Using artificial intelligence to discover therapeutic antibodies

Using artificial intelligence to discover therapeutic antibodies

Using artificial intelligence to discover therapeutic antibodies

Former EMBL staff scientist founds a start-up – DenovAI – for broader, faster and cheaper antibody discovery using advanced machine learning and computational biophysics.

Michael Jendrusch, Kashif Sadiq, Jan Korbel
The EMBL scientists who developed the AI-driven technology to design proteins that DenovAI will harness. From left to right: Michael Jendrusch, Kashif Sadiq, Jan Korbel. ©Kinga Lubowiecka / EMBL.

Antibody’s ability to recognise and mark foreign and harmful molecules and organisms allows therapeutic use of antibodies, including treating or preventing diseases such as cancer, as well as auto-immune and metabolic diseases. But while therapeutic antibodies offer great potential, selecting a promising candidate from billions of potential antibody sequences is laborious, expensive, and, in many cases, ineffective in identifying functional antibodies. To address this challenge, former EMBL staff scientist Kashif Sadiq recently founded a start-up company: DenovAI Biotech. The AI-driven technology to design proteins that DenovAI will harness has been developed by EMBL scientists Michael Jendrusch, Kashif Sadiq and Jan Korbel. It has been licensed to the company by EMBL’s wholly owned commercial subsidiary: EMBL Enterprise Management Technology Transfer GmbH (EMBLEM). Jendrusch will also contribute to the company’s endeavours while completing his PhD at EMBL Heidelberg.
The company will create a platform in the field of de novo protein design, a computational approach to design proteins from scratch, rather than using a known structure. The company will primarily develop an AI-powered biophysics solution that can discover potential antibodies and small protein biologics and suggest which of these could be used therapeutically. In the near future, it may also provide capabilities beyond pharmaceutical biologics, including diagnostics, enzyme and biomaterial design.  DenovAI will build on recent advances in protein structure prediction, artificial intelligence algorithms, computational molecular biophysics techniques, and increased availability of experimentally determined antigen-antibody structures.

Investment from Pfizer, AstraZeneca, Merck, and Teva
Through DenovAI, Sadiq aims to combine AI and biophysics to discover the sequences and structures of antibodies that can recognise and bind strongly to any protein antigen. “This type of approach has not been taken before,” said Sadiq. “We have seen major advances in the field of therapeutic antibodies, from increased antibody library sizes to function-oriented discovery, but the process of developing new drugs is still incredibly slow, vastly expensive, and inefficient.  With the support of AION Labs and its partners, we hope to develop a cutting-edge solution that will disrupt the whole field, cutting discovery timelines from months to days. This could dramatically broaden the scope of antibody therapy to many more diseases.” DenovAI is the second startup to be formed by Israel-based AION Labs, which creates startups through a unique innovation model powered by BioMed X.  After identifying specific industry R&D challenges, it carries out a global talent search for scientist founders. DenovAI is supported by investment from leading pharmaceutical companies Pfizer, AstraZeneca, Merck, and Teva, with close support from Amazon Web Services (AWS) and additional financial backing from the Israel Innovation Authority and the Israel Biotech Fund.