Related Experiment Video
Updated: Aug 15, 2026
06:50
Computational Prediction of Amino Acid Preferences of Potentially Multispecific Peptide-Binding Domains Involved in Protein-Protein Interactions
Published on: January 26, 2024
DNAreader: accurate prediction of DNA-binding residues in structured and disordered proteins using transformers and
Jian Zhang1,2, Sushmita Basu3,4, Jingjing Qian5
1School of Computer and Information Technology, Xinyang Normal University, Xinyang 464000, China.
Nucleic Acids Research
|August 14, 2026
Summary
DNAreader accurately predicts DNA-binding residues in both structured and disordered protein regions. This new tool overcomes limitations of existing predictors, offering improved accuracy for DNA-protein interactions.
Area of Science:
- Biochemistry and Molecular Biology
- Bioinformatics and Computational Biology
- Genomics and Proteomics
Background:
- Protein-DNA interactions are crucial for cellular functions, requiring accurate identification of DNA-binding residues (DBRs).
- Existing DBR predictors are specialized for either structured or intrinsically disordered regions (IDRs), leading to underperformance and cross-predictions.
- Current methods often misclassify residues binding non-DNA ligands as DBRs, limiting their reliability.
Purpose of the Study:
- To develop a novel predictor, DNAreader, capable of accurately identifying DBRs across both structured and intrinsically disordered protein regions.
- To address the limitations of existing predictors, including poor performance on disordered proteins and high rates of cross-prediction.
- To introduce a complementary module, DNAreaderDBIDR, for predicting DNA-binding IDRs.
Main Methods:
- Development of DNAreader, a predictor utilizing a stacked transformer encoder network.
- Implementation of batch training and contrastive learning strategies to enhance predictive performance.
- Validation using two low-similarity test datasets to assess performance against existing tools.
Main Results:
- DNAreader demonstrates statistically superior performance compared to existing DBR prediction tools.
- The predictor exhibits high accuracy for both structured and disordered protein regions.
- DNAreader significantly reduces cross-predictions, improving the specificity of DBR identification.
- The DNAreaderDBIDR module accurately predicts DNA-binding IDRs.
Conclusions:
- DNAreader is the first predictor designed for comprehensive DBR identification in both structured and disordered protein sequences.
- The tool offers improved accuracy and specificity, overcoming limitations of previous methods.
- DNAreader provides a flexible solution for analyzing DNA-binding regions in proteins, with a user-friendly web server and available source code.
