Xlnc1DCNN

Xlnc1DCNN classifies long non-coding RNAs (lncRNAs) and protein-coding transcripts (PCTs) using a one-dimensional convolutional neural network to enable accurate identification and interpretation of transcript sequence features.

Key Features:

• Deep Learning Architecture: Employs a one-dimensional convolutional neural network to process nucleotide sequence data and capture patterns that distinguish lncRNAs from PCTs.
• Interpretability: Provides explanations that highlight sequence regions contributing to classification decisions.
• High Performance: Demonstrated higher accuracy and F1-scores on human test sets compared to other state-of-the-art methods.
• Feature Insights: Identifies lncRNA-associated characteristics such as lack of conserved regions, short patterns of unknown function, and presence of transmembrane helices, versus PCT-associated conserved protein domains or families.
• Database Annotation Insights: Reveals inconsistencies in public database annotations by detecting lncRNA transcripts that contain protein domains, protein families, or intrinsically disordered regions (IDRs).

Scientific Applications:

• lncRNA identification: Classify unannotated transcripts from next-generation sequencing datasets as lncRNAs or protein-coding transcripts.
• Functional hypothesis generation: Use interpretable sequence features to generate hypotheses about lncRNA function and their roles in disease.
• Annotation curation: Inform curation of public databases by flagging transcripts with conflicting domain or disorder annotations.

Methodology:

Xlnc1DCNN applies a one-dimensional CNN to nucleotide sequences to extract features indicative of lncRNA or PCT status. The model architecture is designed to capture both local and global sequence patterns. Interpretability is achieved through mechanisms that highlight which parts of the input sequences contribute most to the classification decision.