Hands-on Tutorials

Tutorial

Reliable Agentic AI for Life Sciences: Foundations to Research Workflows

---

Tutor: Rizwan Qureshi

Agentic AI extends traditional AI systems by enabling models to act, reason, and execute tasks. Unlike standard chatbots, these systems can write code, interact with tools, retrieve knowledge, and perform multi-step workflows. This tutorial introduces how such systems can support data analysis, reproducible research, and scientific discovery in life sciences.

Tutorial

Graph Learning in Practice: Foundations, GNNs, and Applications

---

Tutor: Tabitha Bishenghui Tao

This tutorial introduces graph learning in a simple and practical way. We begin with a short review of MLPs, CNNs, and RNNs, and explain why graph data needs a different model. Students will learn basic graph concepts, node embeddings, Graph Neural Networks, and GNN training. The tutorial also includes three hands-on examples: Iris with GCN, Cora with GAT, and MUTAG with GIN. No previous experience with graph learning is required.

Download tutorial materials

Tutorial

Omics Data Analysis: From Single-Cell to Spatial Multi-Omics Integration

---

Tutor: Le Ou-Yang

Omics experiments simultaneously measure thousands of molecular features, including transcripts, chromatin-accessible regions, and proteins, across individual cells or tissue locations. These datasets are high-dimensional, sparse, and noisy, making feature-by-feature interpretation impractical.

This tutorial introduces a complete omics data analysis workflow, from data organization and modality-specific preprocessing to dimensionality reduction, clustering, spatial-domain identification, multi-omics integration, and interpretable topic modeling.

Tutorial

Reliable and Interpretable AI for Medical Images and Molecular Data

---

Tutor: Debby Dan Wang

AI systems for medical images and molecular data can produce useful predictions, but predictive performance alone does not establish that their outputs are reliable or understandable. This tutorial develops a shared perspective across both modalities, examining how data representations, uncertainty, interpretability, failure analysis, and reproducibility shape responsible model evaluation. Through accessible examples and guided discussion, participants will learn to question model confidence, assess explanations critically, and identify the additional evidence needed before models can be trusted in biomedical settings. The tutorial is designed for a mixed audience and is not tied to a single dataset or methodology.