Leader:
ELIXIR-LU/UNILU
Objectives:
WP4 aims to provide an inventory of artificial intelligence (AI) methods and their use, and the methodologies required to confirm biomedical relevance, test robustness, and validate stratifications. Rigorous validation of stratification results is critical for subsequent clinical research on stratified treatment options.
Description of work:
WP4 will develop guidelines to ensure robustness, reproducibility, and validity of algorithm-driven patient stratification. The guidelines will be based on the mapping and gap analysis from WP2, as well as a workshop held with AI experts.
During the workshop, experts will present, explain and discuss the various machine learning approaches (unsupervised clustering methods for unlabelled data and supervised classification method for labelled training data) capable of identifying compact and distinctive patient strata. The ways in which these approaches could be optimised to ensure biomedical relevance and robustness will be examined in particular. The workshop will lead to a report detailing these recommendations and will include insight on the possibilities and limitations of patient stratification through AI.
WP4 will also identify the needs of main stakeholders (medicines agencies, HTAs, funders) in terms of robustness of stratification. Based on this analysis, the project will propose recommendations on how best to ensure the robustness and reliability of the outcomes of stratification studies.