At BenchSci, our mission is to exponentially increase the speed and quality of life-saving research by empowering scientists with the world's most advanced biomedical artificial intelligence.
Founded In
2015 by Tom Leung, David Q. Chen, Elvis Wianda, and Liran Belenzon
|
Headquarters
Toronto, Ontario
|
Employees
310
|
Years Training AI
5
|
Papers analyzed
14,102,565
|
Reagent products analyzed
64.6 million
|
Reagent use cases identified
75 million
|
Platform Users
over 50,000
|
Health is our most precious asset, and life scientists help protect it. Thanks to public health and pharmaceutical breakthroughs, life expectancy has doubled since 1900. Today, we take treatments like antibiotics, vaccines, insulin, and chemotherapy for granted. But they all took effort by thousands of scientists over decades to research and develop.
Unfortunately, the productivity of this research and development is declining. In the past decade, pharmaceutical companies have doubled their R&D spend. But the number of drugs approved has stayed the same. This means it costs much more to develop each approved drug. In 2010, it cost $1.2 billion to bring a new drug to market. In 2018, this increased to $2.2 billion. That's an 83% increase. Meanwhile, revenue per drug has declined. The result is that R&D returns have fallen by 80% since 2010, to 1.9% in 2018. This is unsustainable, and threatens continued health advances.
A significant contributor to this productivity crisis is preclinical research waste. Preclinical research is the stage in R&D before human trials. Waste at this stage not only increases drug discovery and development costs. It also increases the time before a new treatment can benefit humans. How much waste is there? One study estimates that half of preclinical research spending is waste. This amounts to $28.2 billion in the US alone each year—up to $48.6 billion globally.
The good news is that we know where the waste is. And we know how to solve it. About 36.1% of preclinical research waste is in reagents, for example. Reagents are materials such as antibodies, recombinant proteins, and RNAi used in life science experiments. When an experiment uses the wrong reagents, it may fail or produce unreliable results. Scientists spend a lot of time and money to search for and test reagents to avoid this. Even then, failure is common. When it happens, scientists have to repeat experiments. This all contributes to making drug discovery and development slower and more expensive.
At BenchSci, this problem wasn't theoretical. Our founder, Tom Leung, experienced it first-hand in 2015 while researching cancer. An inappropriate antibody caused him to lose rare patient samples. Could machine learning prevent this in future? And help him choose the right antibodies? He contacted machine learning and data experts at the University of Toronto to find out. These experts, David Q. Chen and Elvis Wianda, joined him to solve the problem. Our fourth cofounder, Liran Belenzon, met the team at the Creative Destruction Lab. Together, they built, tested, and validated a solution to preclinical research waste. The solution used advances in deep learning to teach a computer how to read and think like a PhD biologist.
BenchSci launched in July 2017 with our first application, AI-Assisted Antibody Selection. It helped scientists select the appropriate antibodies faster, reducing experimental failure. How? First, we collect relevant open- and closed-access scientific papers and product catalogs. Second, we extract relevant data points from them with proprietary machine learning models. Third, we build relationships between the data points with proprietary bioinformatics ontologies. Finally, we make the results searchable in an intuitive interface.
The response to AI-Assisted Antibody Selection exceeded our loftiest ambitions. Within two years, more than 3,600 research institutions and 15 of the top 20 pharma companies used it. More than 31,000 scientists began relying on BenchSci for their experiments. Based on this response, and the impact of our work, we earned the trust of investors. This includes Gradient Ventures, Google's AI fund.
Today, we're building on our success with antibodies to address other critical reagents—including recombinant proteins, siRNAs, and CRISPR tools—as well as model systems such as animal models and cell lines. We're also developing other machine learning-powered applications to reduce preclinical research waste. In everything we do, we're guided by our mission and hope for the future. By helping scientists be more successful, we can improve R&D productivity. And that means continued good health for all of us.