Advances in AI can help advance genomics by helping to predict the effects of changes to DNA structures. This can be used to help medical diagnostics, create better vaccinations and support research into crop breeding.
AI can predict the effects of changes in DNA
Stockholm, 26 November, 2018
A paper published today in the world’s leading scientific journal in genomics, Nature Genetics, reviews the success of artificial intelligence (AI) techniques, especially deep learning, in analyzing the genome and predicting the effects of changes in DNA. The paper concludes that artificial intelligence has already demonstrated impressive potential in genomics, and that the technology can potentially be used for synthetic biology by learning how to automatically generate new DNA sequences with specific desirable properties.
The paper assesses many recent breakthroughs using a particular type of AI known as ‘deep learning.’ The co-authors of the paper are James Zou of the Stanford University School of Medicine and the Chan-Zuckerberg Biohub; Mikael Huss of the operational AI platform company, Peltarion, and the Karolinska Institutet in Sweden; Abubakar Abid of the Stanford School of Medicine; and Ali Torkamani, Pejman Mohammadi and Amalio Telenti, of the Scripps Research Institute.
The authors have also launched an interactive online tutorial to help introduce deep-learning AI to other genomic researchers and scientists. Aimed at both AI experts and genomics researchers, the tutorial provides an easy-to-use playbook on how to harness the power of AI specifically for the purpose of genomics research.
Predicting the effects of changes in DNA sequences is crucial in many fields of genomics and is an integral part of new medical diagnostics, vaccine development, and innovations in plant breeding. AI can help by spotting patterns in large quantities of data, which would have been difficult, if not impossible, for humans to see or understand.
In genomics, researchers need to understand not just the effect of a single change of DNA, but multiple changes; this work is currently done in lab experiments, which is costly and time-consuming. Using AI, researchers can save time and money by building an AI-model which can allow them to do more with less, and faster.