Identification of multiple RNAs using feature fusion.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Deep sequencing identifies numerous novel transcripts, increasing the need for efficient computational analysis.
• Non-coding RNAs (ncRNAs) have diverse functions but lack a standardized classification system.
• Distinguishing RNA functional categories is crucial for understanding their roles.

Purpose of the Study:

• To develop a computational algorithm for classifying RNA transcripts into distinct functional categories.
• To address the heterogeneity and classification challenges of non-coding RNAs.
• To create a tool that distinguishes between coding, housekeeping, small non-coding, and long non-coding RNAs.

Main Methods:

• Developed RNA classifier (RNAC) using alignment-based genomic descriptors.
• Extracted statistical, local binary patterns, and histogram features.
• Fused features to build classification models with extreme gradient boosting.
• Validated performance on multiclass and binary classification problems across four species.

Main Results:

• RNAC achieved >93% accuracy in both multiclass and binary RNA classification.
• The approach outperformed existing methods in predicting coding potential.
• Feature fusion demonstrated effectiveness for enhanced classification performance.
• Validated the utility of RNAC across multiple species.

Conclusions:

• RNAC is a valuable computational tool for accurate RNA identification and classification.
• The method provides an efficient solution for categorizing diverse RNA types.
• The findings highlight the importance of feature fusion in improving RNA classification accuracy.