Our model

We designed the computational framework as shown to develop a new HP-PPI predictor based on the collected PPI data between human and Yersinia pestis :

  1. The human-Y. pestis PPIs were downloaded from HPIDB and PATRIC.

  2. Five different encoding schemes were introduced to construct feature vectors for protein pairs between human and Y. pestis.
      3.      Three sequence-based encodings:
    • AC : Auto covariance
    • CKSAAP : The composition of k-spaced amino acid pairs
    • PseTC : Pseudo-tripeptide composition
      • Two host network properties-related encodings:
    • NetTP : Network topology properties
    • NetSS : Sequence similarity measurements between pathogen protein and host protein's partners
  3. Individual predictive model for each encoding scheme was inferred by Random Forest.

  4. The five individual models was integrated into a final powerful model by the Noisy-OR algorithm.

Citation

Lian, X., Yang, S., Li, H., Fu, C., and Zhang, Z. (2019). Machine-Learning-Based Predictor of Human-Bacteria Protein-Protein Interactions by Incorporating Comprehensive Host-Network Properties. J. Proteome Res. 18, 2195–2205.

Prediction

Tips

If your input protein sequence contains less than 35 amino acids or contains non-standard amino acids, our model will not be able to process the prediction.

Human protein|sequences

OR upload your own file (human protein sequences in FASTA format) :

Bacteria protein|sequences

OR upload your own file (bacteria protein sequences in FASTA format) :



Contact