A Support Vector Machine based method to distinguish long non-coding RNAs from protein coding transcripts.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Increasing RNA sequencing data necessitates methods to differentiate long non-coding RNAs (lncRNAs) from protein-coding transcripts (PCTs).
• Accurate classification of RNA transcripts is crucial for understanding gene regulation and function.

Purpose of the Study:

• To develop and validate a Support Vector Machine (SVM) based computational method for distinguishing lncRNAs from PCTs.
• To identify key features, including nucleotide pattern frequencies and open reading frame (ORF) lengths, that best discriminate between lncRNA and PCT classes.

Main Methods:

• Employed a Support Vector Machine (SVM) algorithm.
• Utilized features derived from nucleotide pattern frequencies (selected via Principal Component Analysis - PCA) and relative ORF lengths.
• Trained and tested the model using transcript data from multiple species, including human, mouse, and zebrafish.

Main Results:

• Achieved high classification accuracies: 98.21% for human, 98.03% for mouse, and 96.09% for zebrafish.
• Demonstrated superior performance compared to existing tools, with an approximate 3% increase in accuracy.
• Validated the model's robustness by cross-species classification and testing on diverse organisms, including rat, pig, and fruit fly, with >84% accuracy.
• Successfully re-annotated two uncharacterized Swiss-Prot sequences as likely lncRNAs.

Conclusions:

• The proposed SVM method provides a highly effective and accurate approach for distinguishing lncRNAs from PCTs.
• The integration of PCA-selected nucleotide pattern frequencies alongside ORF features enhances classification performance.
• The developed model exhibits cross-species applicability, indicating its potential for broad use in transcriptomic data analysis.