Phylogenies as graphs: structured neural networks improve host origin predictions from paramyxovirus sequences
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Accurately identifying the host of a virus from its genome sequence is a task with important applications in zoonotic disease surveillance and filling data gaps for metagenomic sampling. Machine learning approaches have seen broad application in making host predictions directly from viral genome sequences. However, most host prediction models do not incorporate information on viral phylogeny, which is strongly correlated with both genome composition and host. We apply a novel graph neural network (GNN) approach which explicitly represents viral phylogeny in model architecture to predict hosts of origin within the paramyxoviruses. We conduct rigorous benchmarking against non-structured neural networks and predictions made using phylogeny alone, showing that GNNs carry distinct advantages over other methods when making predictions where training data is sparse. Validation across different phylogenetic scales shows that simple phylogenetic prediction is effective in many applications and that phylogeny contributes a large proportion of the predictive power of host prediction models, with genome compositional features providing additional power only for specific predictions outside the range of the training data. This novel modelling approach and model validation framework are flexible and can be applied to other viral families.
Author summary
Advances in genome sequencing technology in the last two decades have resulted in a huge increase in the number and diversity of available viral genome sequences. However, many of these genomes do not have reliable associated data regarding the host which the source virus infects. Predicting the host a virus infects based on its genome sequence is therefore important both for the surveillance of emerging diseases from animal reservoirs and to fill data gaps which will enable future research. Computational models using machine learning have been widely applied to this problem. However, most previous applications do not consider the phylogenetic relationships between viruses, which has a strong correlation with host. Here, we have applied graph neural networks to predict the hosts of the paramyxoviruses, a family of viruses that contains numerous endemic and emerging threats to human and animal health. Our approach allows us to explicitly represent the relationship between viral sequences in the model architecture, helping improve predictions. We thoroughly assess our model’s performance relative to alternatives, finding that phylogeny is very important to host prediction and that our novel graph neural network approach results in more accurate predictions compared to other methods.
