Program

Tutorials

Hands-on research-agent training plus a ProVR session. Course outlines are below.

Tutorials complement the main scientific program with practical, guided work. Session I runs on the afternoon of October 17; Session II and ProVR run on the afternoon of October 18.

Training Course: Building Research Agents in Practice (1)

Hands-on development of research agents — from RNA sequence and transcriptome analysis to building personalized knowledge bases.

Instructor: Dechao Bu (Associate Investigator, Institute of Computing Technology, Chinese Academy of Sciences)
When: Day 2 · 17:20–18:20
Where: E3 Lecture Hall

Outline

Topic 1: Building Agents for RNA Sequence and Transcriptome Analysis
Session 1: Based on user-input RNA sequences and instructions, the agent autonomously invokes tools to analyze sequence features and motifs.
Session 2: Mine expression profile data from colorectal cancer transcriptome sequencing, autonomously invoking tools to screen differentially expressed genes and analyze functional pathways.
Topic 2: Building Personalized Knowledge Bases
Session 1: Based on one’s own research direction — for example, osteoarthritis — download literature and build a personal knowledge base.
Session 2: Using various prompts with the knowledge base, summarize literature threads, expand research ideas, and generate new hypotheses.

Training Course: Building Research Agents in Practice (2)

Hands-on development of research agents — from protein sequence, structure, and function analysis to multi-agent systems for complex research tasks.

Instructor: Yang Yang (Associate Professor, Chongqing University of Posts and Telecommunications; Assistant Investigator, Academy of Military Sciences Innovation Institute)
When: Day 3 · 15:45–16:45
Where: E3 Lecture Hall

Outline

Topic 1: Building Agents for Protein Sequence, Structure, and Function Analysis
Session 1: Given user-provided candidate protein sequences and validation requirements, the agent autonomously invokes tools to localize catalytic motifs and stratify the candidates.
Session 2: For candidate antibiotic-resistance enzymes, the agent performs batch structure folding and molecular dynamics simulations, autonomously scheduling GPU resources to assess stability and calibrate against natural-enzyme benchmarks.
Topic 2: Building Multi-Agent Systems for Complex Research Tasks
Session 1: For complex research tasks, a commander agent coordinates multiple specialized agents working in parallel, enabling automatic decomposition of the analysis workflow and synthesis of the results.
Session 2: In response to reviewer concerns, the agent autonomously designs and launches supplementary computational experiments on GPU servers, closing the evidence loop with real data support.

Tutorial: ProVR — an AI+VR platform for protein design

This tutorial introduces how ProVR can be used for interactive protein design, local structural reconstruction, and protein–protein docking in an immersive 3D environment. Participants will work directly with protein structures, redesign local regions, refine structural models, and further optimize protein interfaces after docking.

Instructor: Danfeng Yang (Engineer, Zhongyuan Artificial Intelligence Industry Technology Research Institute); Jianquan Zhao (Engineer, Institute of Computing Technology, Chinese Academy of Sciences)
When: Day 3 · 16:45–17:15
Where: E3 Lecture Hall

Outline

Topic 1: Interactive Protein Design, Local Reconstruction, and Structural Refinement
Session 1: Interactive Design and Refinement of Existing Protein Structures
Starting from an existing protein structure, participants will inspect the local molecular environment and directly redesign selected regions in 3D. The session will cover residue mutation, side-chain adjustment, loop editing, and manipulation of secondary-structure elements. Participants will also learn how these design operations are recorded as actual changes to the underlying molecular structure, which can then be further edited, refined, undone, or saved.
Session 2: Local Structural Reconstruction with Fragments and Computational Tools
Participants will reconstruct selected protein regions using natural structural fragments. They will define the region to be replaced, use spatial guide points and secondary-structure preferences to describe the desired geometry, search for suitable fragments, and adjust candidate fragments directly in 3D. The reconstructed structure can then be refined with OpenMM. When needed, ProteinMPNN can be used for sequence redesign and AlphaFold 3 for structure prediction. The predicted structures can then be loaded back into ProVR for comparison and further modification.
Topic 2: Interactive Protein–Protein Docking and Interface Engineering
Session 1: Direct Protein–Protein Docking in 3D
Participants will directly manipulate two proteins in 3D—for example, holding an antigen in one hand and an antibody in the other—and bring them together to explore possible binding poses. They can freely move and rotate either protein while observing how the binding interface changes in real time. ProVR provides information on intermolecular distances, contacts, interface residues, and steric clashes, helping participants adjust the relative position and orientation of the two proteins and compare different docking arrangements.
Session 2: Editing and Refining the Docking Interface
After obtaining a suitable docking pose, participants will continue working directly on the binding interface. They can select and mutate interface residues, drag or adjust side chains, and modify nearby loops or other local structural regions. Local refinement can then be applied to improve interface packing and reduce unfavorable contacts. Participants can repeatedly adjust the docking pose, inspect the interface, modify local residues or structures, and refine the complex until a satisfactory model is obtained. The final complex can then be saved or exported for further structural or computational analysis.

Instructor and room details may still be refined in the final agenda.