Predict human-virus protein interactions
DeepHVI predicts interactions between a given human and viral protein pair, or identifies potential natural interaction partners from a single protein sequence.
Two independent tasks
Two sequences to assess a pair. One sequence to predict natural partner sequences.
Pairwise Interaction Prediction
Enter a human protein sequence and a viral protein sequence to obtain the interaction probability, predicted label, and classification threshold.
Predict a pairInteraction Partner Sequence Prediction
Enter one human or viral protein sequence. The model predicts a partner embedding and retrieves the closest natural sequences from the corresponding reference library.
Predict partner sequencesSubmitting a prediction task
Choose your task
Select pairwise interaction prediction or partner sequence prediction according to your research question.
Add your sequences
Enter the required sequences, check the input and encoding limits, and submit the task.
Review and download
Use the private job link to view results and download them before expiry.
Protein representations and physicochemical features
ESM-2 encodes sequence features, and AAindex describes physicochemical properties. DeepHVI uses these features to classify protein pairs or predict partner embeddings for retrieval from natural protein libraries.
Read the methodsPrediction examples
The examples use sequences from the versioned training dataset and have been checked with the deployed model.
Single human-virus protein pair
Example sequences are from the versioned DeepHVI classification training dataset. Loading an example fills the form without submitting a task.
Human input, viral partners
Example sequences are from the versioned DeepHVI classification training dataset. Loading an example fills the form without submitting a task.
Cite DeepHVI
Please cite the DeepHVI paper when using this website in your research.
Wang X, Luo J, Cai X, Liu R, Li Y, Hon C. DeepHVI: A multimodal deep learning framework for predicting human-virus protein-protein interactions using protein language models. Biosafety and Health. 2025;7(4):257–266. DOI: 10.1016/j.bsheal.2025.07.005.
User Guide
Predictions are intended for research and require experimental validation.