CRISPR Gene Editing with Machine Learning

Genome editing—which concerns changing or shutting down parts of the genetic code—has long been a goal of molecular biology. One day it may let us fix parts of the genome in a bespoke manner, potentially helping to treat HIV, cancer, and other diseases. In the past decade, the CRISPR system has revolutionized gene editing. Already, it’s used day-to-day to probe biological systems, yielding better understand of mechanisms of disease and enabling drug screens, for example. Separately, the data-driven discipline of machine learning, a sub-discipline of artificial intelligence, is changing our world, from speech and visual understanding, to automated personal assistants. In this talk, I will explain how we combined these two fields, of CRISPR and machine learning, to achieve a more effective way to shut down genes. Specifically, CRISPR can be deployed in many places in the genome; however, some of these places don’t work, and some accidentally wreak havoc elsewhere in the genome even when they do work. We used machine learning to capture regularities of the CRISPR system to predict effective places for attack, while avoiding accidental damage elsewhere in the genome. The resulting tool is now used worldwide, saving researchers both time and money.