Investors Flock to AI-based Drug Discovery
The application of convolutional neural network technology continues to expand beyond machine vision use cases to the red-hot drug discovery market.
The application of AI to new drug development has moved in fits and starts, including IBM’s (NYSE: IBM) decision last year to pull the plug on its Watson AI software for pharmaceutical research. Lately, investors are warming to new small molecule drug discovery efforts, including recent efforts aimed at developing therapies for COVID-19.
The latest example comes from Atomwise, a San Francisco-based company that claims to have developed the first convolutional neural network for new drug discovery. Atomwise announced a hefty $123 million Series B funding round this week, bringing its investment total to more than $174 million.
The oversubscribed round was led by B Capital Group and Sanabil Investments, and included returning investors Data Collective VC, BV Investment Partners, Tencent, Y Combinator, Dolby Ventures and AME Cloud Ventures. Two unidentified “global insurance companies” also chipped in.
Atomwise said it would use the funds to scale its AI platform while expanding its drug discovery partnerships with leading researchers via its flagship “AtomNet” platform. The goal is commercializing its drug discoveries while ramping production via partnerships with Eli Lilly and Co. (NYSE: LLY), Bayer AG (OTCMKTS: BAYRY), China’s Hansoh Pharmaceuticals (HKG: 3692) and other large drug manufacturers. It is also working with several biotechnology companies, and has so far signed deals valued at $5.5 billion.
Abraham Heifets, CEO and co-founder of Atomwise, claimed AtomNet has so far succeeded “in finding small molecule hits for more undruggable targets than any other AI drug discovery platform.”
Among the hurdles faced in AI-based drug discovery is the difficulty of accessing emerging technologies like convolutional neural networks. CNN technology can help researchers analyze large volumes of experimental data, including drug target sites and disease categories.
The company claims its deep learning technology for “structure-based” small molecule drug discovery has so far generated more than 16 billion molecules for virtual screening. Along with drug target sites, the platform also generates protein classes and homology models—atom-level models of target protein structures.
Atomwise has so far been awarded 19 patents for its drug discovery technologies. Among its 285 current drug discovery partnerships are 15 with university researchers pursuing “broad-spectrum” therapies for COVID-19.
Along with accelerating pharmaceutical research, the computational drug discovery platform is addressing “biology problems previously believed to be unsolvable by researchers and delivering that capability to everyone from academics to big pharma,” said Raj Ganguly, B Capital Group’s co-founder and managing partner.