About Workshop
Artificial Intelligence has transformed protein structure prediction through models such as AlphaFold, enabling rapid advances in structural biology and drug discovery. However, ensuring the reliability, confidence, and biological validity of AI-generated protein structures remains a critical challenge.
This 3-day international workshop introduces participants to practical methods for validating and interpreting AI-predicted protein structures using structural bioinformatics tools, public protein databases, Explainable AI (XAI), and Google Colab. Through hands-on sessions, participants will learn to assess prediction confidence, evaluate structural quality, and develop trustworthy AI workflows for protein research.
Aim
To provide participants with comprehensive knowledge of trustworthy AI in protein structure prediction by integrating structural bioinformatics, machine learning, confidence assessment, validation techniques, and Explainable AI to improve the reliability and interpretability of computational protein models for biological research and drug discovery.
What Participants Will Learn
- Understand the principles of AI-driven protein structure prediction.
- Explore AlphaFold confidence metrics and structural quality assessment.
- Retrieve and analyze protein structures from public databases.
- Perform structural validation using widely used bioinformatics tools.
- Compare experimentally determined and AI-predicted protein structures.
- Apply Explainable AI (XAI) methods to interpret AI-based biological predictions.
- Understand uncertainty estimation and failure modes in protein AI.
- Design reproducible validation workflows for structural bioinformatics research.
- Explore industrial applications of trustworthy AI in biotechnology and pharmaceutical research.
Structure
📅 Day 1: Foundations of AI-Driven Protein Structure Prediction and Structural Validation
- Introduction to protein structure prediction and structural bioinformatics
- Evolution of AI in protein structure prediction
- AlphaFold, RoseTTAFold, and modern protein foundation models
- Protein Data Bank (PDB) and AlphaFold Protein Structure Database
- Protein structure representation and confidence metrics (pLDDT, PAE)
- Structural quality assessment and validation concepts
- Biological interpretation of predicted protein structures
- Challenges and limitations of AI-generated protein models
Hands-on 1: Exploring AI-Predicted Protein Structures
Participants will retrieve experimentally determined and AI-predicted protein structures from the Protein Data Bank (PDB) and AlphaFold Protein Structure Database, visualize protein structures using Mol* or PyMOL, compare confidence scores (pLDDT and PAE), and identify high-confidence and low-confidence structural regions.Hands-on 2: Basic Protein Structure Validation
Participants will use publicly available structural validation tools to examine stereochemical quality, structural consistency, residue confidence, and basic geometric validation of protein models and generate a preliminary validation report.📅 Day 2: Structural Validation, Explainable AI, and Confidence Assessment
- Principles of trustworthy AI in structural biology
- Confidence estimation and uncertainty quantification
- Common failure modes of AI protein prediction
- Structural comparison using RMSD and TM-score
- Explainable AI (XAI) concepts for biological prediction
- Interpreting feature importance in AI-based structural models
- Benchmarking AI predictions against experimental structures
- Best practices for reproducible protein AI validation
Hands-on 1: Comparing Experimental and AI-Predicted Protein Structures
Participants will use Google Colab to compare experimentally solved protein structures with AI-predicted models by calculating structural similarity metrics such as RMSD and TM-score, identifying structural deviations, and interpreting model confidence.Hands-on 2: Explainable AI for Protein Prediction Assessment
Participants will explore simplified Explainable AI (XAI) techniques to interpret machine learning predictions using protein structural features. They will visualize feature importance, analyze confidence estimates, and investigate potential sources of prediction uncertainty.📅 Day 3: Trustworthy Protein AI for Drug Discovery and Future Perspectives
- AI-assisted functional annotation of proteins
- Integrating protein structures with biological pathways and interaction networks
- AI in protein engineering and therapeutic target discovery
- Trustworthy AI in pharmaceutical research and regulatory science
- Foundation Models and Large Language Models (LLMs) in structural biology
- FAIR principles, reproducibility, and responsible AI
- Emerging trends: Protein language models, multimodal biological AI, autonomous protein design, digital biology, and AI-guided drug discovery
Hands-on 1: Protein Function Prediction and Biological Interpretation
Participants will analyze an AI-predicted protein structure alongside protein annotation resources such as UniProt, InterPro, and Pfam to infer protein function, conserved domains, and biological significance.Hands-on 2: Building a Trustworthy Protein AI Validation Workflow
Participants will develop a complete validation workflow by integrating protein structure retrieval, confidence assessment, structural comparison, functional annotation, and AI-assisted interpretation into a reproducible analysis pipeline using Google Colab and publicly available bioinformatics resources.Important Dates
Registration Ends
4:30 pm
Workshop Dates
2026-10-05
5:00 pm
5:00 pm
What You Will Gain
Outcomes
- Understand AI-based protein structure prediction technologies and their limitations.
- Retrieve, visualize, and analyze protein structures from public repositories.
- Assess prediction confidence using established structural validation metrics.
- Compare AI-generated and experimentally determined protein structures.
- Apply Explainable AI methods to improve biological interpretation and model transparency.
- Design reproducible validation workflows for protein AI research.
- Evaluate the reliability of AI-generated protein models before downstream biological experiments.
- Explore emerging applications of trustworthy AI in structural biology, biotechnology, and pharmaceutical research.