Solving The Life Science Reproducibility Crisis with Machine Learning
How BenchSci is refactoring biomedical research
When C. Glenn Begley and Lee M. Ellis published their research in 2012 demonstrating that 47 of 53 landmark cancer research papers could not be reproduced, the biomedical community opened its eyes to the pervasiveness of what’s now known as the ‘reproducibility crisis’ in life science research. Many published scientific results simply fail to be replicable and as a result, research labs waste hundreds of millions a year in research budgets while critical innovations are left uncovered, and patients left at risk.
Reproducibility issues partly stem from market inefficiencies affecting the use of the most common tool in biomedical research: antibodies — immune proteins that researchers leverage to understand and identify molecules of interest. The market around antibodies suffers from structural information asymmetry between manufacturers and researchers, and this lack of transparent data in the commercial reagent and antibody market makes it tremendously difficult to properly assess reagent quality, specificity and consistency.
About 5 years ago, Dr. Thomas Leung ran directly into this problem while conducting research for his PhD in Epigenetics at the University of Toronto. Sourcing for antibodies was a tedious chore that involved reviewing peer research papers and reference materials for countless hours. Once he did manage to identify the right research antibodies for his work, their quality and validation methods varied so widely batch-to-batch (even from the same vendor) that his experiments, budget and time would all go to waste… repeatedly.
Because of his growing frustration about this problem, Dr. Leung thought: What if there was a platform that instantly sifted through research publication data to enable researchers to find the exact antibody information they need to conduct their research? By combining millions of data points from research papers, such a platform would help researchers understand which experiments and results are reproducible. This would create enormous value for the industry as a whole, by freeing millions of precious hours and hundreds of millions of dollars in research budgets to be better allocated.
But there was a huge challenge. Creating a knowledge graph to validate the roughly 5 million commercially available antibodies that react to human proteins is a herculean task. To build such a platform, the company had to be world-class in three distinct disciplines: sales & marketing, life sciences and machine learning.
As a researcher with a strong foundation in collaborative solution-finding, he knew that building a successful product would mean working with other top-tier co-founders. He set out to recruit Liran Belenzon (CEO), a repeat founder, and David Q. Chen (CTO) & Elvis Wianda (CDO) who brought unparalleled experience in both software engineering, bioinformatics and life sciences.
When we first met the BenchSci team in 2016, I didn’t believe they would be able to build and commercialize such a technologically complex product at scale. A few months later, I was blown away by the speed at which they’ve proven me wrong.
After 2 years of R&D, BenchSci is now able to bring researchers the power of scale in a simple format. Their platform consumes vast amounts of published research papers and unstructured data, when combined with their computer vision and natural language processing algorithms, extracts the most relevant knowledge to be easily searched, sorted, and compared. BenchSci’s technology takes seconds to accomplish what previously took 10 scientists a full year to process. Using BenchSci, researchers can find reliable antibodies on average 24x faster and 75% cheaper than traditional methods. Customers now include 14 pharmaceutical companies (including 7 of the world’s top 10), and 910 academic research institutions (including Harvard, UCLA, Stanford, and MD Anderson) with a pipeline that is growing on a weekly basis.
We believe this compression of time, and simplification of vast sums of data into actionable insight has the potential to reshape research in labs across the world. I am excited to announce that Inovia has partnered with the BenchSci team by leading its Series A financing as we set out to refactor life science research.
