Zhanglab forum

My name is Daim Saood, and I recently uploaded a set of single protein sequences to the i_Tasser online bioinformatics lab. I am reaching out to seek clarification on the interpretation of the output graphs generated by the deep neural network.

The results include three graphs: a contact map, a distance map, and a hydrogen bond network graph. While I appreciate the detailed information provided, I find myself in need of further clarification to better comprehend the significance of these graphs in the context of my research.

Could you kindly provide some guidance or explanatory notes on how to interpret these graphs effectively? I believe a clearer understanding will greatly contribute to the discussion and analysis section of my research paper.

Hi Daim,

Thank you for using our software, D-I-TASSER.

The figures you asked for are a contact map, distance map, and hydrogen binding network that is used to guide the D-I-TASSER folding simulation (structure prediction). Those spatial restraints are predicted by DeepPotential, AttentionPotential, and AlphaFold2 deep learning methods.

For detailed information about the D-I-TASSER, you can check our CASP15 report published recently.

Zheng, W, Wuyun, Q, Freddolino, PL, Zhang, Y. Integrating deep learning, threading alignments, and a multi-MSA strategy for high-quality protein monomer and complex structure prediction in CASP15. Proteins. 2023; 91(12): 1684-1703. doi:10.1002/prot.26585

(https://onlinelibrary.wiley.com/doi/10.1002/prot.26585)

Briefly, contact means the residue pair (i,j) that has a distance below 8 angstroms. Each black point in the map figure means a predicted contact.
The point in the distance map means a predicted distance between residue pair (i,j), where the distance is colored as a heatmap style with the rainbow color bar from red to blue, corresponding to the distance from 0 to 22.
The definition of the hydrogen network is slightly complex, and you can check this paper (figure2) for details.

Zheng, W, Li, Y, Zhang, C, et al. Protein structure prediction using deep learning distance and hydrogen-bonding restraints in CASP14. Proteins. 2021; 89(12): 1734-1751. https://doi.org/10.1002/prot.26193

Let us know if you have any further questions.

Best Regards
Wei Zheng