Related Experiment Video
Updated: Aug 6, 2025
06:58
Determining if DNA Stained with a Cyanine Dye Can Be Digested with Restriction Enzymes
Published on: February 2, 2018
8.9K
DeepOM: single-molecule optical genome mapping via deep learning
Yevgeni Nogin1, Tahir Detinis Zur2, Sapir Margalit2
1Russel Berrie Nanotechnology Institute, Technion, Haifa 320003, Israel.
Bioinformatics (Oxford, England)
|March 17, 2023
Summary
DeepOM, a novel deep learning method, enhances optical genome mapping by improving DNA image alignment to genomic references. This computational tool boosts accuracy and efficiency for genomic analysis in human and microbiology applications.
Area of Science:
- Genomics
- Computational Biology
- Molecular Diagnostics
Background:
- Extracting genomic information from single DNA molecules via optical imaging presents significant challenges.
- Advancements in molecular diagnostics require precise and efficient genome mapping techniques.
Purpose of the Study:
- To introduce DeepOM, a novel computational method for optical genome mapping.
- To enhance the success rate of aligning DNA images to genomic references using deep learning.
Main Methods:
- Development of DeepOM, a method employing a convolutional neural network.
- Training the neural network on simulated images of labeled DNA molecules.
- Evaluation of DeepOM using acquired images of human DNA molecules in nano-channels.
Main Results:
- DeepOM demonstrates a significant advantage in alignment success rate for DNA molecules shorter than 50 kb.
- Benchmarking against Bionano Solve shows improved performance.
- The method enhances yield, sensitivity, and throughput in optical genome mapping.
Conclusions:
- DeepOM offers a powerful computational approach to optical genome mapping.
- The tool has broad applications in human genomics and microbiology.
- Publicly available source code facilitates further research and development.
