Multimodal AI for Systems Biology

Multimodal AI for Systems Biology

Dates

2-6 March 2026

To foster international participation, this course will be held online

Course overview

This course introduces multimodal artificial intelligence (AI) for systems biology. Over five half-days, participants will learn to process, curate, and analyze multimodal datasets, explore fusion paradigms, and apply multimodal AI techniques to integrate diverse biological data types. The program combines lectures with hands-on exercises to build skills and confidence in running multimodal analysis workflows independently, using freely available R and Python tools.

Target Audience

Advanced MS, PhD, and postdoctoral researchers, computational biologists, and data scientists interested in multimodal AI for biological data integration.
Prerequisites: Basic experience with R and/or Python.

Learning outcomes

By the end of this course, participants will be able to:

Import, structure, and manage multimodal datasets
Describe key paradigms for supervised multimodal data integration
Perform data aggregation, transformation, and preprocessing for downstream analysis
Apply causal reasoning frameworks within multimodal AI
Evaluate and quantify uncertainty in multimodal AI models
Analyze multimodal data with appropriate statistical and machine learning methods
Build and execute reproducible multimodal analysis workflows using publicly available tools

Session content

Live sessions will be held daily from 14:00 to 18:00 CET. Between sessions, participants can engage with instructors and peers via Slack (platform details provided before the course).

Monday, March 2

14:00–14:15: Welcome and practicalities
14:15–14:45: Introduction to multimodal AI and systems biology
14:45–15:15: Unsupervised integration of multimodal data
15:15–15:30: Break
15:30–16:00: Supervised and unsupervised multimodal learning — problem setup and data flow
16:00–16:30: Overview of integration frameworks and fusion paradigms
16:30–17:30: Hands-on: Exploring data structures and harmonization in R/Python; unsupervised multimodal fusion
17:30–18:00: Q&A

Tuesday, March 3

14:00–14:15: Recap
14:15–14:45: Early fusion and conventional machine learning approaches
14:45–15:15: Late fusion and ensemble-based frameworks
15:15–15:30: Break
15:30–16:00: Bayesian models for intermediate fusion
16:00–17:30: Hands-on: Implementing various fusion strategies in R/Python
17:30–18:00: Q&A

Wednesday, March 4

14:00–14:15: Recap
14:15–14:45: Introduction to horizontal integration
14:45–15:15: Meta-analysis of multiview datasets
15:15–15:30: Break
15:30–16:00: Multistudy multiview factor models
16:00–16:45: Hands-on: Implementing meta-analysis strategies in R/Python
16:45–17:30: Hands-on: Implementing multistudy multiview factor models in R/Python
17:30–18:00: Q&A

Thursday, March 5

14:00–14:15: Recap
14:15–14:45: Diagonal integration and Multiple Instance Learning (MIL)
14:45–15:15: Attention-based MIL architectures and interpretability
15:15–15:30: Break
15:30–16:15: Hands-on: Implementing deep MIL in R/Python
16:15–17:30: Hands-on: Visualizing attention weights and biological interpretation
17:30–18:00: Q&A and discussion