LigRNA: High-accuracy prediction of RNA-ligand binding sites via dynamic chemical complementarity and language model encoding
- 1University of Science and Technology of China Suzhou Institute for Advanced Research
Abstract
RNA-binding small molecules represent a validated therapeutic modality, yet pinpointing their exact RNA binding sites remains critical for mechanistic understanding and drug optimization. Experimental methods, while precise, are costly and low-throughput, necessitating robust computational alternatives. Approaches based on static feature concatenation may overlook input-dependent relationships between RNA residues and ligand functional groups. Here we present LigRNA, a graph neural network framework that predicts RNA-ligand binding sites through joint cross-modal encoding of both molecules. LigRNA integrates structure-guided language model embeddings with physicochemical features-solvent-accessible surface area and secondary structure types, to construct comprehensive RNA representations. Ligands are encoded as attributed molecular graphs via graph attention networks, preserving atom-level chemical topology rather than compressing molecules into fixed fingerprints. A cross-attention mechanism then enables dynamic, ligand-dependent modulation of RNA nucleotide attention, capturing context-specific interaction patterns. Benchmarked on TE9 and TE18 datasets, LigRNA achieves F1 scores of 0.73 and 0.58, improving on GATRsite by 7.4% and 18.4%, respectively. Ablation studies identify the RNA residual encoder as the most consistent contributor across the two benchmarks. The source code is freely available at https://github.com/WillowWang0216/LigRNA.
Related
- Synergy and Complementarity: The Generative Basis of Chemical Organizations
- Kinematic encoding of the speed-accuracy trade-off in approach/avoidance decisions when facing social threats
- PredIDR3: A new output-encoding scheme and abundant negative source provide more information for deep learning-based protein intrinsic disorder prediction
- Comparing serial and parallel language input: Shared neural representations with distinct encoding strengths
- Highlighting the strengths of new wearable fNIRS, precision accuracy microgenetic research designs, and new NeuroDOT processing algorithm to measure change over time in clinical populations for Speech-Language and Hearing Researchers
BibTeX
@misc{wang2026ligrna,
title = {{LigRNA: High-accuracy prediction of RNA-ligand binding sites via dynamic chemical complementarity and language model encoding}},
author = {X. Wang and H. Zhu and H. Li and A. Chang and F. Zhang and H. Chen and C. Wang and K. Chen and S. K. Zhou and P. Xiong},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.15.759315},
url = {https://www.biorxiv.org/content/10.64898/2026.10.15.759315v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - LigRNA: High-accuracy prediction of RNA-ligand binding sites via dynamic chemical complementarity and language model encoding
AU - X. Wang
AU - H. Zhu
AU - H. Li
AU - A. Chang
AU - F. Zhang
AU - H. Chen
AU - C. Wang
AU - K. Chen
AU - S. K. Zhou
AU - P. Xiong
PY - 2026
DA - 2026/10/15
PB - bioRxiv
DO - 10.64898/2026.10.15.759315
UR - https://www.biorxiv.org/content/10.64898/2026.10.15.759315v1
AB - RNA-binding small molecules represent a validated therapeutic modality, yet pinpointing their exact RNA binding sites remains critical for mechanistic understanding and drug optimization. Experimental methods, while precise, are costly and low-throughput, necessitating robust computational alternatives. Approaches based on static feature concatenation may overlook input-dependent relationships between RNA residues and ligand functional groups. Here we present LigRNA, a graph neural network framework that predicts RNA-ligand binding sites through joint cross-modal encoding of both molecules. LigRNA integrates structure-guided language model embeddings with physicochemical features-solvent-accessible surface area and secondary structure types, to construct comprehensive RNA representations. Ligands are encoded as attributed molecular graphs via graph attention networks, preserving atom-level chemical topology rather than compressing molecules into fixed fingerprints. A cross-attention mechanism then enables dynamic, ligand-dependent modulation of RNA nucleotide attention, capturing context-specific interaction patterns. Benchmarked on TE9 and TE18 datasets, LigRNA achieves F1 scores of 0.73 and 0.58, improving on GATRsite by 7.4% and 18.4%, respectively. Ablation studies identify the RNA residual encoder as the most consistent contributor across the two benchmarks. The source code is freely available at https://github.com/WillowWang0216/LigRNA.
ER -