Multi-feature fusion for deep learning to predict plant lncRNA-protein interaction

Area of Science:

• Plant molecular biology
• Bioinformatics
• Genomics

Background:

• Long non-coding RNAs (lncRNAs) are critical regulators of cellular processes.
• Many plant lncRNAs remain functionally uncharacterized.
• Understanding lncRNA-protein interactions is essential for functional genomics.

Purpose of the Study:

• To develop an accurate computational model for predicting plant lncRNA-protein interactions.
• To address the challenge of functionally uncharacterized plant lncRNAs.

Main Methods:

• An integrative prediction model, DRPLPI, was developed.
• Multi-feature fusion incorporating structural and sequence features (tri-nucleotide composition, gapped k-mer, recursive complement, binary profile).
• A multi-head self-attention long short-term memory encoder-decoder network for feature extraction.
• A meta-learner combining categorical boosting and extra trees for robust prediction.

Main Results:

• DRPLPI achieved high prediction performance in experiments on Zea mays and Arabidopsis thaliana.
• Area Under the Precision-Recall Curve (AUPRC) values of 0.9820 (Zea mays) and 0.9652 (Arabidopsis thaliana) were obtained.
• The proposed method demonstrated significant improvement over existing state-of-the-art methods.

Conclusions:

• DRPLPI provides a robust and accurate approach for predicting plant lncRNA-protein interactions.
• The model facilitates functional characterization of unannotated plant lncRNAs.
• This work enhances the toolkit for plant functional genomics research.