DeepInteract: Deep Neural Network Based Protein-Protein Interaction Prediction Tool
Problem
Predicting which proteins physically interact is slow and costly to determine experimentally, yet essential for understanding cellular mechanisms.
Why it matters
Protein-protein interactions underpin cellular function and disease, so accurate computational prediction accelerates biological discovery.
What the AI did
DeepInteract, an integrative domain-based method, used a Deep Neural Network trained on DIP and KUPS data to predict protein-protein interactions.
Results
On independent dataset of 34,100 Saccharomyces cerevisiae PPIs: accuracy 92.67%, precision 98.31%, sensitivity 86.85%, specificity 98.51%.
Impact
Benchmark performance suggests a useful prediction tool, but only dataset validation is stated, no real-world adoption.
Record
| Venue | Current Bioinformatics |
|---|---|
| Published | 2017 |
| Work type | article |
| Authors | Sunil Patel, Rashmi Tripathi, Vandana Kumari, Pritish Kumar Varadwaj |
| Citations | 50 (OpenAlex, at import time) |
| Model or method | Deep Neural Network (DNN), 'DeepInteract' |
| Organization | not stated in source |
| Primary source | https://doi.org/10.2174/1574893611666160815150746 |
| Open access | closed |
| Keywords | protein-protein interaction, deep neural network, prediction, bioinformatics, Saccharomyces cerevisiae |
Importance rating
real but narrower advance
PPI prediction accelerates biological discovery per abstract; a useful research tool but narrower advance, no disease or population scale stated.