top of page

Gene-ius Code: AI's DNA for Life-Saving Drugs

10/29/23

Editorial team at Bits with Brains

Scientists are now questioning if the same AI that can draft an email can also write a DNA sequence for a life-saving drug.

Ginkgo Bioworks, a company that works on the genetics behind a wide range of products from pharmaceuticals to fragrances, is one of the pioneers in this field. The company recently announced a five-year deal with Google Cloud to host AI focused on synthetic biology.


Ginkgo's approach to bringing generative AI to biotech is based on the premise that DNA is just code. Just like a book is read from start to finish, a gene, which is comprised of DNA, can be read in the same way. The company is attempting to apply predictive training, like large language models, to genes. The idea is to understand the 'grammar' of genomes, which are not random sequences of DNA bases but a highly structured product of 4 billion years of evolution.


The company is working on training a gigantic neural net, a foundation model, to write DNA, similar to how ChatGPT or GPT-4 learned to write English. The hope is that this foundation model, combined with antibody data, could be more effective at drug discovery, designing new crop protection, and other biotech applications.


The company is leveraging its vast collection of genomic data for this neural net training. While there have been early successes, this project is still in the experimental stage. The challenge is to see if this AI can outperform the current state of the art computational tools in generating new DNA designs.


Sources:

https://www.ginkgobioworks.com/

Sources

bottom of page