Evaluating warfarin dosing models on multiple datasets with a novel software framework and evolutionary optimisation

Warfarin is an effective preventative treatment for arterial and venous thromboembolism, but requires individualised dosing due to its narrow therapeutic range high individual variation. Many machine learning techniques have been demonstrated in this domain. This study evaluated the accuracy of most promising algorithms on International Pharmacogenetics Consortium dataset a novel clinical South African patients. Support vectors linear regression were amongst top performers both datasets performed comparably recent stacked ensemble approaches, whilst neural networks one worst datasets. We also introduced genetic programming automatically optimise model architectures hyperparameters without human guidance. Remarkably, generated models found match performance best hand-crafted by experts. Finally, we present software framework (Warfit-learn) warfarin research. It leverages successful preprocessing, imputation, parallel evaluation, with goal accelerating research making results domain more reproducible.