iEnhancer-DCSV: Predicting enhancers and their strength based on DenseNet and improved convolutional block attention...

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Gene enhancers regulate transcription and expression, making their prediction crucial.
• Current experimental methods for enhancer identification are time-consuming and costly.
• Developing rapid and accurate computational tools is essential for genomic research.

Purpose of the Study:

• To propose a novel computational predictor, iEnhancer-DCSV, for identifying gene enhancers and predicting their strength.
• To enhance the accuracy and efficiency of enhancer prediction using deep learning techniques.
• To provide an accessible tool for researchers without requiring complex mathematical expertise.

Main Methods:

• Developed iEnhancer-DCSV using a modified densely connected convolutional network (DenseNet) and an improved convolutional block attention module (CBAM).
• Employed one-hot and nucleotide chemical property (NCP) coding for sequence representation.
• Utilized channel and spatial attention for feature evaluation, followed by a fully connected neural network and ensemble learning for prediction.

Main Results:

• Achieved 78.95% accuracy and a Matthews correlation coefficient (MCC) of 0.5809 for enhancer recognition (first layer).
• Attained 80.70% accuracy and an MCC of 0.6609 for enhancer strength prediction (second layer).
• The iEnhancer-DCSV method demonstrated high performance on the test set.

Conclusions:

• iEnhancer-DCSV offers a computationally efficient and accurate approach for enhancer prediction.
• The deep learning architecture effectively extracts and evaluates relevant genomic features.
• This tool can accelerate genomic studies by simplifying enhancer identification and strength assessment.