Backed by Google’s AI fund, Gradient Ventures, BenchSci spent 3 years teaching a computer to read biomedical papers like a PhD scientist.
Today, BenchSci is the world's most comprehensive AI-driven biomedical discovery platform. Powered by state-of-the-art machine learning, our AI has read millions of scientific documents. This includes more than 9 million biomedical papers and more than 5.8 million compound specifications.
Using BenchSci, scientists now have the world's biomedical knowledge at their fingertips. In an instant, they can get novel scientific insights, including for materials and methods. These are insights they can trust, derived from millions of successful experiments. This catalyzes their research and increases the velocity of discovery.
BenchSci was born in the birthplace of modern machine learning, Toronto. Its creators are biomedical scientists, computational biologists, and machine learning engineers. They experienced first-hand the expensive inefficiency of data gaps and literature searches. Together, they developed technology to overcome them. This technology uses machine learning to read scientific documents and distill their knowledge. This knowledge is then stored in a graph of relationships. Researchers can query the graph to ask complex questions and receive evidence-based answers.
Just over a year from launch, BenchSci was already powering research in 15 of the top 20 pharmaceutical companies and more than 2,000 academic institutions. Our first application is resolving the antibody crisis. Leveraging our comprehensive biological knowledge graph, BenchSci guides research antibody selection. Researchers can find antibodies by target, technique, and 15 experimental variables.
We're now expanding the questions our knowledge graph can answer. In 2019, our AI will unlock insights into siRNAs, plasmids, and recombinant proteins. Our platform will also scale to encompass more compounds, cell lines, and tissues. And it will branch out to new data sources, including patents.
This evolution will allow us to address broader discovery challenges. These include patient cohort identification and patent decoding. BenchSci will also be capable of guiding experimental design by understanding knowledge about disease pathways, proteins, and genes currently encoded in the world's published scientific literature.