EVlncRNA-Dpred

EVlncRNA-Dpred predicts experimentally validated long non-coding RNAs (EVlncRNAs) from high-throughput transcriptomic data using deep learning to prioritize functional lncRNA candidates.

Key Features:

• Deep learning algorithms: A three-layer deep neural network (DNN) using K-mer features is combined with a small convolutional neural network (CNN) using one-hot encoding to distinguish EVlncRNAs from mRNAs and high-throughput lncRNAs (HTlncRNAs).
• Species-specific models: Separate models are developed for human (h), mouse (m), and plant (p), which are concatenated to form EVlncRNA-Dpred (h), EVlncRNA-Dpred (m), and EVlncRNA-Dpred (p).
• Performance improvement: Demonstrates improved classification performance versus the support-vector-machine-based EVlncRNA-pred, with an AUC of 0.896 on human datasets compared to 0.582 for the predecessor.

Scientific Applications:

• Candidate prioritization: Ranks lncRNA transcripts to prioritize candidates for low-throughput experimental validation.
• Transcriptome screening: Screens lncRNA transcripts derived from high-throughput sequencing or other transcriptomic datasets to identify likely EVlncRNAs.
• Research domains: Supports studies in genomics, molecular biology, and disease research that require identification of functionally relevant lncRNAs.

Methodology:

A three-layer deep neural network using K-mer features and a small convolutional neural network using one-hot encoding are combined as a dual-model approach, with species-specific models for human, mouse, and plant concatenated into EVlncRNA-Dpred (h/m/p).