Problems with Dmax

[Iván]

Hi to everyone, 

First of all, thank you very much for all the effort you have put into the design and writing of the tutorials. They really help.
To come to the point, I have some doubts about the dmax calculation with GNOM:

1) On the one hand, although it is a more theoretical question, what exactly does the N in the "Force to 0 at Dmax" option represent?

2) On the other hand, I almost always have doubts when it comes to determining the Dmax. Most of the time the automatic function overestimates the Dmax by a lot and I have to modify it manually. The thing is that, as you mention in the tutorials, the best thing to do is to switch between Y and N when forcing the Dmax to 0. The problem is that I almost never see that the curve reaches zero in a clear way when the N option is activated; it usually meanders above zero and it makes the estimation very difficult, because I always have doubts if I am really overestimating the dmax (I attach some screenshots in the "Choosing DMAX" file)

3) In this same example I considered a dmax of 79 as optimal, but when truncating the data to process the DAMMIF, the P(r) changes and is not optimal. This means that I have to modify the dmax, but I don't know if I have to do it before truncating the data for dammif or when this option is already activated ("Truncation problem" file)

Thank you very much in advance

Best regards,

Iván

[Jesse Hopkins]

Hi Ivan,

The force Dmax to zero does essentially what you would think. If you set it for yes ('Y') then the P(r) function is mathematically constrained to zero at the selected maximum dimension. If you set it to no ('N') then it isn't. Obviously, from a physical point of view the protein should have some distance where there are no more pairs of electrons, and at that maximum distance (Dmax) the P(r) function (the r^2 weighted electron pair distance distribution) will naturally go to zero. 

Problems finding Dmax usually mean there is some amount of polydispersity in your sample. This could be due to aggregation or oligomerization, so that you have some amount of larger object in solution, contributing to a long tail of the P(r) function out to high distances. The other common case is polydispersity in conformation, for example from flexibility in your system. If a protein is adopting some small amount of extended conformations, for example, those contribute to a long tail of the P(r) function out to high distances as well. You'll want to look at the profile in other ways (e.g. Guinier and Kratky plots) to determine which of those you're seeing.

It's okay to accept the tail of the P(r) function to be a tiny bit above zero without forcing. Most people, myself included, would be okay with the P(r) function you show with Dmax=100 and forcing turned off. Your 85 isn't great, it's quite visibly above zero. Probably somewhere between those two there's a point where it first comes down to mostly zero, and that's what I'd chose as the maximum dimension for this particular system. But with something like this there's room for reasonable people to disagree, which is why we say that uncertainty in dmax can be quite large, 10% is a common number people throw around.

For your second issue, I suspect it's because you don't have a good initial Dmax. 79 is obviously too short for that data (you can see this in the first image you sent comparing different values). When you truncate, you're fitting a different region of the profile, and that's changing it. Ideally if you have a good dmax chosen, truncating should have little or no effect on the shape of the P(r) function.

So, my recommendation is try to find a Dmax in the 85-100 range, and that will probably take care of the second issue.

All the best.

- Jesse

----

Jesse Hopkins, PhD

Deputy Director

BioCAT, Sector 18

Advanced Photon Source

--
You received this message because you are subscribed to the Google Groups "BioXTAS RAW" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bioxtas_raw...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bioxtas_raw/c3b70545-5d92-408d-b8f4-dc31326f261bn%40googlegroups.com.

[Jesse Hopkins]

One more thing. I don't know which tutorial you're referring to, but if you haven't already you should check out the general IFT tutorial (rather than the RAW specific one on how to do it using the software):

https://bioxtas-raw.readthedocs.io/en/latest/saxs/saxs_ift.html

----

Jesse Hopkins, PhD

Deputy Director

BioCAT, Sector 18

Advanced Photon Source