Key Takeaways
- Gabriele Corso developed open-source software to predict molecular structures and functions, countering restrictive AI models like AlphaFold3 from Google DeepMind.
- Corso’s models, Boltz-1 and Boltz-2, along with generative AI tool BoltzGen, are freely available for modification and use by researchers worldwide.
- His public benefit corporation, Boltz, aims to enhance accessibility to these tools for the pharmaceutical and biotech industries.
Corso’s work stems from his computer science background, focusing on software that predicts the structure and function of complex molecules, including proteins and potential drugs. His open-source initiative allows researchers to utilize these tools to propel advancements in molecular biology.
In 2024, Google DeepMind introduced AlphaFold3, an AI model aimed at understanding biologically significant molecules, claiming enhanced accuracy over traditional physics-based methods. During this time, Corso was working on Boltz-1, a competing model developed through his PhD thesis at MIT. Despite the model’s significant capabilities, AlphaFold3 was encumbered by usage limitations that frustrated researchers, including Corso.
As a response to these limitations, Corso released Boltz-1 as an open-source project, allowing unrestricted access to its code, model weights, data sets, and benchmarks. This initiative sparked further developments, leading Corso and his team to launch Boltz-2 and BoltzGen, a generative AI tool for proteins that interact with molecular targets, under the same open-source licenses.
In January, Corso founded Boltz, a public benefit corporation designed to support the ongoing development of innovative models and associated software and services. This corporation is positioned to serve not only academics but also the pharmaceutical sector. Corso reports that all of the top 20 global pharmaceutical companies, alongside thousands of biotech firms and numerous scientists, currently leverage his open-source models.
Corso emphasizes the mission of Boltz: “Everything we do will be accessible. Some will be open source; some will be available by API. We will continue to make them available as widely as possible.” This commitment aims to democratize access to advanced molecular prediction tools, fostering collaboration and innovation across the scientific community.
The content above is a summary. For more details, see the source article.