HYBRID TOXICOLOGICAL MODELS
During the initial R&D process, a substantial number of potential drugs are designed, but the majority fail the safety assessment. This contributes significantly to the widespread elimination of drug candidates in the development process, resulting in a 10-15 year timeline and billions of dollars in costs for a single drug's development. In the competitive drug development market, factors such as cost, speed, and reliability drive the competition. Our AI-powered technology, a key trend shaping the future of drug development, offers the potential to reduce development costs, hasten time to market, and foster viable innovations.
The AI-driven accelerated safety assessment of drug candidates relies on a hybrid model comprising two systems: an in silico virtual prediction system for adverse effects and wet lab in vitro experimentation, enabling exploration of unconstrained chemical space in contrast to any standalone in-silico model
HYBRID IN SILICO/IN VITRO SCREENING MODEL FOR CARCINOGENIC RISK EVALUATION OF SMALL MOLECULES IN HUMANS
This model is based on short-term wet-lab experiments for chemicals. Subsequently, the carcinogenic potential of chemicals is determined after inputting experimental data into the AI/ML in-silico model. The short-term experimental test-system analyses molecular key events associated with genotoxic and non-genotoxic mechanisms of carcinogenesis. The AI/ML in-silico model is a multitask neural network based on a pre-trained large language model, leveraging information about general human gene expression patterns and interactions. The output of the AI/ML in-silico model is a multiclass classification, categorizing chemicals into “human carcinogen”, “probably human carcinogen”, or “human non-carcinogen''. Additionally, it provides information on the adverse outcome pathway (AOP), revealing the mechanism of action (MOA) and associated key events (KEs), which serves as the cornerstone of a toxicological knowledge framework designed to support chemical risk assessment through mechanistic reasoning.