Enhancer-LSTMAtt

Enhancer-LSTMAtt identifies and classifies enhancers in DNA sequences, distinguishing enhancers from non-enhancers and strong enhancers from weak enhancers to support studies of gene regulation.

Key Features:

• Bi-directional long-short term memory (Bi-LSTM): Captures sequential dependencies in DNA sequences to model nucleotide context.
• Feed-forward attention mechanism: Focuses the model on relevant sequence regions to enhance prediction accuracy.
• Deep residual neural network architecture: Integrates Bi-LSTM and attention modules within a residual framework.
• End-to-end deep learning: Learns directly from DNA sequence inputs to output enhancer classifications without handcrafted features.
• Classification capabilities: Differentiates enhancers versus non-enhancers and discriminates strong enhancers from weak enhancers.
• Validation strategy: Performance assessed against 19 state-of-the-art methods using 5-fold and 10-fold cross-validation and independent tests.

Scientific Applications:

• Enhancer recognition: Identification of enhancer elements within genomic DNA sequences.
• Enhancer strength classification: Distinction between strong and weak enhancers for regulatory element characterization.
• Gene regulation studies: Support for analyses of transcriptional regulation by pinpointing regulatory DNA segments.
• Method benchmarking: Comparative evaluation of predictive performance against other computational enhancer prediction methods.

Methodology:

Implements a bi-directional LSTM integrated with a feed-forward attention mechanism within a deep residual neural network in an end-to-end deep learning framework, trained and evaluated using 5-fold and 10-fold cross-validation and independent tests and compared to 19 state-of-the-art methods.