TABLE 1.

Regulator’s considerations and possible regulatory actions by subject of potential interest in the different steps of the drug lifecycle.

Drug lifecycle steps	Data related subjects	Models and applications related subjects	Considerations/possible regulatory actions
A- Drug discovery	Use of chemical and pharmacological big data Genetic association analysis, pathway mapping, molecular docking, and signature profile matching data	Mode of action/AI prediction Ranking of promising drug compounds Drug repurposing	If determinant in the body of evidence: Assessing relevance of data used/ability to check comprehensiveness and relevance models used
B- Non-clinical testing and toxicity prediction	Use of toxicological big data	Specific models for toxicity prediction (ex: RASAR)	Assessing data used: ability to check comprehensiveness and relevance New regulatory standards for toxicity prediction if used in animal full replacement methods Evolution of pharmacopeial monographs
C-Translational and clinical research	Synthetic data	Digital twins	Ability to check relevance of the synthetic data
	EHR, data from various sources (omics, imaging, pathology, etc.) real-world data	Clinical trial design optimization	Checking/assessing data used Relevance of potential regulatory action to be determined
	EHR, use of real-world data	Selection of patients	Expected regulatory adaptations needed
	Use of clinical big-data (imaging, pathology, omics, etc.)	Analysis of clinical trials	Enhancement of classical routine clinical trial analysis Relevance of potential regulatory action to be determined. Use of clinical big-data already in research hospitals (clinical trials)
D- Pharmaceutical manufacturing, QC, QA	Use of in-house firms’ data	Specific models and applications: process design, optimization, in-control advanced process control, used before process or on-site	Potential confidentiality issues. Regulations to be adapted Knowledge of model precision, accuracy, and state-of-the-art. For continuous production or models used in live production control, specific training of inspectors Evolution of pharmacopeial monographs
E- Pharmacovigilance (PV)	Usual and “Non-conventional” big-data use: social networks, medical forums	Early or wider detection of PV signals/detection from more sources and big-data	Validate relevance and usefulness of data Check and test relevance of detections: part of model training. Might be put in routine production soon