A deep learning model for plant lncRNA-protein interaction prediction with graph attention

Area of Science:

• Plant molecular biology
• Bioinformatics
• Computational biology

Background:

• Long non-coding RNAs (lncRNAs) are key regulators, but their interactions with proteins (LPIs) in plants are understudied.
• Experimental characterization of plant lncRNAs and their protein partners is limited, hindering functional understanding.

Purpose of the Study:

• To develop and validate GPLPI, a computational method for predicting plant lncRNA-protein interactions (LPIs).
• To leverage sequence and structural information for accurate LPI prediction in plants.

Main Methods:

• GPLPI utilizes a generative model with long short-term memory (LSTM) and graph attention for feature extraction.
• Frequency chaos game representation (FCGR) captures evolutionary features.
• Manifold regularization and L2-norm enhance feature discrimination and prevent overfitting.
• Catboost and regularized Logistic regression with L-BFGS optimize final interaction predictions.

Main Results:

• GPLPI achieved high prediction accuracies: 85.76% for Arabidopsis thaliana and 91.97% for Zea mays.
• The model demonstrated superior performance compared to existing state-of-the-art methods.
• GPLPI effectively integrates sequence, structural, and evolutionary data for LPI prediction.

Conclusions:

• GPLPI offers a robust and accurate approach for predicting plant lncRNA-protein interactions.
• This method can accelerate the discovery of functional lncRNA-protein complexes in plants.
• The findings contribute to a deeper understanding of gene regulation by lncRNAs in plants.