Multi-Omics Data -Session 8- eXplainable Artificial Intelligence (XAI) for predictive purposes

6 Views

  • Thanks! Share it with your friends!

    You disliked this video. Thanks for the feedback!

RockSinCity
Published
Multi-Omics Data Integration in Human Exposome Studies - Session 8 - - workshop series - coordinated by ISGlobal (https://www.isglobal.org/en/) and the ATHLETE Project (https://athleteproject.eu/)

Session 7 is titled: "Multi-omics and eXplainable Artificial Intelligence (XAI) for predictive purposes", led by Dr. Augusto Anguita, a biological data scientist specialised in the analysis of complex epidemiological datasets such as those composed of clinical, omics, biochemical, and environmental data. At ISGlobal he is an investigator of the EU-H2020 ATHLETE exposome project. His main technical skills include a strong statistical, programming and data visualisation background, with special emphasis on the use of machine learning models for multi-omics data analysis.

Talk outline:

This talk will cover machine learning strategies for leveraging multi-omics data with predictive purposes in epidemiological studies. For that, a real showcase of clinical outcome prediction from multimodal molecular data in a prospective cohort will be presented. This talk will put special emphasis on the interpretability of generated predictive models, explaining how to use post-hoc explainers for the full exploitation of machine learning algorithms and multi-omic information. The talk will be structured as follows:
· Introduction to the real data showcase “Insulin Resistance Prediction in Paediatric Populations”
· Tips for multi-omics data processing and ML model selection and construction
· Use of SHAP values (SHapley Additive exPlanations) for the extraction of global and local explanations
· Use of SHAP values for the study of interactions
Sign in or sign up to post comments.
Be the first to comment