Flnc

Flnc identifies full-length long noncoding RNAs (lncRNAs) directly from RNA sequencing (RNA-seq) data using machine learning to detect both novel and annotated lncRNAs, including single-exon variants.

Key Features:

• Machine learning integration: Machine-learning models integrate four feature types—transcript length, promoter signature, presence of multiple exons, and genomic location—as predictive inputs.
• Full-length and single-exon detection: Identifies full-length lncRNAs and includes single-exon variants directly from RNA-seq data.
• No transcription-initiation profiling required: Operates without transcriptional initiation profiles such as H3K4me3 ChIP-seq.
• High AUROC: Reports an Area Under the Receiver Operating Characteristic (AUROC) exceeding 0.92.
• Improved accuracy: Achieves reported identification accuracy greater than 85%, exceeding typical conventional accuracies (<50%).
• Reduced false discovery: Reduces false discovery relative to traditional coding-potential transcript detection approaches that exhibit 30–75% false discovery rates.
• Novel and annotated lncRNA detection: Detects both novel and previously annotated full-length lncRNAs from RNA-seq datasets.

Scientific Applications:

• lncRNA discovery and annotation: Discovery and annotation of novel and annotated full-length lncRNAs from existing RNA-seq datasets.
• Single-exon lncRNA inclusion: Inclusion of single-exon lncRNAs to improve transcriptome completeness.
• Regulatory role studies: Facilitates studies of lncRNA regulatory roles in human development and disease by improving identification accuracy.

Methodology:

Applies machine-learning models that integrate transcript length, promoter signature, presence of multiple exons, and genomic location to RNA-seq-derived transcript data without requiring H3K4me3 ChIP-seq.