In 2008, François-Henri Boissel was leading a charmed life. He was a young, successful investment banker working in Tokyo, Japan. And then the market crashed.
He thought of sticking it out, waiting until things improved, but then he remembered a conversation he’d had with his father, Jean-Pierre, in the summer of 2007, and it started gnawing at him.
His father had had a long career in clinical research and had always dreamed of using mathematics to "find truly innovative therapies and dramatically improve patient outcomes," François recalls. The pair had discussed the idea of using mathematical modeling to improve innovation in the pharmaceutical industry, but François had put that idea to the side because he was enjoying the banker's life and the pharmaceutical industry seemed risky. But in 2008, things changed.
"After having spent a number of years analyzing companies through financial statements and market research reports, I was curious to actually get my hands dirty," François says. He was 28, single, and had no kids. "It was the ideal setup to take on serious risk."
The result was Novadiscovery, a startup founded in 2010. In essence, this fledgling company is trying to build a community of virtual patients that scientists and drug companies can use as on-demand digital lab rats. Its goal isn’t to understand how patients interact or behave, but to help curb the costs of discovering new drugs by providing a means of screening potential drug candidates -- and screen them quickly -- using mathematics and intelligent algorithms.
"This is going on before you get anywhere near a person. It’s the first point of research," François told Wired. "It’s a major disruption."
In 2008, when he first left the banking game, François moved back to France and spent the next year brainstorming with his father on how they would try to solve some of the inefficiencies that had plagued the drug pharmaceutical industry for decades. “Our skill-sets were very complementary. [My father] would bring the fundamental science, and I would contribute my business expertise,” François says.
After several months spent ironing out concepts, and recruiting scientists and engineers, Novadiscovery was born. Nova is part of a growing group of companies that are turning to model-based approaches to circumvent some of the inefficiencies that have plagued the pharmaceutical industry in recent years. Pfizer, for example, published a paper in May on the cost benefits of incorporating predictive quantitative modeling into their R&D pipeline.
"This won’t replace [clinical] trials in humans and animals, but it will inform much earlier in the process which [molecules] are worth spending on and which ones should be cut," François says.
Currently, pharmaceutical companies can invest 10 to 15 years and billions of dollars in basic research before they know whether their drug candidate is a dud. There isn’t a reliable way to predict how well a potential drug will work in people so a majority of funding pays for failure. The end result is an industry rife with wasted resources, little innovation, mediocre products and astronomical prices.
Novadiscovery’s approach is an attempt to fix that problem by making biological research more predictive, says Bernard Munos, founder of the InnoThink Center for Research in Biomedical Innovation, a think tank that focuses on innovation in the pharmaceutical industry. In Europe, he says, Nova is helping lead the shift from a business model that relies heavily on serendipity to one driven by mathematics, analytics and computation.
The company is modeling the human body and its diseases using a similar integrative, data-rich approach other researchers have used to device computer models of much simpler organisms like bacteria. The human body is a much more complex system, so the challenge is exponentially greater. Nova's goal is not to simulate every protein or cell in the body, but instead to build a model with enough detail to be able to accurately represent "the fundamental fabric of disease," François says. His objective is clear and specific: to identify new therapies from which patients can benefit.
To do so, Nova scientists and engineers are building a population of virtual patients using real-world data from epidemiological studies, clinical trials, census information and the wealth of disease-related knowledge buried in scientific publications. They've turned this unstructured human data into functional relationships represented by mathematical equations that capture the mechanisms of human diseases. These equations are then turned into computer code that can compute probable outcomes.
Currently, the company is focusing on developing a library of models for cancer, cardiovascular disease, infectious diseases, and immunology, but in theory, their platform could be applied to other conditions, assuming they have the relevant data.
The company has piloted their technology in small proof-of-concept studies with some promising results, but it hasn’t yet applied its algorithms to a large R&D program. That will be its next big challenge.
If Nova's technology –- and others like it -- is shown to work and is more widely adopted, it could bring drug development squarely into the personalized-medicine era. These types of algorithms should eventually be able to take into account individual risk factors like smoking, weight, diet, age, gender, geographic location, and previous medical history. Basically, patients would have a digital version of themselves which clinicians could use to assess possible treatments, lowering the chances patients will suffer from side effects.
François is confident that will indeed be the future. "This feels," he says, "like it’s just the beginning of our journey to accelerate the industry’s transition to a model of sustainable innovation."
