Non-standard proteins in the lenses of AlphaFold3 - case study of amyloids

Alicja W. Wojciechowska,Jakub W. Wojciechowski,Malgorzata Kotulska

DOI: https://doi.org/10.1101/2024.07.09.602655

2024-10-05

Abstract:The recent release of AlphaFold3 raises a question about its powers and limitations. Here, we analyze the potential of AlphaFold3 for correct reproduction of amyloid structures, which are an example of multimeric proteins with low representation in protein structure databases, which may also be characterized by polymorphism. We show that AlphaFold3 is capable of producing amyloid-like assemblies that have significant similarity to experimental structures (TM-score>0.5), although its results are impacted by the number of monomers forming the predicted fibril and a protein of choice. AlphaFold3 produces structurally diverse models of some amyloid proteins, which could reflect their polymorphism observed in nature. We hypothesize that the lower emphasis on multiple sequence analysis (MSA) in AlphaFold3 improves the results quality, since for this class of proteins sequence homology may be misleading in their structural similarity. However, the structural landscape obtained from the modeling does not reflect the real one governed by thermodynamics. Finally, AlphaFold3 enables for the first time, structural modeling of fibril-like structures to a certain extent, possibly including their polymorphic nature. Still individual benchmarking is necessary for optimal modeling.

Bioinformatics

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve

This paper aims to evaluate the capabilities and limitations of the newly released AlphaFold3 in predicting the structures of amyloids. Amyloids are a class of polymeric proteins that are underrepresented in the protein structure database and exhibit polymorphic characteristics. Specifically, the paper focuses on the following aspects:

1. **AlphaFold3's ability to predict amyloid structures**:
- Investigate whether AlphaFold3 can accurately reproduce the amyloid structures observed in experiments.
- Analyze the similarity between the amyloid models generated by AlphaFold3 and the experimental structures.

2. **The impact of polymer length on prediction results**:
- Explore the effect of different numbers of monomers on the prediction results.
- Evaluate how the diversity and quality of the predictions change with the length of the polymers.

3. **AlphaFold3's performance in handling polymorphic amyloids**:
- Examine whether AlphaFold3 can capture the polymorphic characteristics of amyloids.
- Compare the prediction results of pathological and functional amyloids to explore the differences between them.

4. **Comparison between AlphaFold3 and AlphaFold2**:
- Compare the performance differences between AlphaFold3 and AlphaFold2 in predicting amyloid structures.
- Analyze the improvements in AlphaFold3 in reducing dependency on multiple sequence alignments (MSA) and their impact on prediction results.

5. **Adherence to physicochemical rules**:
- Check whether the structures generated by AlphaFold3 conform to physicochemical laws, such as energy distribution and thermodynamic stability.

Through these studies, the paper hopes to provide new insights into the structural prediction of amyloids and evaluate the strengths and weaknesses of AlphaFold3 in handling these complex proteins.