[RASMB] precision limit in quantifying protein aggregates by sedimentation velocity with Sedfit

[John Philo]

Yi-Ming and RASMB,

My perspective on this question differs somewhat from the others who have
responded. 

It certainly is possible to obtain much better reproducibility than you
describe, but not necessarily for every sample, and not necessarily without
significant effort. 

Just within the last couple of weeks I ran each of two samples in
triplicate. These were moderately homogeneous pharmaceutical samples with
1-4% total aggregate. For one sample the total aggregate content came back
X%, (X + 0.1)%, and (X - 0.1)%. For the other it was Y%, (Y + 0.2)%, (Y -
0.1)%

Those results are certainly among the best I've obtained, but for total
percentage aggregate the standard deviation I obtain is usually no worse
than 0.5% and often 0.3% or better (that is +/- 0.3% out of the total of
100%, not 0.3% of the actual aggregate content). Everything I'm discussing
here is for absorbance data and assumes you have near optimal signal/noise.

I also did a study a few months back, where I had 4 replicate measurements,
and we calculated the standard deviations in the areas for the individual
aggregate peaks (there were something like 6 different species as I recall).
Consistent with what I see from running simulations, the highest variability
is for dimer, since its separation from monomer is poorest, but even the
dimer standard deviation was quite good at ~0.3%. As you went out to larger
aggregates the variability in peak area dropped, reaching 0.2% to 0.1% for
octamer and larger.

While those results demonstrate it is possible to obtain excellent precision
and reproducibility, that represents what I can now do after using this
approach on literally thousands of samples. I certainly didn't get those
sorts of results when I first started. Good reproducibility requires a great
deal of care with experimental details, and a strong understanding of how
the analysis software works so you can avoid the pitfalls. Despite all that,
I of course still see the occasional outlier.

I also must point out that there are certain proteins for which the
reproducibility is poorer, and exactly why that is true is often not clear
to me. For one type of protein I see aggregation induced during the SV
experiment, and the extent of aggregate formation is a characteristic of
individual centerpieces (the specific centerpiece, not just the type). Thus
presumably this represents some sort of surface-induced aggregation. 

Turning to the other part of your question, definitely yes your
reproducibility issue could be related to experimental problems.
Centerpieces that are scratched on the interior walls of the channels, or
where the center rib was bent when there was a leak at fairly high speed,
cause weird effects. Generally though those things will give you data that
fit poorly (systematic residuals).

Overall my advice is that if you really want to evaluate and improve
reproducibility and/or test for instrument or centerpiece problems you have
to run lots of replicates---a single repeat can alert you to a problem, but
likely won't get you very far toward solving it.

John Philo

-----Original Message-----
From: rasmb-admin at server1.bbri.org [mailto:rasmb-admin at server1.bbri.org] On
Behalf Of Yi-Ming_Li at hgsi.com
Sent: Thursday, October 21, 2004 4:07 PM
To: rasmb at server1.bbri.org
Subject: [RASMB] precision limit in quantifying protein aggregates by
sedimentation velocity with Sedfit


----------------------------------------------------------------------------
------
The older archived RASMB emails can be found at:
http://rasmb-email.bbri.org/rasmb_archives
and current archives at http://rasmb-email.bbri.org/pipermail/rasmb/
Search All the Archives at: http://rasmb-email.bbri.org/rasmb_search.html
----------------------------------------------------------------------------
------

Hi! Everyone:
We routinely use Sedfit to analyze sedimentation velocity data to detect the
aggregates in our protein products. Many of our results are consistent with
the size-exclusion HPLC regarding the percentages of the aggregates in
protein products. But we also found that sometimes the repeatability of
sedimentation velocity analysis is not very good. For example, in order to
test the repeatability of the method, we loaded the same sample (a
monoclonal antibody) to two different cells and make a sedimentation
velocity run. Then we analyzed the data from the two different cells using
sedfit to get C(S) profiles. From the C(S) profiles we calculated the
percentage of the species. We got the following results:

Cell 1:
99.07% @ 6.4 s
0.74% @ 9.7 s
0.1% @ 11.3 s
0.08% @ 16.9 s
rmsd: 0.003517

Cell 2:
97% @ 6.4 s
1.43% @ 8.9 s
0.42% @ 10.9 s
0.27% @ 13.6
0.15% @ 18.2
rmsd: 0.003533

According to Dr. John Philo, sedimentation velocity is an excellent method
to detect and quantify the aggregate in protein products. Therefore I think
that the variations here are beyond the range we expect. Moreover, the
sedimentation coefficient of the dimmer and other oligomers are quite
different. Could anyone tell me, based on your experience and knowledge, the
precision limit of the software for quantitative determination of the
percentage of protein aggregate in a non-interactive system? If the
precision limit of the software is not a problem, what are the possible
sources for such large variations? Is it possible that the variations came
from the instrument or centerpieces?

Thanks!

Yi-Ming



_______________________________________________
RASMB mailing list
RASMB at rasmb-email.bbri.org
http://rasmb-email.bbri.org/mailman/listinfo/rasmb