GeoStaS and Residue Cross Correlation: Clarifications needed

[George Tzotzos]

The difference between the two is adequately described. However, some issues need to be clarified:

1. The correlation matrix cm ← dccm(modes) is calculated over all modes. This implies that the high-frequency modes dominate the slow modes. If this is correct, it would be difficult to identify domain motions from the cross-correlation matrix.
2. GeoStaS, gs <- geostas(modes, k=2), uses the “slowest” 5 modes by default (I think). The GeoStaS results would be therefore incompatible with the cross-correlation matrix calculated over all modes.
3. The key of the GeoStaS heatmap. An interpretation is needed. I assume that red corresponds to correlated motions and white to anti-correlated ones.

‌

[Mahnaz Tehrani]

@George Tzotzos

Good point George, it’s also my question as well. I need more information on how to interpret the AMSM plot. Do you know how the color in residue number axes on the AMSM plot has been defined? And how to visualize pdb of AMSM result with domain assignment as shown here:

Grantlab / bio3d-geostas — Bitbucket

Thank you

Bests

Zahra

[George Tzotzos]

Hi Mahnaz,

Point 1. The key of the GeoStaS heatmap. Have a look at J. Romanowska, K. S. Nowinski and J. Trylska, J Chem Theory Comput 2012 Vol. 8 Issue 8 Pages 2588-99. Accession Number: 26592104 DOI: 10.1021/ct300206j

The key of Figure 2 of this paper ranges from 0.0 (dark blue - anticorrelated motions) to 1.0 (yellow - correlated motions). In the Bio3D implementation of GeoStaS, I assume, that white is used instead of dark blue and orange instead of yellow.

Point 2. The following works for me

pdb ← read.pdb(“mypdb”)

modes ← nma.(pdb)

gs ← geostas(modes, k=2) %% finds 2 domains

mktrj(modes, mode=7, chain=gs$grps) %%% should be able to visualise the two domains in pymol or similar

Hope this helps

George

[Mahnaz Tehrani]

@George Tzotzos

Thank you so much. I will check it our. so far this command given in the tutorial link did not work for me.

# write pdb with domain assignment write.pdb(pdb, chain=gs$atomgrps, xyz=xyz)

Bests

Zahra

[Mahnaz Tehrani]

@George Tzotzos

Dear George, I have done couple of test runs and I come up with more questions, can you please send me your email address to share results if it is possible.

I appreciate your help/consideration

Bests

Zahra

[George Tzotzos]

[e-mail]