lncRNA_Mdeep

lncRNA_Mdeep predicts long non-coding RNAs (lncRNAs) from transcript sequences using an alignment-free multimodal deep learning framework that integrates Open Reading Frame homology (OFH), k-mer, and raw sequence modalities.

Key Features:

• Multimodal Deep Learning Framework: Integrates OFH (Open Reading Frame homology), k-mer, and sequence modalities combining hand-crafted features and raw sequence data.
• Alignment-Free Approach: Performs classification without relying on sequence alignment methods.
• High Prediction Accuracy: Achieved 98.73% accuracy in 10-fold cross-validation on human datasets and 93.12% independent test accuracy, outperforming eight state-of-the-art methods by 0.94%–15.41%.
• Cross-Species Validation: Validated across 11 cross-species datasets.
• Probability Output: Produces probability scores indicating whether a transcript is an lncRNA.

Scientific Applications:

• Functional Characterization: Supports identification of lncRNAs for studies of lncRNA functional mechanisms and gene regulation.
• Disease Research: Aids investigation of lncRNA involvement in complex diseases and disease pathology.

Methodology:

Input comprises three modalities (OFH, k-mer, sequence) combining hand-crafted features and raw sequence data; a multimodal deep learning framework learns high-level abstract representations from these modalities; the model is alignment-free and outputs probability scores for lncRNA classification.