One of the most powerful features of ConAn is its ability to correlate features of interest or external perturbations with the contact map. This can allow a precise identification of key interactions (between residues or groups of residues) responsible for some macroscopic quantity. One "macroscopic quantity" is time. If we identify parts of our protein that have a significant time drift, that might mean that the reference structure is a poor one, or that our simulation needs to be run longer!The 50 ns equilibrium simulation of ubiquitin, for example, shows the following time correlation:
Oops! There is significant rearrangement of the linker region between β3 and β4 (observed also in the previous section as a "tug of war" between R42 and D58). If we want to faithfully simulate ubiquitin, we would need to simulate "longer" (perhaps several milliseconds as we recently learned! but I digress). This kind of analysis would be very difficult with "fitting" methods, perhaps one could look at time correlations with pricipal components, but here we have direct access to a very large amount of information.
At constant velocity pulling experiments, the very first part of the unfolding trajectory sees a linear increase in force (as a "force ramp"). In that case, this Pearson correlation with time represents the force response of the protein. We examine this for the first 10 ns (of a total of 250) of "unfolding" of ubiquitin in which only minor changes in the contact map can be observed:
Residue pairs in orange get closer together as the external force increases (and could therefore be considered "force-bearing") and blue ones farther apart ("force-compliant"). More sampling than 10ns on a single frame would definitely be helpful, of course!
How about some other macroscopic variable? Let us try to explain the radius of gyration of ubiquitin. If we check how the contact map correlates with this observable, we get:
If we identify the positively correlating interactions, we find out that it is the final loop with everything else and the flexible loop with β3 that are mainly responsible for changes in the radius of gyration. This makes a lot of sense, of course.
Hello, I think CONAN is a wonderful tool. But I do not how to draw the inter-residue cross-correlation (Your paper Figure S1), and I can not find the option.
ReplyDeleteDear visitor, thanks for the kind words. The inter-residue cross-correlation can be obtained with the keyword
DeletePEARSON_INTER yes
This is only available for a symmetric contact map, as far as I know.
Thank you for your reply!~
Delete1. I found the tool will generate a figure named "inter-residue Pearson correlation".
"inter-residue Pearson correlation" is "inter-residue cross-correlation" ? They are a same thing and same mean?
2.The Fig.S2 of the paper, Pearson correlation with radius of gyration for pairwise distances also puzzled me, and I do not how to draw it.
Dear visitor,
Delete1) Yes, those two things are the same. I am sorry if the nomenclature was inconsistent.
2) You are saying that it puzzled you but you are not telling me how it is puzzling you. The idea is pretty simple: you have A(t) = an observable and r_{ij} (t) = a pairwise distance. We can get the correlation between these two time series, and we can repeat this computation many times for all possible r_{ij}.
The option you are looking for is:
PEARSON_OBS obs.txt
See some example input here:
https://github.com/HITS-MBM/conan/blob/master/docs/Example_inputs.md
or in your CONAN distribution (~/conan/example_inputs) for the format.
I hope that helps.
Csaba
I try to further learn the tool again and my problems are solved. Thank you so much for your help!~~
Delete