Full atom reconstruction

[JMB-Scripts]

Hi,
After applying Calvados on the test file hnRNPA1S, I encountered issues when trying to regenerate a full-atom model from the MD results. For some reason, the C-alpha atoms in the structured region of the protein appear shifted, preventing pulchra and PD2 from accurately reconstructing the peptide bonds. However, in the flexible region, both pulchra and PD2 seem able to successfully rebuild the backbone. Do you know what might be causing this issue?

Cheers,
JM

[sobuelow]

Hi! This probably has to do with the fact that the algorithm uses the residue centers-of-mass rather than the C-alpha atoms in the coarse-graining of the structured regions (see the paper https://doi.org/10.1002/pro.5172). I believe @fancaoErik has a solution for reconstructing mixed representation coarse-grained structures using cg2all.

[fancaoErik]

Hi! If you are using center-of-mass representation for structured regions, then you should consider using cg2all (https://github.com/huhlim/cg2all) to reconstruct all-atom trajectories. cg2all provides several options based on what representation is used (--cg). Based on our current workflow, ResidueBasedModel should be used to reconstruct multi-domain proteins consisting of both IDRs and folded regions, and the algorithm itself will figure it out how to handle Cα beads of IDRs; CalphaBasedModel should only be used to reconstruct IDPs coarse-grained trajectories.

[JMB-Scripts]

Thanks a lot, i ll check asp