Exploring Protein-DNA Binding Residue Prediction and Consistent Interpretability Analysis Using Deep Learning
Bibliographic record
Abstract
A bstract Accurately identifying DNA-binding residues is a crucial step in developing computational tools to model DNA-protein binding properties, which is essential for binding pocket discovery and related drug design. Although several tools have been developed to predict DNA-binding residues based on protein sequences and structures, their performance remains limited, and proteins with crystal structures still represent only a small fraction of DNA-binding proteins. Additionally, the process of extracting handcrafted features for protein representation is labor-intensive. In this study, we combined the strengths of pre-trained protein language models and attention mechanisms to propose a sequence-based method: an attention-based deep learning approach for accurately predicting DNA-binding residues, incorporating a contrastive learning module. Our method outperformed all other sequence-based models across two prevalent benchmark datasets. Furthermore, we developed a structure-based graph neural network (GNN) model to demonstrate the impact of the contrastive module. A common limitation of existing models is their lack of interpretability, which hinders our ability to understand what these models have learned. To address this, we introduced a novel perspective for interpreting our sequence-based model by analyzing the consistency between attention scores and the edge weights generated by the GNN model. Interestingly, our results show that large-scale pre-trained protein language models, together with attention mechanisms, can effectively capture structural information solely from protein sequence inputs.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".