You have probably heard of ChatGPT and DALLE-E, a new class of AI-powered software tools that can create new images or write text. The algorithm brings to life any idea you may have by putting together fragments of what it has previously seen – such as images annotated with meta-descriptions of what they represent – to generate original content from user-defined input. But now generative AI technology is revolutionizing drug discovery. Absci Corporation (Nasdaq: ABSI) is using machine learning to transform the field of antibody therapeutics: Absci has put out a press release today announcing the ability to create new antibodies with the use of generative AI. Their founder and CEO, Sean McClain, will be presenting the news later today at the annual JPM Healthcare Conference happening this week in San Francisco.
Absci has been around since 2011. In less than a decade, the company has transformed from a garage lab into a massive facility with 77,400 square feet of lab space in Vancouver, Washington, and opening additional sites in New York City and Switzerland. Absci has been working with top antibody therapeutics makers in the world and had developed state-of-the-art screening capabilities that allow them to test and validate nearly 3 million unique AI-generated designs each week. In addition to this unprecedented wet lab capability, they have recently added an AI-powered computational platform, called Integrated Drug Creation™ platform, that allows them to analyze a billion molecules computationally per week and to test and validate more than 100,000 of these in the lab.
Similar to the way AI has transformed the world of information, it is now bringing breakthroughs to biology and medicine. Absci recognized this opportunity right away. In 2021 they brought on Joshua Meier from Meta to spearhead the de novo, meaning “from scratch”, protein design using large language models. His unusual job title, the “Chief AI Officer”, gives you a good idea of what his role is at the company and where the field of antibody therapeutics is moving: “There's a totally new paradigm for designing proteins,” Meier says. “We are doing things we thought were impossible even 5 years ago. And today, a programmer can write code that 18-24 months later can save someone’s life.”
Drug discovery has a biological data problem
Just like other generative protein design companies in this space, such as Cradle, Basecamp Research, Arzeda, Biomatter Designs, and Cambrium, Absci is using machine learning to design new proteins from scratch. Proteins are biological machines that run the programs of biology. They perform different functions in our bodies and those functions are specified by the DNA sequence of the gene which encodes that protein. So, to design a protein to perform a particular task, you just need to “write” the code that will make the protein fold in a specific 3D shape to do what it needs to do. The challenge is figuring out the connection between the sequence and function.
Antibodies happen to be some of the most challenging proteins to design. Antibodies are part of our immune system, and they can be biologically programmed to bind to the specific tissues or cell types to fight cancer, for example, or to destroy infecting viruses. Specific binding is notoriously hard to achieve, however, and even the best designs often fail in clinical trials. But Absci has been tackling the antibody design problem for quite some time now. They have helped their clients build better therapies that bind with high specificity, are stable in the body for a long time, and do not generate the immune response that would cause the body to reject the treatment. And now they are taking that knowledge to build their own pipeline with the help of generative AI.
AI-aided protein design requires massive amounts of data to train the models on. The field of antibody development has struggled to provide the high-quality biological data that is needed for good AI models to generate results: “Drug discovery has a biological data problem,” says CEO McClain. Absci has been collecting antibody data over the past decade, which has allowed the company to build its zero-shot generative AI method. “Zero-shot” refers to the process that involves designing antibodies to bind to specific targets without using any training data of other antibodies known to bind those specific targets. What that does is provides more sequence diversity and eliminates bias in de novo designs.