[RASMB] Precision of c(s) method

[Peter Schuck]

Hi Matt,

I agree with what was said before, but would like to add an aspect.

Physically, the information about oligomers resides in the slope of the 
solution plateaus.  This can be addressed with very high precision, because 
there's a lot of data points in those plateaus.  One can easily simulate 
noisy data and ask what fraction of oligomer could be detected in the 
best-case scenario, assuming a perfect experiment.  Other people have done 
that, too; I did this for a simulated antibody trimer, assuming a 
signal/noise ratio of 200:1 (i.e. 0.005 fringes noise and 1 fringe loading 
concentration), and found that 0.2% should be detectable.  For something 
that sediments closer to the leading edge of the monomer boundary, it might 
get more difficult, but that would depend also on how well you can 
characterize the monomer (maybe that can be done separately).  What I would 
recommend is to use the SEDFIT integration tool and do a Monte-Carlo 
statistics on the integrated values.  Alternatively, I would probably also 
try the SEDPHAT hybrid discrete/continuous model, describe the monomer as a 
discrete species (perhaps predetermined or constrained in a global fit 
including data from pure monomer), and describing the larger material as a 
continuous section of c(s) [local to each data set].

It is a valid question whether or not in a real experiment you can reach 
the theoretical limit, but my feeling from talking to several colleagues is 
that with careful experimentation you should get close, provided also that 
the sample is behaving hydrodynamically ideal.  (Clearly, you should not 
rely on the details of the c(s) analysis if it doesn't give a good fit.)

If you're asking separately how much dimer, trimer, or tetramer etc. 
exists, this is more difficult to answer than the question "how much 
material sediments in a certain s-range", or "how much total oligomers are 
there".  The latter question can be addressed with much higher precision, 
since by integrating over a large s-range (which would exclude the monomer) 
you essentially are not relying on fitting individual boundaries into the 
slightly sloping plateaus, but instead essentially only measuring the total 
increase of the plateaus (which corresponds to the loading concentration of 
large material).   This is what the integral of c(s) over the s-range of 
large material gives you.  This is essentially equivalent to integrating 
the ls-g*(s), except that with the c(s) you may be able to perhaps set the 
range of interest closer to the monomer, and also that you will be able to 
use a larger data basis by including scans ranging from the initial partial 
depletion at the meniscus until very late where the trailing edge of the 
monomer boundary has disappeared (which will help characterizing the monomer).

Regards,
Peter





At 04:41 PM 3/23/2006, you wrote:
>Content-class: urn:content-classes:message
>Content-Type: multipart/alternative;
>         boundary="----_=_NextPart_001_01C64EC2.9075B0DA"
>
>Hi folks.
>
>            I was wondering if people could give their views on the degree 
> of precision that can be expected from the c(s) method, or other methods 
> for fitting SV data, for assessing the degree of aggregate formation in a 
> protein sample. (Monoclonal antibodies, in particular.) The FDA is 
> starting to ask for "orthogonal methods" to back up SEC data on aggregate 
> formation. Is it reasonable to expect that SV can give the same degree of 
> reproducibility when dealing with, say, 1 to 5% aggregated species? Has 
> anybody done any systematic determinations of the precision of the 
> method? (Subjective impressions are also welcome!)
>
>            Thanks,
>
>            Matt
>
>Matthew Parker, Ph.D.
>Analytical and Pharmaceutical Sciences
>Immunogen, Inc.
>128 Sydney St.
>Cambridge, MA 02139
>(617) 444-2028
>(617) 995-2544 fax
><mailto:matthew.parker at immunogen.com>matthew.parker at immunogen.com
>
>_______________________________________________
>RASMB mailing list
>RASMB at rasmb.bbri.org
>http://rasmb.bbri.org/mailman/listinfo/rasmb
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20060324/54bf0abe/attachment-0001.html>