DNAcycP: a deep learning tool for DNA cyclizability prediction
Problem
Measuring intrinsic DNA cyclizability/bendability experimentally (via loop-seq) is laborious; a fast computational predictor was needed.
Why it matters
DNA mechanical properties influence DNA-dependent biological processes, so cheap prediction aids genome-wide studies.
What the AI did
A deep-learning tool (DNAcycP) was trained on loop-seq data to predict intrinsic DNA cyclizability scores (C-scores) from sequence.
Results
High-fidelity predictions vs experimental data; verified on independent in vitro selection dataset; no numeric metrics stated
Impact
Enables genome-wide cyclizability prediction and revealed conserved high bendability at nucleosome dyads in yeast and mouse.
Record
| Venue | Nucleic Acids Research |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Keren Li, Matthew S. Carroll, Reza Vafabakhsh, Xiaozhong A. Wang, Ji-Ping Wang |
| Citations | 40 (OpenAlex, at import time) |
| References | 71 |
| Model or method | deep learning (specific architecture not stated) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1093/nar/gkac162 |
| Open access | gold |
| Keywords | deep learning, DNA cyclizability, loop-seq, nucleosome, prediction tool |
Importance rating
real but narrower advance
Abstract states DNA mechanical properties play a critical role in every DNA-dependent biological process, but the tool itself is a research aid for genome-wide cyclizability studies — a real but narrower scientific advance.