeling can provide accurate predictions of the protein structure.…”