AI in Clinical Applications (Theory)

AI in Pharmacokinetics & Dose Optimization Review of pharmacokinetic parameters (Cmax, Tmax, AUC, clearance) Modeling concentration-time relationships: Linear regression in PK data modeling, multiple regression for dose adjustment based on patient variables (age, weight, renal function) Interpretation of regression coefficients in clinical context Error metrics (RMSE, R²) in PK modeling Limitations of linear modeling in nonlinear pharmacokinetics

AI in Drug Safety & Pharmacovigilance Overview of pharmacovigilance systems Structure and data formats from real systems (e.g. FAERS, EudraVigilance) Conceptual overview of frequency analysis and signal detection Logistic regression for ADR risk prediction Confusion matrix and clinical performance metrics Sensitivity, specificity, precision, recall Bias and confounding in observational datasets

AI in Personalized Medicine & Risk Stratification Concept of precision medicine Patient covariates and therapeutic response Logistic regression for disease risk prediction Classification of responders vs non-responders Evaluation metrics in healthcare prediction Ethical implications of predictive modeling

AI in Clinical Decision Support & Real-World Data Structure of Electronic Health Records (EHR) AI in Clinical Decision Support Systems (CDSS) Regression models for outcome prediction Classification models for risk scoring Real-world data analytics Limitations of AI in clinical environments Accountability and interpretability

Guided Supervised Learning Project Pharmaceutical and Clinical Applications Students will undertake a guided supervised learning project (individually or in groups) applying regression or classification models to pharmaceutical or clinical datasets. The project should include the following steps: Identification and definition of a relevant pharmaceutical or clinical problem Selection of an appropriate publicly available dataset Identification of predictor variables and outcome variables Data preprocessing and preparation for analysis Application of suitable supervised learning methods (e.g., linear regression or logistic regression) Evaluation of model performance using appropriate metrics Interpretation of results in the context of pharmaceutical or clinical relevance Discussion of limitations, potential biases, and ethical considerations Presentation of findings to faculty mentors or peers. Suggested project topics (not limited to): Dissolution rate prediction, tablet hardness prediction, stability degradation modelling, quality control batch failure prediction, assay variability modelling, impurity prediction, moisture impact on formulation stability, coating thickness prediction, solubility enhancement modelling, tablet defect classification, adverse drug reaction (ADR) prediction, therapeutic response prediction using patient datasets, hospital readmission risk prediction, dose requirement prediction using pharmacokinetic data, diabetes risk prediction from health markers, antibiotic treatment success prediction, ICU stay duration prediction, medication adherence analysis, and disease severity classification.