TExCNN: Leveraging Pre-Trained Models to Predict Gene Expression from Genomic Sequences.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Understanding DNA sequence and gene expression is crucial.
• Deep learning can predict gene expression from genomic data.
• Traditional methods struggle with DNA sequence patterns.

Purpose of the Study:

• To develop a novel framework for predicting gene expression from DNA sequences.
• To integrate pre-trained models for enhanced DNA sequence representation.
• To improve upon existing gene expression prediction methods.

Main Methods:

• Introduced TExCNN, a framework integrating DNABERT and DNABERT-2 pre-trained models.
• Generated DNA sequence embeddings using these pre-trained models.
• Employed a convolutional neural network architecture for prediction.

Main Results:

• TExCNN achieved a higher R² score (0.622) than DeepLncLoc (0.596).
• Performance improved with longer DNA sequences (R² of 0.639 for 50,000 bp).
• Incorporating additional biological features further enhanced prediction accuracy.

Conclusions:

• Pre-trained models significantly boost gene expression prediction accuracy.
• TExCNN performs optimally with longer DNA sequences.
• The TExCNN pipeline is adaptable for various prediction scenarios.