<HTML>
<HEAD>
<TITLE>Re: [RASMB] Sedimentation equilibrium question</TITLE>
</HEAD>
<BODY>
<BLOCKQUOTE>Greetings, everyone<BR>
<BR>
Jo makes a point (below) which is in fact a general one. Assuming your software allows it (and I know that Jo and I both use pro Fit, which makes it easy) then <U>every</U> time you fix a parameter in a fit you should at least give thought to constraining that value within error bounds. Unless it is something like a radial value, where error can be taken as <U>nil</U>. <BR>
<BR>
For a protein, the vbar value is indeed critical in evaluating sigma. Errors in solvent density you can pretty much forget about, so final error in the (1 - vbar* rho) term ought to be about 3 times greater than the uncertainty in (usually computed via SEDNTERP) vbar value you use. I would reckon that ±1% in sigma would be OK for most cases.<BR>
<BR>
That may sound a bit optimistic. But we need to bear in mind that if you do have a system of some polydispersity, then the sigma value tends to be significantly correlated with the (floated) baseline (E) value. So - whilst I would 'give thought' to floating, I am not sure it would always be very positive thought. There is a danger that shoving in one extra parameter into a fit one may merely permit that fit to take advantage of the elbow room given by data error to seek out an optimal chi-squared/whatever.<BR>
<BR>
Arthur<BR>
<BR>
</BLOCKQUOTE><BR>
<BLOCKQUOTE><TT>Dear Sarah,<BR>
<BR>
If I may chip in again, Arthur is basically correct - one usually knows <BR>
the actual monomer mass very accurately from sequence - but there is <BR>
still the problem of the v-bar to use for the residual mass correcting <BR>
for Archimedian buoyancy [i.e. M(1 - v-bar*rho)], which is why it is <BR>
usual to float the mass, at least within bounds.<BR>
One can also get interesting problems, such as one protein I worked on <BR>
which appeared as a dimerising, non-ideal dimer - which would never <BR>
dissociate (under non-denaturing conditions) to monomer. It was a <BR>
couple of years later that my colleagues solved the crystal structure <BR>
and showed that it was a coiled-coil dimer, which would snap together on <BR>
synthesis and not dissociate again!<BR>
<BR>
Jo<BR>
<BR>
Arthur Rowe wrote:<BR>
> Hi Sarah<BR>
><BR>
> There are lots of good suggestion coming your way - so I will not <BR>
> duplicate. But one little point does occur to me. You sound to be <BR>
> floating the (reduced) molecular weight of your protein. Assuming that <BR>
> what you have is a chemically pure system (i.e. there is only one <BR>
> covalently bound entity present (you say it is 84 kDa, purified) then <BR>
> you should be /fixing/ this parameter for the monomeric species, <BR>
> whatever mode of analysis you use.<BR>
><BR>
> Oh, and one little thing more. There are dangers in the "try every <BR>
> model you can think of, select the one which has best <BR>
> residuals/whatever as being _correct_" approach. It has long been <BR>
> appreciated that simple theory shows that if you keep on doing this, <BR>
> then you will always, eventually, come up with something that looks <BR>
> good. Doesn't mean it's right, though. The famous example of this is <BR>
> the search (a few years back) for ways of predicting earthquakes - <BR>
> which eventually came up* with a statistically good correlation <BR>
> between occurrence of terrestrial earthquakes and the phases of a moon <BR>
> of the planet Uranus!<BR>
><BR>
> Kind regards<BR>
><BR>
> Arthur<BR>
> *the authors /did/ report this with tongue firmly in cheek, I should add.<BR>
> -<BR>
> *************************<BR>
> Arthur Rowe<BR>
> Lab at Sutton Bonington<BR>
> tel: +44 115 951 6156<BR>
> fax: +44 115 951 6157<BR>
> *************************<BR>
><BR>
><BR>
> *From: *Tom Laue <tml@cisunix.unh.edu><BR>
> *Organization: *CAMIS-UNH<BR>
> *Reply-To: *Tom.Laue@unh.edu<BR>
> *Date: *Thu, 29 Mar 2007 09:06:47 -0400<BR>
> *To: *Sarah Siegel <ssiegel@scripps.edu><BR>
> *Cc: *rasmb@server1.bbri.org<BR>
> *Subject: *Re: [RASMB] Sedimentation equilibrium question<BR>
><BR>
><BR>
> Hi Sarah-<BR>
> Have you done sedimentation velocity analysis of your samples?<BR>
> What does<BR>
> it show with respect to the concentration dependence of the g(s)<BR>
> patterns? This sort of analysis will tell you whether it is<BR>
> worthwhile<BR>
> doing equilibrium sedimentation by: 1) showing whether there is<BR>
> aggregate, 2) showing if there is significant mass action<BR>
> association/dissociation in the concentration range where you are<BR>
> working and 3) making sure there are no smaller species.<BR>
> With respect to the equilibrium analysis- fit the individual data<BR>
> sets<BR>
> first to a simple model (average molar masses as a function of<BR>
> concentration).Essentially doing what Jo and Ariel suggested.<BR>
> There are<BR>
> several programs to help with this: Sedanal has Biospin built into<BR>
> it,<BR>
> MStar, Segal. The trick with any of these analyses is making sure the<BR>
> signal is equal to the concentration, and it is the presence of any<BR>
> baseline offset (i.e. zero signal does not equal zero concentration)<BR>
> that is the usual concern. For absorbance data, the baseline offset<BR>
> usually is small and can be determined experimentally by pelleting<BR>
> your<BR>
> material (or at least making sure some data are acquired at a rotor<BR>
> speed where the meniscus region is entirely depleted of protein).<BR>
> Interference optics are tougher to deal with since the baseline<BR>
> offset<BR>
> may change with rotor speed (due to window stresses). You may use<BR>
> HeteroAnalysis, fitting to an ideal single component and fit for the<BR>
> baseline offset. In this latter case, M will be between the weight<BR>
> and<BR>
> z- average, typically closer to the latter.<BR>
> See if the individual data sets reveal systematic behavior with<BR>
> respect<BR>
> to concentration and rotor speed. If the M increases with<BR>
> concentration<BR>
> (I would suggest acquiring data over a wider initial concentration<BR>
> range), then you have a mass action association, plus you have<BR>
> some idea<BR>
> (within a decade) of the monomer concentration range where the mass<BR>
> action association is important to protein behavior. If M decreases<BR>
> significantly with increasing rotor speed, you probably have<BR>
> irreversible aggregate (but you would know this clearly from velocity<BR>
> analysis).<BR>
> Best wishes,<BR>
> Tom<BR>
><BR>
> Sarah Siegel wrote:<BR>
> > Hi Everyone,<BR>
> ><BR>
> > I am a beginner in the world of AUC and I'm having a lot of trouble<BR>
> > fitting sedimentation equilibrium data. I was hoping someone could<BR>
> > help me figure out if my problem is in the way I'm fitting or<BR>
> the way<BR>
> > I set up my experiment. I did attend last spring's AUC workshop at<BR>
> > UConn, but this is my first time really running an equilibrium<BR>
> > experiment. The system I have is an 84 kDa purified protein<BR>
> that may<BR>
> > be in equilibrium with dimer, possibly also some larger species.<BR>
> I am<BR>
> > trying to get an idea of what the dominant species present are.<BR>
> I ran<BR>
> > two samples, at 1.0 and 0.5 mg/mL, collecting interference data<BR>
> at a<BR>
> > range of speeds from 8,000 rpm to 20,000 rpm. I watched the<BR>
> approach<BR>
> > to equilibrium using Match in HeteroAnalysis. I have been using<BR>
> > HeteroAnalysis to try to fit the data globally. I have tried<BR>
> fitting<BR>
> > to several models, including ideal and nonideal systems,<BR>
> > monomer-dimer, monomer-trimer, monomer-tetramer, and several three<BR>
> > species equilibria. So far the best fit I have has rms<BR>
> deviations of<BR>
> > 0.027 using a nonideal model, however this fit also tells me the<BR>
> > molecular weight of the protein is 138 kDa, which is larger than<BR>
> > monomer and smaller than dimer, and the residuals do not have a<BR>
> random<BR>
> > pattern. Does anyone have any ideas about another way I should be<BR>
> > fitting this, or other data I should collect? Thanks for your help!<BR>
> ><BR>
> > Sarah<BR>
> ><BR>
> > Sarah Siegel<BR>
> > Graduate Student<BR>
> > The Scripps Research Institute<BR>
> > La Jolla, CA<BR>
> > _______________________________________________<BR>
> > RASMB mailing list<BR>
> > RASMB@rasmb.bbri.org<BR>
> > http://rasmb.bbri.org/mailman/listinfo/rasmb<BR>
> ><BR>
><BR>
> -- <BR>
> Department of Biochemistry and Molecular Biology<BR>
> University of New Hampshire<BR>
> Durham, NH 03824-3544<BR>
> Phone: 603-862-2459<BR>
> FAX: 603-862-0031<BR>
> E-mail: Tom.Laue@unh.edu<BR>
> www.bitc.unh.edu<BR>
> www.camis.unh.edu<BR>
><BR>
> _______________________________________________<BR>
> RASMB mailing list<BR>
> RASMB@rasmb.bbri.org<BR>
> http://rasmb.bbri.org/mailman/listinfo/rasmb<BR>
><BR>
><BR>
><BR>
> This message has been checked for viruses but the contents of an <BR>
> attachment may still contain software viruses, which could damage your <BR>
> computer system: you are advised to perform your own checks. Email <BR>
> communications with the University of Nottingham may be monitored as <BR>
> permitted by UK legislation.<BR>
><BR>
> ------------------------------------------------------------------------<BR>
><BR>
> _______________________________________________<BR>
> RASMB mailing list<BR>
> RASMB@rasmb.bbri.org<BR>
> http://rasmb.bbri.org/mailman/listinfo/rasmb<BR>
> <BR>
<BR>
-- <BR>
Dr P.J.G. Butler,<BR>
MRC Laboratory of Molecular Biology,<BR>
Hills Road, Cambridge, CB2 2QH, UK.<BR>
Tel. +44 (0)1223 402296<BR>
<BR>
</TT></BLOCKQUOTE><TT><BR>
</TT>
</BODY>
<br/>
<p>
This message has been checked for viruses but the contents of an attachment
may still contain software viruses, which could damage your computer system:
you are advised to perform your own checks. Email communications with the
University of Nottingham may be monitored as permitted by UK legislation.
</p>
</HTML>