CapsProm: a capsule network for promoter prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Identifying DNA promoter regions is crucial for gene regulation understanding but remains a challenge.
• Existing methods, like convolutional networks, show promise but lack cross-organism versatility.
• Current approaches often require manual architecture redesign for new datasets.

Purpose of the Study:

• To develop a versatile capsule network architecture, CapsProm, for accurate promoter sequence identification.
• To enable efficient promoter prediction across diverse eukaryotic and prokaryotic organisms.
• To provide a standardized model adaptable to various DNA datasets with minimal effort.

Main Methods:

• Proposed a novel architecture named CapsProm, utilizing capsule networks.
• Applied the architecture to raw DNA sequence data from five different organisms.
• Evaluated performance against baseline methods using metrics like F1-score.

Main Results:

• CapsProm demonstrated accurate promoter identification in raw DNA data.
• The architecture proved versatile across both eukaryotic and prokaryotic datasets.
• Outperformed the baseline method in five out of seven tested datasets based on F1-score.

Conclusions:

• CapsProm offers a versatile and effective solution for DNA promoter region identification.
• The proposed architecture simplifies the development of promoter prediction models for new datasets.
• CapsProm shows competitive performance and potential for broader application in bioinformatics.