[RASMB] MW calculation question

John Sumida jpsumida at u.washington.edu
Tue Aug 13 16:05:35 PDT 2013


Dear John P.

 

I have fitted the entire envelop of boundary scans in SEDANAL.  That is
after determining which scans of the entire dataset showed signal, I fit
those scans.  

 

A typical "Best fit std. dev" value for one sample concentration was ~0.27
for a total signal of 310 fluorescence counts.  I based my goodness of fit
statement on the fact that this was the simplest model that produces the
lowest std dev for the fit. S value for a single concentration was
determined to be 8.2. The global analysis covering the entire range of
concentrations (0.2 - 1.00mg/ml polymer) resulted in an s value of 7.88 for
the full range of boundary scans, and 7.9s for a smaller subset of scans
correspond approximately to the first half of the experiment. 

 

John S.

 

From: John Philo [mailto:jphilo at mailway.com] 
Sent: Tuesday, August 13, 2013 2:34 PM
To: 'John Sumida'; rasmb at rasmb.org
Cc: 'Walter Stafford'
Subject: RE: [RASMB] MW calculation question

 

John S., first I think all responders have been assuming that in your
SEDANAL single-species fit you were fitting the entire run (all the scans
where something is still in the cell), and then also doing the same in your
c(s) fitting. If that is not correct then the picture is different. And
obviously you shouldn't directly compare a single species fit from SEDFIT to
the one from SEDANAL unless you are fitting the same scans. But I think you
must have been fitting the entire run in SEDANAL or you would not have
gotten such a gross over-estimate of D and under-estimate of M.

 

Yes if you really have 20% of other species in the sample then I would not
expect that you would get a good fit of the entire run as a single species,
yet I think that is what you said about the SEDANAL results. Anyway no I was
not proposing that you would try to select a subset of scans to focus only
on the main boundary. Yes you could effectively cut out the
faster-sedimenting minor species by eliminating the early scans from the
fit, but any slowly-sedimenting minor components will always be in the cell
whenever the main boundary still is.

 

The g(s*) procedure I was proposing to effectively ignore the minor
components and focus the analysis on the major species is not something you
can do in SEDANAL---you can only do this in DCDT+, or by porting the g(s*)
distribution into some other software and fitting with an appropriate
function. The basic idea is that after transforming the raw data into
sedimentation coefficient space you can reduce the effects of the slow- or
fast-sedimenting junk in the sample by simply chopping off the upper and
lower ends of the g(s*) distribution and fitting only the middle part of the
main peak (to a single species model). An example of this approach is shown
in Fig. 6 of reference 1 below, and then ref 2 has a more extensive
discussion about it and a validation of the statistical estimates of its
precision via 15 replicates of one experiment.

 

Also FYI the fact that fitting a heterogeneous sample as a single species in
whole boundary analysis leads to significant errors in estimating mass and
diffusion coefficient was discussed in the first paper on whole boundary
analysis with personal computers, ref. 3 from 1994. When a sample of BSA
containing ~10% dimer was fitted as a single species the monomer mass was
low by 20%.

 

John P.

------------

(1) Philo, J. S. (2006). Improved methods for fitting sedimentation
coefficient distributions derived by time-derivative techniques. Anal.
Biochem. 354, 238-246
http://www.jphilo.mailway.com/Anal_Biochem_gofs_fitting_2006.pdf

(2) Philo, J. S. (2011). Limiting the sedimentation coefficient for
sedimentation velocity data analysis: Partial boundary modeling and g(s*)
approaches revisited. Anal. Biochem. 412, 189-202 

(3) Philo, J. S. (1994). Measuring sedimentation, diffusion, and molecular
weights of small molecules by direct fitting of sedimentation velocity
concentration profiles. In: Modern analytical ultracentrifugation.
T.M.Schuster and T.M.Laue, eds. Birkhauser, Boston, pp. 156-170
http://www.jphilo.mailway.com/svedberg.pdf

  _____  

From: John Sumida [mailto:jpsumida at u.washington.edu] 
Sent: Tuesday, August 13, 2013 12:25 PM
To: jphilo at mailway.com; rasmb at rasmb.org
Cc: 'Walter Stafford'
Subject: RE: [RASMB] MW calculation question

Dear John P.  Your comment regarding the effects of fitting the whole set of
boundary data is very helpful.  

 

But just to make sure I am clear on this, I believe what you are saying is
to use dcdt in order to determine the dataset that is consistent with the
major (80%) species observed in sedfit, and then proceed with the single
species model in SEDANAL? 

 

I will also check the single species fit in SEDFIT - I had not tried that
yet - but I think that this also should only use those scans that correspond
to the single species otherwise the fit could prove to be quite poor?

 

Thank you all for all the comments.  They are not only helpful and
informative but encouraging as well.

 

Best regards

John Sumida

Analytical Biopharmacy Core

University of Washington.

 

From: John Philo [mailto:jphilo at mailway.com] 
Sent: Tuesday, August 13, 2013 10:05 AM
To: rasmb at rasmb.org
Cc: 'Walter Stafford'; 'John Sumida'
Subject: RE: [RASMB] MW calculation question

 

John S., I agree with Walter Stafford that heterogeneity is causing the
single-species fit to underestimate the mass, but I don't think there is any
reason to think the primary cause is heterogeneity of monomer mass or
conformation. You said yourself that the c(s) analysis gives one major peak
that is only 80% of the total area. If the sample is only 80% pure, why
would you expect a single-species fit to give the correct MW? If c(s) is
giving one narrow main peak then it is not its ability to emulate multiple
overlapping species that is primarily improving the fit quality relative to
a single-species fit, it is the addition of the 20% of minor species
(presumably with significantly different sedimentation coefficients).  

 

The strength of whole-boundary analysis is that it fits all the scans, but
that means that to get good results it must account for all species that are
present, whether or not they are things you are interested in! That is, a
weakness of whole boundary analysis is that it isn't particularly easy to
exclude impurities, aggregates, fragments or other 'junk' from the data
analysis, and your results simply demonstrate that problem. This is a case
where a quick and simple g(s*) analysis via the dc/dt approach, and then
fitting only the main peak of that distribution (excluding the junk at high
and low sedimentation coefficients) would probably correctly identify the MW
of your 80% major species (but always the best precision should come from
whole boundary analysis).

 

Further, fundamentally your question is an apples vs. oranges comparison.
You say you want to compare the two different analysis programs, but you are
trying to compare two fundamentally different fitting models. You can easily
do a single-species fit in SEDFIT and then compare that to the one from
SEDANAL (and I'm sure the results will be similar).

 

John P.

 

  _____  

From: rasmb-bounces at list.rasmb.org [mailto:rasmb-bounces at list.rasmb.org] On
Behalf Of Walter Stafford
Sent: Monday, August 12, 2013 6:30 PM
To: John Sumida
Cc: <rasmb at rasmb.org>
Subject: Re: [RASMB] MW calculaiton question

I think this was already answered. I think your sample is heterogeneous.
Small variations on composition (I.e molar mass heterogeneity) will cause
peak broadening that will appear as diffusion. If you include enough data in
the fit. It should be apparent in the residuals. C(s) will add enough
components to account for the heterogeneity (albeit with the same f/fo) and
should give better residuals. Sedanal will not because fitting to a single
species (I. E, the wrong model) should not give as good of a fit. The
heterogeneity will cause Sedanal to return a value that is too large and
therefore, a molar mass that is too small. 

Walter Stafford 

wstafford3 at walterstafford.com

 


On Aug 12, 2013, at 20:57, "John Sumida" <jpsumida at u.washington.edu> wrote:

Dear RASMB,

The following was posted on the SEDFIT user list and it has been suggested
that I also post this message to the RASMB list.  Thus I am reposting my
question to RASMB and am including some edits and additional information for
clarity and correctness.

Objective

I am trying to reconcile results obtained using two different approaches,
SEDANAL and SEDFIT.

Observations

In SEDFIT, analysis of sed velocity FDS data returns an s20w of 8.5s and
MW=1.3 MDa for the major peak, comprising >80% of the total loading
concentration, determined in a c(s) fit.  Please note after checking with
our collaborators this value is consistent with the MW returned from DLS and
static light scattering experiments.  Thus the SEDFIT result is consistent
with the DLS and static light scattering data which estimate a MW of 1.1
MDa. 

In SEDANAL, the simplest model necessary to fit the data well was a single
species model.  Using this model, analysis of the same data-set returns an
s20w of 8.8s, similar to the value calculated in SEDFIT, but a MW of 655 kDa
is calculated.

My question:

Why does the value of the MW returned for SEDANAL and SEDFIT differ by a
factor of ~2 when the s20w in each case are similar (8.5s versus 8.8s).  My
purpose here is that I believe the apparent difference being returned from
these parallel analyses is saying something fundamentally important about
the behavior of these materials in buffer and our assumptions thereof.

Background.

1.     <!--[if !supportLists]--><!--[endif]-->Rotor speed was 30 krpm,
vbar=0.917 (measured), solution density = 1.00506 g/cm3 (measured), and
viscosity =0.0100281 Poise (measured).  Temperature = 20oC.

2.     <!--[if !supportLists]--><!--[endif]-->The material being studies is
a polymer micelle with a CMC of 14 micrograms/ml. 

3.     <!--[if !supportLists]--><!--[endif]-->The polymer was run over a
series of concentrations ranging from 0.2 mgs/ml to 1.00 mgs/ml.  

1.     <!--[if !supportLists]--><!--[endif]-->Thus under the conditions of
the experiment I am at least 14 times the CMC at the lowest concentration.

4.     <!--[if !supportLists]--><!--[endif]-->A major peak is observed in
the initial c(s) distribution comprising >80% of the total loading and the
position of this peak (7.7 s-exp; s20w=8.5) does not shift with
concentration.  

5.     <!--[if !supportLists]--><!--[endif]-->From the analysis of raw SV
data, SEDANAL returns an experimental sedimentation coefficient of 7.9s and
a calculated s20w of 8.8s.  

6.     <!--[if !supportLists]--><!--[endif]-->A global analysis in SEDPHAT
was performed over the entire concentration range, transforming the initial
c(s) distribution into a set of 8 discrete species.  

1.     <!--[if !supportLists]--><!--[endif]-->Four of these 8 species
survived a critical chi square analysis suggesting that these four species
were important in retaining the quality of the fit.  

2.     <!--[if !supportLists]--><!--[endif]-->Of these four discrete
species, three were grouped beneath the major peak observed in the initial
c(s) analysis.

3.     <!--[if !supportLists]--><!--[endif]-->The calculated weight averaged
s20w for these three species was 8.1s. 

4.     <!--[if !supportLists]--><!--[endif]-->The  s20w values were checked
and confirmed with a manual calculation, SEDNTERP verstion 1.09, as well as
the calculate "s(20,w) from s(xp) in SEDFIT.

7.     <!--[if !supportLists]--><!--[endif]-->Using the ratio of s/D in the
Svedberg equation and values for diffusion estimated by assuming 655 kDA,
(the MW returned in SEDANAL), I calculate a MW of 655 kDa.  Thus I believe
that the value for diffusion being estimated in SEDANAL is consistent with
the 655 kDA molecular weight.

1.     <!--[if !supportLists]--><!--[endif]-->Estimates of the diffusion
coefficient in SEDFIT and SEDNTERP (and thus presumable also from SEDANAL)
are not very different. 

2.     <!--[if !supportLists]--><!--[endif]-->The diffusion coefficient
estimate in SEDFIT is 3.19E-7 cm2/sec and the value calculated in SEDNTERP
3.67E-7 cm2/sec (Teller approximation).

8.     <!--[if !supportLists]--><!--[endif]-->Using the calculate M(s)
function in SEDFIT and providing the experimental sedimentation coefficient
noted above I need to input an ff0<1 (0.8967) in order to arrive at the
655kDA MW returned by SEDANAL. 

1.     <!--[if !supportLists]--><!--[endif]-->I understand that this is
nonsensical as ff0 cannot be less than 1. 

2.     <!--[if !supportLists]--><!--[endif]-->Notably SEDNTERP version 1.09
also calculates a frictional ratio<1; namely SEDNTERP calculate ffp=0.9449. 

3.     <!--[if !supportLists]--><!--[endif]-->The c(s) analysis in SEDFIT
returns an ff0=1.47.

Thus to summarize:

1.     <!--[if !supportLists]--><!--[endif]-->Relatively similar s20w values
are determined in both SEDFIT and SEDANAL for the same data-set.

1.     <!--[if !supportLists]--><!--[endif]-->8.5s from SEDFIT

2.     <!--[if !supportLists]--><!--[endif]-->8.8s from SEDANAL 

2.     <!--[if !supportLists]--><!--[endif]-->SEDFIT calculates a MW almost
exactly 2 times the value of the MW returned by SEDANAL for the same s20w.

3.     <!--[if !supportLists]--><!--[endif]-->The estimates of diffusion in
both programs are quite similar

4.     <!--[if !supportLists]--><!--[endif]-->The frictional ratio ff0 in
the SEDFIT analysis is 1.47 whereas the calculated ffp in SEDNTERP based on
the observe MW in SEDANAL is <1.  

5.     <!--[if !supportLists]--><!--[endif]-->The frictional ratio
calculated in SEDNTERP, assuming the MW=1.3 MDa, is 1.54

6.     <!--[if !supportLists]--><!--[endif]-->There does appear to be
heterogeneity in the major peak observed in the c(s) analysis.

Thank you in advance for your comments and suggestions.  I apologize for the
length of this post, but in fairness, I am attempting to provide information
that would enable an informed response.

Best regards

John Sumida

University of Washington

Analytical Biopharmacy Core

_______________________________________________
RASMB mailing list
RASMB at list.rasmb.org
http://list.rasmb.org/listinfo.cgi/rasmb-rasmb.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20130813/0dd776a8/attachment.htm>


More information about the RASMB mailing list