Assessing comparative importance of DNA sequence and epigenetic modifications on gene expression using a deep...

Area of Science:

• Genomics
• Epigenetics
• Bioinformatics

Background:

• Gene expression is controlled by DNA sequences and epigenetic modifications, but their interplay is complex.
• Predicting mRNA abundance requires understanding these intricate regulatory mechanisms.

Purpose of the Study:

• To develop a deep convolutional neural network framework, iSEGnet, for predicting mRNA abundance.
• To utilize DNA sequence and epigenetic data within genes and cis-regulatory regions for prediction.

Main Methods:

• Implemented iSEGnet, a deep convolutional neural network.
• Trained and validated the model using ENCODE data from six cell lines/types.
• Employed Integrated Gradients for identifying critical regulatory segments.

Main Results:

• iSEGnet outperformed existing machine learning models in predicting mRNA abundance.
• Identified key regulatory regions near promoters and transcription termination sites.
• Validated identified segments against transcription factor binding sites (ChIP-seq) and explored cancer-related transcription factors.

Conclusions:

• iSEGnet provides a powerful framework for understanding gene expression regulation.
• The model accurately pinpoints crucial DNA sequence and epigenetic elements influencing gene activity.
• iSEGnet has potential applications in identifying novel transcription factors in cancer biology.