A Hybrid Prediction Method for Plant lncRNA-Protein Interaction

Area of Science:

• Plant molecular biology
• Bioinformatics
• Genomics

Background:

• Long non-protein-coding RNAs (lncRNAs) play crucial roles in biological processes, including gene regulation and plant pathogen resistance.
• lncRNA-protein interactions are vital for cellular functions but remain understudied in plants compared to animals.

Purpose of the Study:

• To propose a novel computational method, PLRPIM, for predicting lncRNA-protein interactions specifically in plants.
• To address challenges in feature selection and compression for deep learning models in lncRNA-protein interaction prediction.

Main Methods:

• PLRPIM integrates deep learning (stacked sparse autoencoder for feature extraction) and shallow machine learning (Random Forest and Light Gradient Boosting Machine fusion).
• The method utilizes k-mer sequences to extract high-level, sequence-based features.

Main Results:

• PLRPIM demonstrated superior performance on Arabidopsis thaliana and Zea mays datasets compared to existing prediction tools.
• Achieved high accuracy (89.98% for Arabidopsis, 93.44% for Zea mays) and AUC (0.954 for Arabidopsis, 0.982 for Zea mays) via 5-fold cross-validation.

Conclusions:

• PLRPIM effectively predicts plant lncRNA-protein interactions, offering a valuable tool for researchers.
• The findings facilitate further investigation into lncRNA functions and regulatory mechanisms in plants.