Artificial intelligence (AI) is revolutionising drug discovery, enabling scientists to identify novel therapeutic candidates in a fraction of the time it once required. Until now, the main obstacle for researchers has been the ability to quickly screen and characterise large libraries of candidates for an efficacious therapeutic, as well as the inability to spot failed candidates earlier in the process. Consequently, pharmaceutical companies have spent significant development dollars on molecules that will not prove to be suitable therapeutics – prolonging the time-to- market and increasing the cost of life-saving medicines. With the introduction of generative AI and machine learning (ML), it is now possible to select more promising candidates using fewer resources, saving time and money. But how are AI/ML algorithms created and fine-tuned to become reliable methods for candidate selection?
Therapeutic antibodies traditionally undergo five sequential stages leading up to preclinical development. First, the drug’s target is investigated for its relevance in the disease based on primary research. Then the antibody is generated by immunising an animal and growing its immune cells that produce the antibodies. These antibodies are initially tested in the lab, and those that look most promising pass on to the next step as lead candidates. These lead candidates undergo optimisation, a critical stage in which they are honed for biological activity and properties that will impact a future drug’s efficacy, safety, and developability.
To design a promising lead drug candidate for further development, you need to find antibodies with several different properties, all falling within a small range. It’s like searching for the proverbial needle in a haystack, except the haystack is larger than the known universe, and the needle is smaller than a speck of sand. Using traditional drug discovery methods, this can be a long, expensive process with little probability of success. AI-guided design of drug candidates can dramatically speed up both the discovery and the lead optimisation process as well as produce unique, generative variants that could never be found using standard methods.
In recent years, numerous antibody-related databases have emerged, offering valuable resources for training machine- learning models. However, many of these databases can often lack critical information, such as affinity, aggregation parameters, or epitope data.1 The Carterra® high-throughput surface plasmon resonance (HT-SPR) platforms are playing an essential role, both upstream and downstream in the discovery process. Josh Eckman, CEO of Carterra, when asked about the role of Carterra’s HT-SPR technology in AI-driven drug discovery has said, “We believe we can help in this [process] by providing high-resolution binding information on epitope, affinity, and kinetics at the earliest stages of screening.”
The platforms provide reliable data on affinity maturation using its kinetic and epitope software to feed the AI model, as well as verifying the model’s predictions. This validation is then fed back into the model, continually strengthening its intelligence.
AI Drug Discovery in Action
AI promises to revolutionise drug discovery, but advances in drug creation also continue to depend on scalable wet lab technologies to produce and validate biological data at scale. One company leading the way in this combined approach is Absci Corporation. Established in 2011 and headquartered in Vancouver, Washington (United States) the team is using a zero-shot AI approach, which designs antibodies without prior learning on the specific target and are, therefore, generating candidates unlike those found in existing databases. It created a proprietary Integrated Drug CreationTM platform, which combines the data to train, the AI to create, and the wet lab capabilities to validate millions of AI-generated designs. Jens Plassmeier, PhD and Senior Vice President for Biologics Discovery Technologies at Absci, noted that this platform enables their team to develop new therapeutics using the same AI technology celebrated for generating text and images from natural language prompts.
Working from massive biology datasets, generative AI is applied to design optimal drug candidates based on target affinity, safety, manufacturability, and other traits. Absci supports its generative AI designs with its wet lab’s extensive validation capabilities, which includes the Carterra LSA®. David Eavarone, PhD and Director of High-Throughput Screening at Absci, said, “This workflow can take us from AI-designed antibodies to wet lab-validated candidates in as little as six weeks. The quality and scale of wet lab data give us incredible training data, propelling our iterative design-build-test-learn cycle.”