[RASMB] graphing AUC data

[Walter Stafford]

Hi Tara,
	The choice to use the sequence molecular weight is the 
correct one.  I think that if you had had significant amounts of 
other species present (monomer or aggregates), the plots would have 
exhibited noticeable curvature.  I would say that the error you do 
have is quite small given the potential sources of error. For 
example, an error in v-bar gets nearly tripled in the calculation. Is 
your density correct? An error in radial calibration could also 
contribute. Did you run at other loading concentrations and did you 
get the same results at each concentration? How many times have you 
repeated this run. Do you have an idea what the run-to-run error is? 
You may find that 4.8% is not too far from the error expected.

The value of sigma you used here looks to be about 1.2 (cm^-2); At 
sigmas this low you cannot expect to get much 
resolution/fractionation and, therefore, it becomes extremely 
difficult to fit to a unique model. This is because, at sigmas below 
about 2.0, the three parameters, simga, ln(A1) and delta are highly 
cross correlated. Although you didn't fit for delta, if you force it 
to a particular value, (zero for example), that is in error, the 
other parameters, sigma and ln(A1), will compensate to give a very 
good fit. At sigmas below 1.0 , what you get is something like, but 
not exactly, a weight average for the sample.

My suggestion would be to run this sample at three loading 
concentrations at several speeds at higher sigmas like 2,3 and 4 and 
perform a global fit. The higher sigmas will give you better 
fractionation/resolution between species, and the range of loading 
concentrations will allow you to differentiate between reversible 
interactions and heterogeneity should there be more than one species 
present.

Walter

At 4:22 PM -0700 10/13/04, John Philo wrote:
>Tara,
>
>You are making the graphs correctly, but your own statements 
>about your results show why these lines have different slopes. When 
>you use the sigma calculated by SEDNTERP using the sequence mass for 
>a trimer, then your graph will give a line corresponding to that 
>sequence mass. However, you said that the fit from NONLIN 
>corresponds to a molecular mass that is 4.8% different from the 
>sequence mass for trimer, and one can easily see a 4.8% difference 
>in slope on such a graph. So yes, of course, the line using the 
>NONLIN result goes through the data very well (by definition if it 
>is a good fit), and the line using the sequence mass won't go 
>through the data as well because your data don't actually correspond 
>to the molecular mass for pure trimer.
>
>If you are trying to show the expected slopes that correspond to 
>trimer, dimer, and tetramer then you should use the values based on 
>the known monomer mass. But fundamentally the graph isn't coming out 
>like the published results because your data is telling you the 
>sample is not pure trimer (unless the vbar or density are way off, 
>but presumably you are using the same vbar as the other labs used).
>
>You didn't say whether the NONLIN result is above or below the 
>expected trimer mass. If it is below, it might simply be because at 
>the concentration you used there is significant reversible 
>dissociation to monomer (perhaps the other labs were plotting data 
>taken at a higher concentration, where the fraction monomer is very 
>low). If the concentrations are similar, and your mass is low, I'm 
>going to guess your sample is a mixture of trimer and some denatured 
>'incompetent monomer' that cannot associate to form trimers. On the 
>other hand, if NONLIN says the average mass is above that for 
>trimer, your sample probably contains some aggregate.
>
>'Hope this helps,
>
>John Philo
>Alliance Protein Laboratories
>
>-----Original Message-----
>From: rasmb-admin at server1.bbri.org 
>[mailto:rasmb-admin at server1.bbri.org] On Behalf Of Tara Suntoke
>Sent: Wednesday, October 13, 2004 10:48 AM
>To: rasmb at server1.bbri.org
>Subject: [RASMB] graphing AUC data
>
>Hi all,
>I am fairly new to AUC and have a question about Sednterp and 
>graphing AUC data.
>
>I spun a peptide that is a clean trimer- other people have published 
>the same result for the same peptide.  I think the data is good for 
>the following reasons:  the residuals are random, the data fits best 
>to a single species, and has a square root of variance of 5.9 e-3. 
>In addition, the molecular weight I get from the experiment, (24.8kD 
>for the trimer) is within 4.8% of the calculated MW in Sednterp 
>(based on composition).
>
>I am trying to make a graph of ln(A) vs. r^2, and show that the 
>peptide corresponds to a trimer and not a dimer or tetramer.  In 
>order to graph the lines of the dimer, trimer, and tetramer, I used 
>the equation of a line as follows:
>
>ln(A) = (sigma)(r^2) + ln(A1).  I obtained the constant ln(A1) from 
>the WinNonlin program.
>
>My question is about which value of sigma I should use.  I can use 
>the value of sigma obtained from Nonlin, which is based on the data. 
>Alternatively, I can use the sigma that Sednterp gives me, based on 
>the peptide composition and buffers.  When making these graphs 
>however, the data fits exactly to the line for a trimer which uses 
>the Nonlin sigma, and doesn't exactly follow the slope of the line 
>that uses the Sednterp sigma.  I understand that the sigma obtained 
>from Nonlin comes from the data, and perhaps that is why it fits to 
>the data well.  On the other hand, the sigma from Sednterp is 
>calculated using density and partial specific volumes that are close 
>estimates, but aren't exact. (at least this is my understanding so 
>far...)  So I am wondering which value of sigma is appropriate to 
>use.
>
>I'm not sure whether you will be able to see this at all, but I have 
>enclosed a PDF to show you the different results I get using the 
>different sigma values.  In all the papers that use the same 
>peptide, people report graphs very similar to the lower one, which 
>uses sigma derived from Nonlin.  The data always overlaps the line 
>for a trimer.  However, it isn't clear how they generated those 
>graphs.
>
>I would greatly appreciate any advice you have.
>
>Thanks for your help,
>
>Tara Suntoke


-- 
----------------------------------------------------------
Walter Stafford
mailto:stafford at bbri.org
direct dial:    617-658-7808
receptionist: 617-658-7700
----------------------------------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20041018/aa7f1a32/attachment.htm>