DoseTAIlor: A Web-Based Platform for Personalised Tacrolimus Dose Optimisation Across Multi-Centre Populations Using Interpretable AI

Tacrolimus, commonly used after organ transplantation, has a narrow therapeutic index (NTI) and a high risk of dosing errors, with over 50% of patients suffering serious, potentially life-threatening adverse effects. To address this, we developed an interpretable long short-term memory (LSTM) model trained on 14 years of data from 1,774 patients across Spain and the United States. This LSTM predicts tacrolimus levels using clinical data, with approximately 5% mean absolute error, outperforming state-of-the-art models. Addressing data privacy concerns, we demonstrate that synthetic data can replace real data for training without performance loss. Finally, we launched dosetailor.com, a platform providing personalised tacrolimus dosing recommendations for clinicians. Although focused on tacrolimus, this platform forms a foundation for personalising dosage for high-risk, NTI drugs.