lincRNA predict

lincRNA predict predicts long intergenic non-coding RNAs (lincRNAs) and identifies their transcription sites and regulatory motifs from genomic and RNA sequencing (RNA-seq) data using a two-layer deep neural network with an auto-encoder.

Key Features:

• Deep Learning Framework: Implements a two-layer deep neural network that incorporates an auto-encoder algorithm to capture complex patterns in genomic sequences.
• High Prediction Accuracy: Reports prediction accuracies of 100% and 92.4% on transcription sites across datasets, outperforming support vector machines and conventional neural networks.
• Knowledge-Based Approach: Integrates RNA sequencing (RNA-seq) data and recognizes conserved motifs interspersed with non-conserved regions reflecting regulated lincRNA expression.
• Discovery and Validation: Experimental results validate the capability to uncover novel transcription sites, including previously unreported ones.
• Application in Genomic Research: Facilitates identification and study of lincRNAs to support analyses of their regulatory roles and potential therapeutic targets.

Scientific Applications:

• Gene Regulation Studies: Investigating how lincRNAs influence gene expression and cellular processes.
• Disease Research: Identifying potential lincRNA biomarkers for diseases based on their regulatory roles in genetic pathways.
• Genomic Annotation: Enhancing annotation of non-coding regions within genomes to improve understanding of genomic architecture.

Methodology:

Uses annotated human DNA genome data to collect transcriptional sequences of lincRNAs; constructs and trains a two-layer deep neural network with an auto-encoder on intergenic DNA sequence data, optimizing encoding schemes; and performs feature extraction and correlation analysis to capture essential features and information correlations along genome sequences.