The course will introduce participants to integrative analysis of multi-omics data with a focus on interpretable factor models. We will start with basic data handling, covering the data formats and common workflows for multi-omics data. After a general introduction to Bayesian factor models, participants will become familiar with MOFA, the de facto factor analysis method for multi-omics data to date, as well as its extension to spatial data, MEFISTO. We will then cover several possibilities to incorporate prior domain knowledge in the analysis. Each day is split into a theory and a practical part. In the theory part, the basic principles behind the methods will be discussed. During the practical part participants will run analyses on small datasets. Time permitting, participants may also analyze their own data. The course targets scientists with prior experience in bioinformatics and single-cell data analysis and a working knowledge of Python.
Day 1
theory
- Introduction to AnnData, MuData, scanpy, data handling
- Introduction to multimodal data integration
practical
- Multimodal integration with Muon (weighted nearest-neighbors), MultiVI (?)
Day 2
theory
- Introduction to Bayesian factor models
- Introduction to MOFA and MEFISTO
practical
- Multimodal integration with the MOFA model
- Multimodal integration of spatial data with the MEFISTO model
Day 3
theory
- Integrating prior information: MuVI, SOFA
practical
- Continuing with MEFISTO/NSF
- Multimodal integration with the MuVI model
- Combining prior information and spatial data
Host of the Event:
Cross Topic for Digital Oncology in cooperation with the Cancer Research Academy & Advanced Training.