A sequence-based two-layer predictor for identifying enhancers and their strength through enhanced feature extraction

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Enhancers are crucial regulatory DNA elements controlling gene expression.
• Identifying enhancers and their regulatory strength is difficult due to their dynamic nature.
• Existing machine learning methods for enhancer identification require improved accuracy and efficiency.

Purpose of the Study:

• To develop a robust two-layer prediction model for accurate enhancer identification and strength prediction.
• To enhance feature extraction strategies for improved prediction performance.
• To address the limitations of current machine learning approaches in enhancer prediction.

Main Methods:

• Proposed a two-layer prediction model utilizing an artificial neural network (ANN).
• Implemented an enhanced feature extraction strategy combining Position-Specific Amino Acid Propensity (PSTKNC), Enhanced Nucleic Acid Composition (ENAC), and Composition of k-spaced Nucleic Acid Pairs (CKSNAP).
• Validated the model on a benchmark chromatin dataset from nine cell lines using 10-fold cross-validation.

Main Results:

• The proposed model achieved high accuracy (94.50%) and Matthew's Correlation Coefficient (MCC) of 0.8903 in predicting enhancers.
• The model demonstrated strong performance on an independent test set.
• The combined feature extraction strategy significantly improved prediction accuracy compared to existing methods.

Conclusions:

• The developed two-layer prediction model with enhanced feature extraction offers a significant advancement in enhancer identification and strength prediction.
• The model's high accuracy and MCC indicate its potential for biological applications.
• Further improvements in machine learning algorithms can enhance the efficiency and accuracy of predicting gene regulatory elements.