Main Topic: A critical evaluation of AF-cluster, a method used to predict the structures of metamorphic proteins with AlphaFold2 (AF2).
Key Points:
- Metamorphic proteins can adopt multiple stable conformations, making their structure prediction challenging.
- AF-cluster, initially presented as a solution, is called into question by Schafer et al. due to:
- Inaccurate predictions compared to random sampling.
- Misidentification of non-metamorphic proteins and erroneous conformation predictions.
- Discrepancies in confidence scores for predicted structures.
- CF-random, a method using ColabFold with random MSAs, is introduced as a more reliable and efficient alternative.
- The study highlights the need for:
- Rigorous validation of computational predictions.
- Complementary approaches for accurate protein structure prediction.
- Transparency in methodology to facilitate further research.
