Evotec’s 2025 publication by Pitt et al. explores the real-world deployment of AI and machine learning (ML) in drug discovery. It highlights how AI/ML tools are integrated across the drug discovery pipeline - from chemical space exploration and compound design to protein modeling and safety assessment.

The publication demonstrates the value of AI in enhancing decision-making, designing novel compounds, predicting key properties, and optimizing drug discovery workflows. Techniques such as deep learning-based chemical representations, QSAR/QSPR modeling, active learning, synthetic tractability prediction, and high-throughput transcriptomics-based safety assessments are presented as critical tools.

The "Design-Decide-Make-Test-Learn" (D2MTL) framework is introduced which integrates AI decision-making and feedback loops into traditional drug design cycles.

• AI/ML enhances compound design, property prediction, and virtual screening through tools like autoencoders, QSAR models, and generative design.
• Protein modeling benefits from AlphaFold and related tools, improving structure prediction and drug-target interaction insights.
• Active learning and Bayesian optimization help prioritize compounds efficiently, reducing experimental burden.
• Synthetic feasibility and safety are increasingly assessed using AI-driven retrosynthesis and omics-based toxicity prediction respectively.