Multifaceted Protein-Protein Interaction Prediction Based on Siamese Residual RCNN

This is the repository for PIPR (originally Lasagna in our previous preprint version). This repository contains the source code and links to some datasets used in our paper.

Abstract

Sequence-based protein–protein interaction (PPI) prediction represents a fundamental computational biology problem. To address this problem, extensive research efforts have been made to extract predefined features from the sequences. Based on these features, statistical algorithms are learned to classify the PPIs. However, such explicit features are usually costly to extract, and typically have limited coverage on the PPI information. We present an end-to-end framework, PIPR (Protein–Protein Interaction Prediction Based on Siamese Residual RCNN), for PPI predictions using only the protein sequences. PIPR incorporates a deep residual recurrent convolutional neural network in the Siamese architecture, which leverages both robust local features and contextualized information, which are significant for capturing the mutual influence of proteins sequences. PIPR relieves the data pre-processing efforts that are required by other systems, and generalizes well to different application scenarios. Experimental evaluations show that PIPR outperforms various state-of-the-art systems on the binary PPI prediction problem. Moreover, it shows a promising performance on more challenging problems of interaction type prediction and binding affinity estimation, where existing approaches fall short.

Environment

python 2.7 or 3.6
Tensorflow 1.7 (with GPU support)
CuDNN
Keras 2.2.4

./binary contains the implementation for the binary prediction task. This includes scripts to run on three datasets: Yeast, Human and multi-species.
./type contains that for the interaction type prediction task.
./regression contains that for the binding affinity prediction task. Each folder is attached with a run.sh to show how to run the evaluation program.
./embeddings contains pre-trained amino acid embeddings and the training script.

Datasets

Here we include altogether 6 datasets. New datasets processed in this work are marked ND.

1. The Yeast dataset for binary PPI prediction provided in Guo et al. 2008.
2. The multi-species dataset (C. elegans, D. melanogaster and E. coli) extracted from DIP for binary PPI prediction. (ND)
3. Added another binary PPI prediction dataset from Pan et el. 2010 under the folder sun.
4. The SHS27k dataset for interaction type prediction can be downloaded from here or from the Google Drive. (ND)
5. The larger SHS148k dataset for interaction type prediction can be found in the links above. (ND)
6. Link to the normalized SKEMPI dataset is here.

Reference

This work has been published in the Bioinformatics journal featuring ISMB/ECCB 2019.

DOI: http://dx.doi.org/10.1093/bioinformatics/btz328
Bibtex:

@article{chen2019pipr,
    title={Multifaceted Protein-Protein Interaction Prediction Based on Siamese Residual RCNN},
    author={Chen, Muhao and Ju, Chelsea and Zhou, Guangyu and Chen, Xuelu and Zhang, Tianran and Chang, Kai-Wei and Zaniolo, Carlo and Wang, Wei},
    journal={Bioinformatics},
    volume = {35},
    number = {14},
    pages = {i305-i314},
    year = {2019},
    month = {07},
    publisher={Oxford University Press}
}

MuPIPR (NAR GaB 2020)

Also check out the follow up work in the NAR Genom. Bioinform. paper Mutation effect estimation on protein–protein interactions using deep contextualized representation learning, in which a pre-trained neural language model helps the PIPR architecture to estimate the point mutation effect (e.g. estimating the change of binding affinity and the change of BSA) in PPIs.
The released software is available at guangyu-zhou/MuPIPR.