DNSS2 : Improved ab initio protein secondary structure prediction using advanced deep learning architectures
Problem
Accurately predicting protein secondary structure (alpha-helix, beta-strand, coil) from sequence is a crucial step for contact and ab initio tertiary structure prediction.
Why it matters
Secondary structure prediction underpins broader protein structure prediction, which is essential for understanding protein function.
What the AI did
DNSS2 integrates six advanced one-dimensional deep convolutional/recurrent/residual/memory/fractal/inception networks with more sensitive HMM and MSA profile features to predict 3-state and 8-state secondary structure.
Results
Improves on DNSS1, which exceeded 80% prediction accuracy; DNSS2 systematically benchmarked on independent sets (exact figure truncated in abstract).
Impact
Evidence supports improved secondary structure prediction accuracy over the authors' prior deep learning method.
Record
| Venue | Proteins Structure Function and Bioinformatics |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Zhiye Guo, Jie Hou, Jianlin Jack Cheng |
| Citations | 42 (OpenAlex, at import time) |
| References | 73 |
| Model or method | DNSS2 (deep convolutional/recurrent/residual/memory/fractal/inception networks) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1002/prot.26007 |
| Open access | gold |
| Keywords | protein secondary structure, deep learning, DNSS2, ab initio prediction, benchmark |
Importance rating
significant — thousands to millions affected
Secondary structure prediction is a crucial step for contact and ab initio tertiary structure prediction, foundational to understanding protein function.