Related Experiment Video
Updated: Jul 9, 2025
10:34
Probing RNA Structure with Dimethyl Sulfate Mutational Profiling with Sequencing In Vitro and in Cells
Published on: December 9, 2022
4.2K
AlphaFold predictions are valuable hypotheses and accelerate but do not replace experimental structure determination.
Thomas C Terwilliger1,2, Dorothee Liebschner3, Tristan I Croll4
1New Mexico Consortium, Los Alamos, NM, USA. tterwilliger@newmexicoconsortium.org.
Nature Methods
|November 30, 2023
Summary
Artificial intelligence (AI) protein structure predictions like AlphaFold are powerful tools but vary in accuracy. Experimental validation is crucial for details, especially those not considered by AI models.
Area of Science:
- Structural biology
- Computational biology
- Biophysics
Background:
- AI-driven protein structure prediction, exemplified by AlphaFold, has transformed structural biology.
- However, the accuracy of these predictions is not uniform and does not account for ligands, modifications, or environmental influences.
Purpose of the Study:
- To assess the reliability of AlphaFold predictions by comparing them against experimental crystallographic data.
- To understand the limitations of AI in protein structure prediction.
Main Methods:
- Comparison of AlphaFold-generated protein structure predictions with experimental crystallographic maps.
- Analysis of discrepancies at both global (domain orientation) and local (conformation) scales.
Main Results:
- AlphaFold predictions closely matched experimental maps in numerous instances.
- Significant deviations were observed in some cases, affecting global and local structural features, even for high-confidence predictions.
Conclusions:
- AlphaFold predictions should be viewed as valuable hypotheses requiring experimental verification.
- Prediction confidence levels are critical for interpretation; experimental structure determination is essential for validating details, especially those involving unmodeled interactions.
Related Concept Videos
01:49Conserved Binding Sites
4.2K
Many proteins’ biological role depends on their interactions with their ligands, small molecules that bind to specific locations on the protein known as ligand-binding sites. Ligand-binding sites are often conserved among homologous proteins as these sites are critical for protein function.
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally...
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally...
4.2K
01:22Protein Folding
118.2K
Overview
118.2K
01:13Protein Organization
138.0K
Overview
138.0K