PDA Technical Glossary

A documented program that provides a high level of scientific assurance that a manufacturing process will reliably produce acceptable product. The proof of validation is obtained through rational experimental design and the evaluation of data, preferably beginning from the process development phase and continuing through the commercial production phase. (TR01) A documented program that provides a high degree of assurance that a specific process, method, or system will consistently produce a result meeting predetermined acceptance criteria. (TR26) (TR57) (TR70) (TR74) Documented testing, performed under highly controlled conditions, which demonstrates that a process consistently produces a result that meets predetermined acceptance a result that meets predetermined acceptance criteria. Used to test processes, methods, and systems for which conditions can be controlled in the real world (i.e., after completion of testing, when the process is in use). Transportation processes can be qualified but not validated; in the real world, it is not possible to exert control over all parameters that could affect the transportation process (e.g., weather, customs, traffic delays, mechanical failures, etc.). (TR58) A documented program that provides a high de­gree of assurance that a specific process, method, or system will consistently produce a result meeting predetermined acceptance criteria. (Note: Vali­dation is a lifecycle program which may include development and qualification activities for one or more elements/systems that form a process.) (TR54-5) A documented program that provides a high degree of assurance that a specific process, method, or system will consistently produce a result meeting pre-determined acceptance criteria (17). (TR60-3)

A separate subset of documented and traceable labeled data used to evaluate the performance of machine learning models (distinct from the broader concept of validation as it relates to ensuring the consistency, reliability, and compliance of manufacturing processes). This data may include independent datasets collected from different sources or time periods, which are used to assess the model's performance on unseen data. Through rigorous validation of machine learning models using validation data, pharmaceutical companies can identify and mitigate potential biases, overfitting, and other sources of error, ensuring the reliability and effectiveness of predictive models in supporting drug discovery, development, and regulatory decision-making. As best practice, validation data should be documented and traceable.