Developing FDA-approved drugs involves multiple rounds of molecule optimization for improved safety and efficacy, typically relying on resource-intensive methods that hinder smaller research organizations. To address this, a plug-and-play platform is being introduced to facilitate the creation of more effective and safer drugs, broadening access for promising candidates in the market. Ascent Bio utilizes AI to offer a highly portable, predictive, and user-friendly solution for molecule design, contrasting with complex traditional computational chemistry tools. This accessible platform accommodates users of varying chemical expertise and is poised to benefit underserved biotech companies by significantly speeding up molecular design processes. The technology is based on deep learning, proven effective across diverse molecule datasets, and continuously improved by their expert team with expanded pretrained models and curated datasets.
What is the problem?
To become a FDA-approved drug, a molecule that has been found to have activity against a therapeutic target must undergo many rounds of molecule optimization to improve key properties such as activity and safety. After many conversations with potential customers, investors and industry experts, we’ve learned that conventional approaches to molecule design require a combination of iterative screening and complex in-silico design software. In practice, this process requires immense resources and expertise that can preclude the pipelines of smaller research organizations. Moreover molecules that eventually pass this process and enter the clinic fail >50% of the time due to challenges of efficacy or toxicity. This challenge of molecule optimization needs to be solved in order for more effective and safer drugs to be able to enter the market. With their powerful, plug-and-play platform, they aim to enable better molecules and to allow more organizations to move their promising candidates forward.
What is their solution?
Ascent Bio's unique insight is that AI allows for an unprecedented level of portability, predictive performance, and ease-of-use to be integrated into the molecule design process. Unlike traditional computational chemistry tools that require a high-level of expertise to use effectively, Ascent Bio will be able to serve users with varying chemical expertise. This ability complemented with the speed of their platform will be transformative for many underserved biotech companies with molecular design needs. Their technology is based on years of research developing deep learning approaches to learning rich embeddings of molecules for downstream tasks. They've observed that their approach works for a variety of molecule datasets and they have published several papers examining this as well as submitted a patent application. Leveraging the expertise on their team, they will continually improve the underlying technology of their platform and add to their large set of pretrained models and internally curated data.
