Prediction of Prokaryotic and Eukaryotic Promoters Using Convolutional Deep Learning Neural Networks

TL;DR

This paper demonstrates that convolutional neural networks can accurately identify prokaryotic and eukaryotic promoters across diverse organisms, outperforming previous methods and requiring no prior knowledge of specific promoter features.

Contribution

The study introduces CNN-based models for promoter prediction in various organisms, achieving high accuracy and generalizability without relying on predefined promoter features.

Findings

CNN models achieved high sensitivity and specificity in promoter classification.

The models outperformed previous promoter prediction tools.

The approach is adaptable to newly sequenced genomes.

Abstract

Accurate computational identification of promoters remains a challenge as these key DNA regulatory regions have variable structures composed of functional motifs that provide gene specific initiation of transcription. In this paper we utilize Convolutional Neural Networks (CNN) to analyze sequence characteristics of prokaryotic and eukaryotic promoters and build their predictive models. We trained the same CNN architecture on promoters of four very distant organisms: human, plant (Arabidopsis), and two bacteria (Escherichia coli and Mycoplasma pneumonia). We found that CNN trained on sigma70 subclass of Escherichia coli promoter gives an excellent classification of promoters and non-promoter sequences (Sn=0.90, Sp=0.96, CC=0.84). The Bacillus subtilis promoters identification CNN model achieves Sn=0.91, Sp=0.95, and CC=0.86. For human and Arabidopsis promoters we employ CNNs for identification of two well-known promoter classes (TATA and non-TATA promoters). CNNs models nicely recognize these complex functional regions. For human Sn/Sp/CC accuracy of prediction reached 0.95/0.98/0,90 on TATA and 0.90/0.98/0.89 for non-TATA promoter sequences, respectively. For Arabidopsis we observed Sn/Sp/CC 0.95/0.97/0.91 (TATA) and 0.94/0.94/0.86 (non-TATA) promoters. Thus, the developed CNN models (implemented in CNNProm program) demonstrated the ability of deep learning with grasping complex promoter sequence characteristics and achieve significantly higher accuracy compared to the previously developed promoter prediction programs. As the suggested approach does not require knowledge of any specific promoter features, it can be easily extended to identify promoters and other complex functional regions in sequences of many other and especially newly sequenced genomes. The CNNProm program is available to run at web server http://www.softberry.com.