M4DI
Methods and Models for Multimodal and Multi-scale
Data Integration
Project context
Project abstract
In this context, the IRPs are further organized in 3 main task forces:
Multi-omics
Multi-omics integration
Different methods will be developed, based on statistics, networks and/or deep learning or combination thereof, both supervised and unsupervised
Prior knowledge
Use of prior knowledge
The methods proposed here will leverage prior knowledge with the objective of better inferring cellular heterogeneity, better represent patients and better define health events.
Health databases
Exploration of health databases
This TaskForce will examine different aspects of data integration using data from health databases, including defining disease phenotypes and trajectories, establishing distributed protocols, or linking phenotypes with genotypes
Biases and interpretation
Biases and interpretation
Developing methods based on artificial intelligence in biomedicine raises several specific concerns regarding potential biases and lack of interpretability of the models. Importantly, as these issues concern the field of artificial intelligence for digital health at large, all the IRPs are associated with this task force.
Benchmarks
Benchmarking of the methods
Benchmarking artificial intelligence and numeric methods is fundamental to compare the performances and features of the different frameworks. This task force will examine the strategies allowing a proper assessment of multimodal data integration approaches, to determine the most effective and efficient.