The Groundbreaking Combination of ML and Biology in Pharma
After taking a hiatus from her biomedical background to co-found Coursera, Daphne Koller has returned with a new and inventive startup.
In 2018, Koller founded Insitro, a drug discovery and development company that seeks to combine biology with machine learning. According to Koller, implementing machine learning will hopefully increase the efficiency and success rate of drug production.
While similar companies focus more on the machine learning aspect of this combination, Insitro has purposefully created a system that equally values the input from biologists and data scientists, redefining the industry.
Insitro's unique addition of a laboratory for biologists in addition to implementing a setup for machine learning has led to the firm seeing early success, already negotiating a deal with the biotech behemoth Gilead to develop tools and drugs for non-alcoholic fatty liver disease (NASH).
Insitro believes that the combination of biologists and data scientists working together separates themselves from other drug development companies. According to Koller, the ability to take two groups that are at the cutting edge of their field and integrate them into one allows for increased efficiency and the development of a better fail-fast model.
The current drug discovery process takes about 15 years of development with a 5% success rate.
The low success rate is explained by the many possible paths of action when producing new drugs. The greater the number of paths, the more likely it is for the company to pick the wrong one.
The unlikelihood of the drug succeeding implies that drug discovery is likely to result in major losses in both time and money spent on an ineffective drug.
Machine learning assists researchers in narrowing down the possible paths of action, ultimately leading to a higher success rate.
Machine learning helps predict human clinical outcomes and aspects of a certain disease based on the data collected on the attributes of the cells being tested.
From this data, predictive models are formed.
However, when compared to other fields, the amount of high-quality data in biology is limited. Fortunately, biologists have created tools to generate the data needed for machine-learning.
Data scientists use this data to improve upon their current machine learning algorithm, creating a better predictive model. This cycle of collecting and using data between biologists and data scientists explains why Insitro’s strategy has yielded positive results in pharma research and development thus far.
These positive advancements in drug development may also help pave the way to a COVID-19 solution. Koller states that Insitro is looking into predicting vaccine efficacy with its models.
With the vast potential of the combination of machine learning and biology, the future seems bright for Insitro and the drug development world.
Written by Zachary Ostrow
Edited by Alexander Fleiss & Gihyen Eom