[RASMB] A Complicated Equilibrium Problem

Peter Schuck pschuck at helix.nih.gov
Thu Jan 17 13:24:36 PST 2008


Eric,

provided that the interaction is slow on the time-scale of the SV 
experiment (which I don't know) - the problem of having a broad peak that 
is hypothesized to really contain two separate components is an excellent 
application of the Bayesian extension of c(s) regularization we've 
published recently
http://www.ncbi.nlm.nih.gov/pubmed/17521163?ordinalpos=3&itool=EntrezSystem2.PEntrez.Pubmed.Pubmed_ResultsPanel.Pubmed_RVDocSum
Essentially, rather than subjecting the broad c(s) peaks to a secondary 
analysis to try teasing out relative magnitudes of the underlying 
subpopulations contributing to the peak, you can define upfront the 
Bayesian regularization such that it gives you two baseline-separated peaks 
(if consistent with the raw data) instead of the one broad one.

Note it will never change the quality of fit in order to do that, only use 
the information on your prior expectation to explore the space of possible 
c(s) distributions (i.e. those that fit the data indistinguishably well), 
and give you back the one that is most consistent with your prior 
expectation.  If your expectation is not consistent with the data, the 
result will contradict your expectation.  Otherwise, if it is partially 
consistent, it will add to your expectation additional features essential 
to fit the data (indistinguishably well from the best fit).

In any case, this way you can overcome the 'regularization broadening' and 
you'll get baseline separated peaks for your species, which are then easy 
to integrate.  It is available in the latest version of SEDFIT.

Peter


At 04:00 PM 1/17/2008, you wrote:
>Peter,
>   As far as the c(s) distributions go, it looks as if the s-value of both 
> the dimer and trimer are close to each other; so close that I thought 
> there was only 1 broad peak at the lower concentrations. I can now 
> definitely see that there is a trend of the higher s-value shoulder of 
> that broad peak becoming sharper and more abundant as concentration 
> increases. In short, it does appear as if the s-values only change in 
> height and not value for these two species. With all of this in mind, I 
> will try your suggestions for the ball park estimates.
>    Thank you very much for your time and advice,
>       Eric
>
>Peter Schuck wrote:
>>Hi Eric,
>>
>>if you're able to determine the species concentrations in SV through c(s) 
>>in SEDFIT or by discrete species analysis in SEDPHAT, it must mean that 
>>the interconversion is sufficiently slow such that the oligomers to not 
>>fall apart on the time-scale of sedimentation. If this is true, then the 
>>signals from each oligomer still approximately reflect their equilibrium 
>>populations. In that case, you can simply use the signals of each of the 
>>oligomers, convert into concentration, and plug that into any mass action 
>>law to get the binding constants.
>>
>>With a system as complicated as yours, even if it is pretty clean, I'm 
>>not sure if there is a good chance to do much better. For reasons I've 
>>outlined previously, I don't think that kinetic rate constants will be 
>>sufficiently well determined not to have enormous parameter correlations 
>>and be extremely susceptible from systematic errors arising from slight 
>>inaccuracies in the assumptions regarding buoyancy and purity.
>>
>>On the other hand, even if there's slight dissociation during SV, the 
>>numbers you may be able to get from species analysis may still be pretty 
>>good ball-park estimates. You can find out if that approximation works 
>>well by running your sample at different concentrations. If the s-value 
>>of the peaks for each species appears independent of loading 
>>concentration (i.e. only the relative height changes, but not peak 
>>position), then this should be pretty good.
>>
>>Peter
>>
>>
>>
>>At 02:04 PM 1/14/2008, you wrote:
>>>Hello,
>>>   I have a system that contains four different molecular weight species 
>>> of my protein ( monomer, dimer, trimer, hexamer), all mediated by 
>>> Nickel coordination. I can see this through previous SV experiments, as 
>>> well as the species analysis model in SEDPHAT. I cannot, however, 
>>> determine any dissociation constants using any of the pre-existing 
>>> models in that program. I'm thinking that there might be competing 
>>> monomer-dimer and monomer-trimer-hexamer equilibriums, or some 
>>> combination thereof.
>>>   With all of this in mind, I was wondering if there was a program with 
>>> a model such as this that I might employ, or if it would be possible to 
>>> add my own model to such programs as Heteroanaylsis, Sedphat, etc.
>>>   Thank you for your time and advice in advance,
>>>
>>>     Eric Salgado
>>>     Univ. Cal. San Diego
>>>     Dept. Chemistry and Biochemistry
>>>
>>>_______________________________________________
>>>RASMB mailing list
>>>RASMB at rasmb.bbri.org
>>>http://rasmb.bbri.org/mailman/listinfo/rasmb
>




More information about the RASMB mailing list