Although somewhat of an exaggeration, the title of this article falls in line with the thinking of Insitro’s founder, Daphne Koller. In a recent interview with Bloomberg, Koller said that with most drugs “we do not understand why the work”. Koller found that pharmaceutical companies were not storing adequate data after the completion of an experiment.
Koller explains that her company, Insitro, intends to build data using machine-learning. The goal is to produce the results of an experiment without having to actually conduct the experiment. As we are aware, experiments are costly, complicated, and in some cases even impossible to conduct.
Machine-learning is a fascinating artificial intelligence (AI) technique. It enables systems to learn and improve without being explicitly programmed. The focus is on developing computer programs that will access data and then use that data to “learn for themselves”. It relies on inputting correct data to train the model. If done right, it’s an opportunity to develop novel insights that lead to curing a disease.
Koller explains that she is in the business of building data in order to train machine learning models. She said to think of the models in terms of reducing the expense of experiments.
The drug company, Insitro, founded one year ago by Koller, is well on its way to accomplishing its goals having already raised one hundred million dollars through investors including Jeff Bezos.
Disease-In-A-Dish Models
Koller expressed interest in using her company’s technology in several broad categories such as:
- Approaching complex diseases
- Assisting where there is a lack of an adequate model system
- Providing alternatives in research where animal models used for testing were not effective
The plan is to use in vitro (in a dish) models to accumulate data, which can then be interpreted using machine-learning. Koller likes to use the term “disease-in-a-dish” to describe this process.
A Wide Range of Diseases
Insitro recently partnered with the $84 billion biotech company, Gilead Sciences, in search of a drug therapy for NASH (nonalcoholic steatohepatitis), an increasingly common liver disorder. Gilead provides a major step up due to the abundance of human data it has garnered through the years.
Koller explained that there are many diseases that fit in with their approach– she cites NASH as a good example. She gives her company credit for having the necessary technology, and also gives credit to Gilead for amassing human data from various clinical trials.
The partnership gives them access to two data sources. One source follows the disease by studying human cohorts defined as a “set of people followed over a period of time.”
The second source analyzes the reaction of the disease in vitro. The technologies are then combined, and the data is interpreted. This creates novel insights that can possibly lead to disease cures.
Aiming at Larger Targets
Koller has her eyes on even bigger targets such as Type 2 diabetes or Alzheimer’s. She points out that about 95% of drugs tested in clinical trials fail. Koller believes that this may be due to focusing on the wrong targets, such as genes or proteins that might not even be one of the factors causing the disease.
She points to the recent failures of Alzheimer’s trials and suggests that the failures were due to focusing on the wrong target.
Where Are We Now?
Koller’s answer is that we now have tools that produce an enormous amount of biological data. In the past scientists worked on cancer cell lines which are not relevant to a disease model.
Now scientists can take a skin cell sample and reprogram the cells to stem cell status. Stem cells exist primarily in the womb, so this innovation could be helpful to researchers.
Further, these cells can transform themselves into liver, cardiac or neural cells, which are relevant to many diseases. They can be taken from both healthy people and from patients to see if there is any difference in appearance.