[TALK 13] Structure Prediction and Design using AlphaFold – Sami Chaaban

[TALK 13] Structure Prediction and Design using AlphaFold – Sami Chaaban

by Sami Chaaban

Life & Natural Systems Biology PSALife sciences: general issuesPSAXComputational biology
🎙 Sami Chaaban 👥 10K 📅 February 10, 2026 ⏱ 41 min 👁 413 📄 tutorial 🧭 2026-08-16
Available in: English (current) Français

Keywords

AlphaFoldprotein structurepLDDTPAEColabFold

Summary

Sami Chaaban, a researcher at the MRC Laboratory of Molecular Biology, presents a comprehensive tutorial on using AlphaFold for protein structure prediction and design. He begins by explaining the importance of structure prediction and the historical context, including the CASP competition and the development of AlphaFold2 and AlphaFold3. He then describes the underlying principles, such as multiple sequence alignments and co-evolution, and contrasts AlphaFold with other tools like Rosetta and protein language models. The core of the talk focuses on interpreting AlphaFold outputs: the five predicted structures, the MSA depth, and key confidence metrics like pLDDT and PAE. He provides practical examples from his own research on dynein, illustrating how to use these metrics to assess domain interactions and protein-protein interfaces. He also discusses how to run predictions using ColabFold, both on Google Colab and locally at the LMB, and offers tips on adjusting parameters like recycles and seeds to sample different conformations. Finally, he introduces AlphaFold3, highlighting its ability to handle diverse molecules and its differences from AlphaFold2, and briefly touches on protein design using BindCraft. Throughout, he emphasizes the importance of validating predictions experimentally and being aware of potential pitfalls.

192 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides valuable, practical information for researchers interested in using AlphaFold. It goes beyond a simple overview by offering detailed guidance on interpreting confidence metrics (pLDDT and PAE) and applying them to real-world questions about protein structure and interactions. The speaker’s use of his own research examples adds credibility and demonstrates the utility of the methods. The argumentation is clear and logical, building from basic concepts to more advanced applications. However, the talk is primarily a tutorial rather than a critical evaluation of AlphaFold’s performance, and it does not delve deeply into the limitations or potential biases of the method.

Scientific Rigor, Source Quality, Title Accuracy

The talk is scientifically rigorous, with the speaker clearly knowledgeable about the subject. He accurately explains the principles behind AlphaFold and the interpretation of its outputs. The sources cited in the description are reputable, including the EMBL-EBI AlphaFold training course and the Illustrated AlphaFold blog. The title accurately reflects the content, which covers both structure prediction and design. No comments were provided for analysis.

180 words

Title / Content Match

The title accurately reflects the content, which covers both structure prediction and design using AlphaFold.

Quality & Reliability

8/10

Talk by a researcher at the MRC LMB, providing a practical overview of AlphaFold usage, with clear explanations of confidence metrics and practical tips. The content is based on established methods and personal experience, but lacks formal citations within the talk.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

External References

Contribution & Novelties

This talk provides a practical, user-centric guide to AlphaFold, focusing on the interpretation of confidence metrics (pLDDT and PAE) and offering tips for running predictions and avoiding common pitfalls. It bridges the gap between theoretical knowledge and hands-on application, making it valuable for researchers new to structure prediction. The speaker’s examples from his own research on dynein illustrate the utility of these methods in a real-world context.

Pour aller plus loin :

  • AlphaFold Protein Structure Database — Official database of predicted structures for key organisms.
  • ColabFold — Open-source implementation of AlphaFold for fast predictions.
  • AlphaFold3 — DeepMind’s server for predicting structures with diverse molecules.
  • ModelArchive — Repository for depositing computational models.
  • ChimeraX — Visualization software for structural biology.

118 words

Radar Profile

The radar profile shows high scores in quantity and quality of information, with a slightly lower but still solid technical level. This indicates a well-rounded tutorial that is both informative and accessible, with a strong emphasis on practical application.

Reliability 8/10