DeepPPI: Boosting Prediction of Protein–Protein Interactions with Deep Neural Networks
Problem
Predicting which proteins physically interact is difficult, and most disease-associated protein variants have unknown interaction mechanisms.
Why it matters
Protein-protein interactions underpin disease mechanisms and drug discovery, so better prediction accelerates biomedical research.
What the AI did
DeepPPI used deep neural networks to learn protein representations from common protein descriptors and predict protein-protein interactions.
Results
Test set: Accuracy 92.50%, Precision 94.38%, Recall 90.56%, Specificity 94.49%, MCC 85.08%, AUC 97.43%
Impact
Evidence supports improved PPI prediction on the authors' test dataset only.
Record
| Venue | Journal of Chemical Information and Modeling |
|---|---|
| Published | 2017 |
| Work type | article |
| Authors | Xiuquan Du, Shiwei Sun, Chang-Lin Hu, Yu Yao, Yuanting Yan, Yanping Zhang |
| Citations | 285 (OpenAlex, at import time) |
| References | 80 |
| Model or method | DeepPPI (deep neural network) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1021/acs.jcim.7b00028 |
| Open access | closed |
| Keywords | protein-protein interactions, deep neural networks, DeepPPI, prediction, bioinformatics |
Importance rating
significant — thousands to millions affected
PPI prediction underpins disease mechanism understanding and drug discovery per abstract; no scale or application outcomes stated, method-level advance.