[RASMB] error estimates and F-statistics in Sedphat

[Chad Brautigam]

Mark,
I think I see the problem.  You MUST define an extinction coefficient for your protein (this is an easy step to forget).  You can calculate that with ProtParam or SEDNTERP or the calculator of your choice.  I think that should pretty much clear up your problems.
Some other hints:
With absorbance data, don't fit RI noise unless there are certain circumstances like a drifting baseline.
You should float the concentration.  You probably don't know it precisely enough to keep it fixed.
Your c(s) distributions have a significant species at 0.75 S.  I think you should try to account for that by adding an incompetent species.
Your left fitting limit is uncomfortably close to the meniscus.  Scooch that puppy to the right a bit.
You should set the kinetic parameter to something slow (log(koff)=-6 or so).  You are getting baseline separation between your putative monomer and dimer.  That indicates slow kinetics.  In such a case, you should make sure to incubate your proteins for a long period of time at the desired concentration so that there is enough time for equilibrium to be established (assuming that you are diluting from a concentrated stock solution).
Cheers,
Chad



From: rasmb-bounces at rasmb.bbri.org [mailto:rasmb-bounces at rasmb.bbri.org] On Behalf Of Mark Agacan
Sent: Wednesday, April 27, 2011 8:58 AM
To: RASMB at server1.bbri.org
Subject: Re: [RASMB] error estimates and F-statistics in Sedphat

Many Thanks for the very useful replies, as always.  To answer:

1.  SV c(s) analysis showed the samples were relatively clean, and the sedimentation coefficients output from sedfit showed good reproducibility.  Overlaid plots are attached (Slide 1) and show the monomers landing right on top of each other, with dimers also close, through the concentration range 0.50 - 0.75 mg ml-1.  The expected dimerization constant for the interaction should be around low micromolar.
2.  The averages of these S values were used in the sedphat->monomer-dimer self association model.  As the peaks in c(s) showed very good consistency I thought this would at least provide good starting numbers for the true S values and the software would be able to fit from there.
3.  In sedphat: I loaded the data from the three concentrations, set the fit positions and data limits, entered the concentrations in uM and processed as follows:
a.  Execute the Run command.  When this is done without meniscus, bottom, TI and RI boxes ticked, a single line runs the length of the fit limits.  When these same boxes are ticked and the run command only is executed, a more complete set of fits runs across the data range (slide 6).
b.  fit for S1 and S2 with all other parameters fixed.
c.  fit for TI, RI noise, meniscus, bottom, with all other parameters fixed.
d.  Fit for log10Ka12 and log10Koff with all other parameters fixed.  Protein concentration is not fitted at any stage.
e.  move to ML fitting and float a few parameters to see if the fit improves.
f.  return to simplex fitting and use the outlined procedure to examine F-statistics to derive errors for the binding constants.
4.  When the fits are overlaid on-top of the data, I noticed the start of the fit is well below the baseline (Slide 8).  The data has a flat baseline from meniscus to bottom, but the fit runs at an angle, from low at the meniscus to high beside the bottom.  Is there a parameter somewhere that I have not set or checked that can account for this?
I agree that it is almost irrelevant to calculate errors / confidence intervals when the fit is visually so very poor.

In the attached figures there is only one experiment loaded up for speed of processing and clarity, so I did a single experiment fit, but with the other concentrations also loaded in and using global fitting the results / fit looked the same.

I very much appreciate that you are taking time to give me your advice.
Best Wishes,
Mark


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr Mark Agacan, Scientific Officer for the Division of Biological Chemistry and Drug Discovery,
Wellcome Trust Biocentre, College of Life Sciences, University of Dundee, Dundee, DD1 5EH
Tel: +44 1382 386095    Fax: +44 1382 345764    Mobile: 07525 451 117
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


************************************************************
Please consider the environment. Do you really need to print this email?


The University of Dundee is a registered Scottish charity, No: SC015096

________________________________

UT Southwestern Medical Center
The future of medicine, today.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20110427/dbd75182/attachment.htm>