Developing a high-throughput early developability package to support the pipeline and AI model training
01 Sept 2026
Upstream Development
- How should organizations define meaningful developability risk thresholds when there is no clear industry consensus on pass/fail criteria? At what point should a promising candidate be deprioritized or terminated, and how do teams balance pipeline de-risking against the risk of premature candidate attrition?
- How predictive are early developability assays of late-stage outcomes, and how can organizations continuously assess their translational relevance? Are current early-stage assays serving as reliable indicators, or are they primarily empirical proxies that require ongoing validation?
- How should organizations interpret developability red flags when liabilities are highly context-dependent? Can a universal downselection framework be effective, or should risk assessment be tailored based on modality, molecular format, target biology, and clinical indication?
- When developability risks are identified, how should teams determine whether a candidate should be engineered, reformulated, or terminated? To what extent can molecular engineering, formulation strategies, or manufacturing process optimization successfully mitigate identified liabilities before a candidate is deprioritized?
- How can organizations create an effective feedback loop between AI-driven predictions and wet-lab validation? What operational or organizational barriers are limiting access to predictive models, slowing experimental confirmation, and ultimately reducing the speed and quality of AI model retraining?
