DistRMI: a deep distance-aware neural network for explainable RNA loop motif-small molecule interaction prediction.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology and Bioinformatics
• Drug Discovery and Development

Background:

• RNA's role in disease regulation and its potential as a therapeutic target.
• Limitations of current RNA-targeted drug discovery methods, including accuracy and interpretability.
• The need for precise prediction of RNA-small molecule interactions to accelerate drug development.

Purpose of the Study:

• To develop an accurate and interpretable model for predicting RNA-small molecule interactions.
• To address the limitations of existing methods in guiding lead compound screening and elucidating mechanisms of action.
• To facilitate the discovery of novel RNA-targeted drugs for various diseases.

Main Methods:

• Integration of Transformer and Graph Neural Networks (GNNs) for feature extraction.
• Utilizing distance priors and a distance-aware attention mechanism to capture interaction information.
• Applying the DistRMI model to predict binding preferences between RNA loop motifs and small molecules.

Main Results:

• DistRMI significantly outperforms existing baseline models in prediction accuracy.
• The model demonstrates robust performance with novel RNA loop motifs and small molecules.
• Attention weight visualization highlights the importance of bases near paired bases in RNA loop motifs.

Conclusions:

• DistRMI provides a reliable and interpretable approach for understanding RNA-small molecule interactions.
• The findings facilitate lead compound screening and mechanism of action studies for RNA-targeted drugs.
• This work opens new avenues for RNA-based therapeutic development and disease treatment.