lncScore: alignment-free identification of long noncoding RNA from assembled novel transcripts

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA sequencing (RNA-Seq) is crucial for identifying novel long non-coding RNAs (lncRNAs).
• Current transcript reconstruction methods often yield partial-length protein-coding transcripts, increasing false positive lncRNA identification.
• Sequencing and assembly errors further complicate accurate lncRNA prediction.

Purpose of the Study:

• To develop a novel, accurate, and efficient alignment-free tool for distinguishing lncRNAs from messenger RNAs (mRNAs).
• To address the challenge of identifying lncRNAs, particularly partial-length transcripts.

Main Methods:

• Development of lncScore, an alignment-free tool utilizing a logistic regression model with 11 selected features.
• Comparative analysis against state-of-the-art alignment-free tools like CPAT, CNCI, and PLEK.
• Implementation of multithreading for accelerated prediction and model training.

Main Results:

• lncScore demonstrates superior performance in distinguishing lncRNAs from mRNAs, especially partial-length mRNAs, in human and mouse datasets.
• The tool shows robust performance across diverse species including Zebrafish, Fly, C. elegans, Rat, and Sheep.
• lncScore achieves significantly faster classification and model training times compared to existing tools.

Conclusions:

• lncScore offers an improved solution for accurate lncRNA identification, overcoming limitations of existing methods.
• The tool's efficiency and cross-species applicability make it valuable for genomic research.
• lncScore provides a faster and more reliable approach to lncRNA discovery.