Artificial Intelligence (AI) has been making strides in the field of drug discovery, and DeepMind’s AI model AlphaFold has made significant contributions. In 2020, AlphaFold managed to predict the structures of almost the entire human genome, a groundbreaking achievement that allows a better understanding of protein activity and their potential role in diseases. This is key to accelerating drug innovation.
A recent study published in Chemical Science has successfully applied AlphaFold in the early stages of drug discovery. A global research team including members from the University of Toronto, Stanford University, and Insilico Medicine, used the AI program to identify a potential treatment option for hepatocellular carcinoma (HCC), a common form of liver cancer.
HCC is a major global health issue, responsible for 75% of all liver cancer cases. Despite advances in treatments, significant unmet medical needs remain. Using an AI-driven platform called PandaOmics, the research team analyzed extensive datasets to identify and rank potential therapeutic targets for HCC. Cyclin-dependent kinase 20 (CDK20), a protein with a strong association with HCC, appeared to be a good candidate for AI-powered drug discovery.
CDK20 plays important roles in cell cycle regulation and is overexpressed in various cancers, including HCC. Because of this overexpression and its link to tumor progression, CDK20 is considered an advantageous therapeutic target. The team utilized AI programs like Chemistry42 and AlphaFold to create novel inhibitors without requiring 3D experimental structures.
The PandaOmics platform was used to detect potential targets for HCC, focusing on proteins with structures predicted by AlphaFold2. Targets were selected based on their druggability by small molecules, novelty, and absence from recent clinical trials or existing drugs. After various tests and assays, compounds were examined at different concentrations and cell viability was measured following incubation.
Insilico Medicine integrated AlphaFold’s protein structure predictions into their Pharma.AI platform, using PandaOmics for target detection and Chemistry42 for molecule production. A target pathway for HCC and a hit molecule were identified within 30 days without needing a determined structure. This was followed by AI-led compound optimization and the discovery of a more potent inhibitor, a process that emphasizes AI’s significant influence on speeding up drug discovery.
AlphaFold’s predicted protein structures aided in the swift discovery of CDK20 inhibitors through the combined use of AI in drug discovery. Seven compounds were initially synthesized and evaluated. The subsequent rounds of AI-directed compound creation resulted in an improved inhibitor with enhanced binding affinity and potent CDK20 kinase inhibition. This compound exhibited selective anti-proliferative effects in the HCC cells that overexpressed CDK20, suggesting it could be a viable therapeutic agent. Ongoing work includes further enhancement and detailed analysis of the drug’s properties, underlining AlphaFold’s transformative role in expediting novel drug discovery.