24/7 BIOPHARMA - issue 1 / October 2024

LOGICA

Experimental data fuels AI models

can refine the model’s predictions, making them increasingly accurate. This iterative process is akin to an ongoing dialogue between the laboratory bench and computational algorithms, where each cycle of feedback sharpens AI’s focus and enhances its predictive power. For instance, data from a failed compound can be as informative about target pathways as a successful one, teaching the AI to navigate the chemical space with greater discernment. The ultimate goal of marrying AI with experimental data is to elevate the success rates of drug discovery programs, a long-standing sore spot in this industry. By leveraging AI’s analytical prowess, researchers can identify the most promising compounds early in the discovery process, prioritise them for development and anticipate potential challenges in safety and efficacy. This approach not only accelerates the pace of discovery, but also allows resources to be allocated more effectively, focusing the attention of sponsors on candidates with the highest likelihood of clinical success. Data-driven success in discovery

The synergy between AI and experimental data is pivotal. The ‘wet science’ data serve as the foundation for training AI models, optimising outputs and improving program success. This symbiotic relationship ensures that AI models are not operating in a vacuum, but are continually refined by real world experimental results, leading to more accurate predictions and optimised strategies [ 4] . Experimental data act as the lifeblood of AI in drug discovery, encompassing a wide range of information – from biochemical interactions and phenotypic responses to pharmacokinetics and toxicology profiles. These data do more than just feed AI models; they shape their architecture, guiding the learning process to reflect biological realities. By training AI models with diverse and comprehensive datasets, we ensure that the insights and predictions they generate are grounded in the complex nuances of human biology. Core of AI modelling

AI’s continuous learning cycle

AI’s practical applications

One of the most compelling aspects of AI in drug discovery is its ability to learn and improve over time. Each new experiment contributes data that

The integration of AI in drug discovery, particularly in rapidly identifying treatments for emerging diseases,

35

TWENTYFOURSEVENBIOPHARMA Issue 1 / October 2024