[ad_1]
Daphne Koller do not mind working hard. She joined the computer science department of Stanford University in 1995, where she spent 18 years full time before co-founding the online education giant Coursera, where she spent the next four years and is remained co-chair until last month. Koller then spent just under two years at the Alphabet Longevity Lab, Calico, as the lead IT manager.
This is where Koller recalled his passion for machine learning to improve human health. She was also reminded of what she did not like – a waste of effort – that the drug development industry – which is slow to understand the power of computational methods to analyze biological data sets – is plagued by years.
To be fair, these calculation methods have also improved a lot more recently. It's no wonder that last year Koller seized the opportunity to create another company, a drug development start-up called Insitro, which has since raised $ 100 million in Series A funding, including of GV, Andreessen Horowitz and Bezos Expeditions. As particular, the company has recently partnered with Gilead Sciences to find drugs to treat liver disease called Nonalcoholic Steatohepatitis (NASH) because of all the human data on the disease that Gilead has accumulated over the years .
Subsequently, Insitro could target even larger epidemics, including Alzheimer's or Type 2 diabetes. Certainly, he has reasons to be optimistic about what he can accomplish. As Koller told a group of delighted attendees at an event organized by this publisher a few days ago: "We are now at a historic moment when a convergence of technologies has emerged pretty much at the same time, making it possible to create large-scale and interesting projects, as well as diseases. Relevant datasets to be produced in biology. In parallel, we see. . . Machine learning technologies that can make sense of this data and bring new ideas that hopefully will cure diseases. "
It sounds like a conversation we've heard in recent years, but coming from Koller, we get the impression that we are finally getting closer, despite the mysteries of human biology. Below are excerpts from Koller's interview with Bloomberg reporter Sarah McBride. You can also watch their conversation below.
Why has Insitro entered into a partnership with Gilead (beyond that, it could be lucrative, with milestones of up to $ 1 billion to successfully develop goals for NASH):
There are fairly broad categories for which our technology is well suited. We're really interested in creating what you might call disease models in a box – places where diseases are complex, where we do not really have a model system, where typical animal models that have been used [for years, including testing on mice] It's just not very efficient – and create these "in vitro" models to generate very large amounts of data that can be interpreted using machine learning.
There is a range of diseases that lend themselves to this type of approach. NASH was one of them. So part of that is the relevance of our technology to this disease, and part of the fact that Gilead was a very good partner because it has a lot of human data from some clinical trials. ran [which give us] access to two complementary data sources. One is what happens to the disease in large human cohorts and the other is what happens when we look at what the disease does in vitro, in the dish, then if we can use what we see in the dish using machine learning to predict what we see in the human.
On how Insitro views data differently from large pharmaceutical companies:
Pharmaceutical companies say, "We have a lot of data." And you say, "What types of data do you have?" And it turns out they have traits, each being stored on a separate worksheet in the Someone else's there is metadata that is not even saved.For them, it's like: "Yes, I've done the experiment and obviously, I'm I recorded what I had to do because it does not make sense to throw it in the trash ", but they do not think it's something you build on a business.
We arrive there in a completely different way. We say, "That's the problem you want to solve. If only we had a model that could tell us the result of this experiment without having to do it because it is expensive, complicated or impossible [because it would involve perturbing a living human’s gene]. 'Machine learning has become a great way to build predictive models if you give it the right data to train it. So we are building data for the sole purpose of forming machine learning models. We think of [these models] like little crystal balls that would allow you to avoid doing [these more expensive or complicated] experiences.
On the impact of the National Institutes of Health's "All of Us" research program, which involves collecting data from a million or more people living in the United States in order to accelerate the research and improve health in part by recording individual differences in lifestyle, environment and biology:
I would say if anything that the United States is a little late to play on this one. A number of national cohorts have already been generated in different countries; The two countries currently the most developed are in Iceland and the United Kingdom, but there is also one in Finland and one in Ireland and even in Estonia, where they have taken a large population in that country and measured their genetics, but also measured. many properties on these people, including blood biomarkers and urinary biomarkers, as well as behavioral and physical aspects and imaging. That's what you have now (in these countries) is a data set that tells you: "Nature has disrupted this gene" and "We see this effect on the human being".
[In the UK, specifically, where they started their program five years ago and recruited 500,000 volunteers who agreed to physical and cognitive and blood pressure testing and images of the brain and the abdomen, among other things] it's an incredibly rich dataset [from which] discoveries are made almost every week.
… This is useful not only primarily for gene therapies, but simply as a means of identifying the targets that really make the difference, as most drugs entering clinical trials fail. And by most, I mean 95%. And most drugs fail because they target the wrong things. They target proteins or genes that do not affect the disease they are supposed to affect. The recent and highly visible failures of Alzheimer's drug trials – in fact, several in a row – are almost certainly due to the fact that the protein they were targeting, called beta-amyloid, was simply not the good causal factor in disease.
What researchers can do now with stem cells that would have been impossible just a few years ago:
[There are now] tools that have created not only large amounts of data, but also large amounts of biologically relevant data. So we had the habit of doing experiments on cancer cell lines. . . but this is not a very relevant model for the disease. Today, we can take a small sample of skin cells and use the so-called Yamanaka factor to reprogram these cells to stem cells, which are the cells that exist effectively in the uterus. And these cells are able to differentiate into neural cells, liver cells or heart cells. These are very relevant to the disease because they represent human biology; You can now take these cells from patients and healthy people and see if there are any differences in their appearance.
Readers, we could cite more here, but we strongly suggest watching the conversation with Koller. If you use this text as a starting point, you will want to start listening about 13 minutes later. This is certainly worth the time to hear what she has to say, including about cystic fibrosis, spinal muscular dystrophy in babies and why the "mouse models" on which we have long relied to a lot of diseases seemingly ubiquitous really, really bad. "I hope you enjoy it.
[ad_2]
Source link