iEnhancer-DCLA

iEnhancer-DCLA predicts enhancers and their transcriptional strength from DNA sequences using k-mer word2vec encodings and deep neural networks.

Key Features:

• Enhancer and strength prediction: Predicts both the presence of enhancers and their relative transcriptional strength.
• k-mer word2vec encoding: Transforms k-mers (short DNA sequences) into numerical vectors using word2vec to form an input matrix.
• Convolutional neural networks (CNNs): Captures local sequence features from the input matrix.
• Bidirectional LSTMs (BiLSTMs): Extracts sequential dependencies and contextual information from DNA sequences.
• Attention mechanism: Identifies and emphasizes the most relevant features contributing to enhancer activity and strength.
• Evaluation metrics: Demonstrates improved performance across various evaluation metrics for enhancer and strength prediction.

Scientific Applications:

• Gene regulation and transcriptional dynamics: Study enhancer-mediated regulation and relative transcriptional strength in genomic research.
• Enhancer annotation: Identify and annotate enhancer locations within genomic sequences.
• Functional genomics and regulatory element analysis: Provide insights into regulatory DNA elements affecting transcription.

Methodology:

k-mers are converted to numerical vectors using word2vec to form an input matrix; convolutional neural networks and bidirectional LSTMs extract features; an attention mechanism refines feature importance; the model predicts enhancer presence and relative transcriptional strength and is evaluated across various evaluation metrics.