CLnc-Pred: A Machine Learning Approach to Predict Long Non-Coding RNAs in Crops

Area of Science:

• Genomics
• Bioinformatics
• Plant Science

Background:

• Long non-coding RNAs (lncRNAs) are critical regulatory molecules in plants.
• Existing computational tools for lncRNA identification are often not optimized for crop genomes.
• Accurate lncRNA identification is essential for understanding crop development and function.

Purpose of the Study:

• To develop a crop-specific computational tool for accurate lncRNA and coding RNA (cRNA) classification.
• To address the limitations of existing tools in identifying lncRNAs in diverse crop species.

Main Methods:

• An XGBoost classifier was trained using sequence-intrinsic features from five major crop species (wheat, sorghum, rice, soybean, maize).
• The model's performance was benchmarked against established tools like CPC2 and PLEKv2.

Main Results:

• The XGBoost classifier achieved high performance metrics: 95.30% accuracy, 93.90% precision, 98.40% recall, 96.10% F1-score, and 99.40% AUC-ROC.
• The developed classifier significantly outperformed existing benchmark tools in distinguishing lncRNAs from cRNAs.

Conclusions:

• The XGBoost classifier, deployed as CLnc-Pred, provides an efficient and accurate method for crop lncRNA prediction.
• CLnc-Pred enhances accessibility for crop lncRNA research, supporting functional and regulatory analyses.
• Future development will involve expanding datasets and incorporating multi-class ncRNA classification.