Publicação
Explain-dili: a transparent machine learning framework for predicting drug-induced liver injury
| Resumo: | This document presents a data case examining whether predictive analytics can strengthen MediPharm’s early detection of drug-induced liver injury (DILI). Motivated by emerging liver safety signals from a newly launched therapy, the case provides a clinically inspired dataset through which students analyse data quality, predictive modelling, interpretability, and operational decision-making in pharmacovigilance. The teaching notes outline learning objectives, instructor guidance, discussion paths, and common pitfalls. Individual teaching notes extend the case by exploring missingness, model construction and explainability, thresholds, and responsible deployment. Overall, the case demonstrates how predictive analytics can inform safety monitoring while highlighting the governance required for responsible clinical use. |
|---|---|
| Autores principais: | Msolly, Nessim |
| Assunto: | Drug-induced liver injury (DILI) Predictive modelling in healthcare Clinical decision support systems Model explainability (SHAP) Threshold selection and risk stratification Responsible AI deployment |
| Ano: | 2026 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | This document presents a data case examining whether predictive analytics can strengthen MediPharm’s early detection of drug-induced liver injury (DILI). Motivated by emerging liver safety signals from a newly launched therapy, the case provides a clinically inspired dataset through which students analyse data quality, predictive modelling, interpretability, and operational decision-making in pharmacovigilance. The teaching notes outline learning objectives, instructor guidance, discussion paths, and common pitfalls. Individual teaching notes extend the case by exploring missingness, model construction and explainability, thresholds, and responsible deployment. Overall, the case demonstrates how predictive analytics can inform safety monitoring while highlighting the governance required for responsible clinical use. |
|---|