Overview

In order to grow a deeper understanding of our newly designed protein (YebF-endo-β-1,4-glucanase), it is critical to predict the folding stability and catalytic function. Since no experimental data exist, we turned to computational modeling for protein prediction. We used AlphaFold 3 to run prediction models of the protein to determine whether the protein will fold properly.

Prediction Result

The model achieved a pTM score of 0.67. While the protein sequence itself contains two functional parts, YebF and endo-β-1,4-glucanase, this suggests that the model correctly folded both domains, but isn't sure about how the two domains are relatively placed relative to each other, probably due to the flexible linker between the two domains. This can also be proved through the high plDDT score (>90) in most places besides the parts that connect them, which have a relatively low plDDT score.

The Predicted Aligned Error (PAE) heatmap further proved this interpretation. The graph shows two dark green boxes (low error) connected with a little dark green line. This is expected for a construct connected with a flexible linker.

Conclusion

This model supports the preceding wet lab steps. We can clearly see that both domains are predicted to fold properly with a flexible linker connected. For future plans, it is best to test the expressed protein for further research about the protein structure to gain more insights about its functions.

Abramson, J., Adler, J., Dunger, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500 (2024). https://doi.org/10.1038/s41586-024-07487-w