Protein tertiary structure modeling driven by deep learning and contact distance prediction in CASP13
Problem
Predicting protein 3D structures from sequence, including template-free (ab initio) cases where no similar known structure exists.
Why it matters
Protein structures underpin drug design and understanding of disease and biological function.
What the AI did
Enhanced the MULTICOM system with deep convolutional neural network contact-distance prediction, distance-driven ab initio modeling, and deep-learning-based model ranking.
Results
Ranked 3rd out of 98 predictors in both template-free and template-based modeling at CASP13.
Impact
Independent blind CASP13 benchmark verified the ranking, showing deep learning was the key driver.
Record
| Venue | Proteins Structure Function and Bioinformatics |
|---|---|
| Published | 2019 |
| Work type | article |
| Authors | Jie Hou, Tianqi Wu, Renzhi Cao, Jianlin Jack Cheng |
| Citations | 162 (OpenAlex, at import time) |
| References | 78 |
| Model or method | MULTICOM with deep convolutional neural networks |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1002/prot.25697 |
| Open access | hybrid |
| Keywords | protein structure, deep learning, CASP13, contact prediction, MULTICOM |
Importance rating
major — millions of lives, an industry or a field
Protein structure prediction is a key direction of a foundational scientific field since CASP11; abstract implies broad relevance to biology but states no scale.