[RASMB] Compare MW obtained from Segal vs. MW obtained from Beckman Origin

Tue Mar 11 17:43:00 PST 2003

Chris et al.,

I have not used the Segal software and thus can't comment specifically on
that, but I think the underlying problem here is probably your data itself.
You have a very large baseline offset of ~0.11 OD, and under such
circumstances the apparent molecular weight will be quite sensitive to the
offset value (and very highly correlated to it). I suspect the differences
between the Origin and Segal results have mostly to do with how the baseline
offsets are handled. That large baseline offset also suggests there may have
been something wrong with the sample---it actually looks to me like it may
not really be a constant offset but rather that you have some low mass
contaminant (or even some proteolysis).

Second, you haven't given any confidence limits for these numbers and thus
haven't addressed the question of whether the difference in mass results
from the different programs is even statistically significant. The Origin
software, for example, says your 'Edit 1' result of 67 kDa has a 95%
confidence interval of 62 to 71 kDa, and thus it is certainly not different
from the 65-66 kDa Origin Edit 2 or Edit 3 results. This 67 kDa value
probably is a real difference from the 59 kDa Segal result (which as I said
above probably relates to baseline offsets), but one can't even properly ask
that question unless there is a confidence limit on the Segal result.

Third, regarding your data again (but not the software issue), as Ariel
Lustig noticed too, you were using quite a long solution column (5 mm). Are
you sure that sample was fully at equilibrium within the 24 hr you
apparently gave it? That 24 hr for a 5 mm column is equivalent to <9 hr for
the usual 3 mm column height, which definitely seems too short.

Finally, I would urge some caution with trimming your data. Your message
implies that your criterion for data selection is to get the lowest
variance, but that is only a legitimate criterion if you are fitting to the
correct model. For example, it is easy to make a sample that contains
irreversible aggregates fit nicely as a single species by simply trimming
away the regions at the base where the aggregates are concentrated (and by
rigorously applying your lower variance criterion that is exactly what will
happen). Especially when you have a limited range of concentrations, getting
your data to fit a given model can quickly become a self-fulfilling prophecy
if you just trim away the points that don't fit!

Knowing what is 'good' or 'bad' data in equilibrium experiments is
unfortunately a bit of an art, but in my view should primarily be guided by
your knowledge of the hardware and optical system. Thus one easy, unbiased,
criterion for 'bad' data is that the absorbance is too high (outside the
region of linearity). It is also true that the absorbance data tend to be
bad near the base of the cell, apparently due to reflections, and that bad
region should be the same for all samples in that type of cell (since it is
a matter of cell geometry). 

John Philo
Alliance Protein Laboratories

-----Original Message-----
From: rasmb-admin at rasmb-email.bbri.org
[mailto:rasmb-admin at rasmb-email.bbri.org] On Behalf Of Chin, Christopher
Sent: Wednesday, March 05, 2003 1:51 PM
To: ArielLustig Lustig (E-mail); Rasmb (E-mail)
Subject: [RASMB] Compare MW obtained from Segal vs. MW obtained from Beckman
Origi n

Dear Ariel,

I am a big fan of your Segal program because of the user-friendly feature.
The dynamic fitting pictures that change in front of your eyes for you to
select the best fitting radius region to use is a great time saver in data
analysis.  In the Beckman Origin program, editing equilibrium curve in the
dark by guessing is no fun.  I have to edit several radius regions, one by
one, in order to obtain the best variance value so I can have confident in
the result. 

I do a lot of calculation for Ka at various GuHCl concentrations in order to
estimate the free energy of subunit dissociation using Beckman's origin
program.  I constantly worry that I am not getting the best apparent
molecular weight to use for Ka calculation.  Lately I have combined Segal
and Origin program, meaning that I apply Segal program to get what I think
is the best fitting radius region, I then reapply this same radius region to
edit the subset from the Beckman Origin program (since your program can not
be used to calculate Ka), thinking that with this combine approach I can
surely get the best reliable MW to use for an accurate Ka. 

However, to my surprise, I do not get the same molecular weight even if I
use exactly the same radius boundary values.  Of the 4 set of edited subset
data I tested, the discrepancy various from 4 to 14 %.  Do you know where is
this discrepancy coming from? Is Beckman fitting equation different than the
equation you used for Segal program?  Is 4-14% discrepancy acceptable?  I
think I can live with 5-6% discrepancy. I am including the raw data
00008.ra1 for anyone who is interested in looking into this.  I have not
included NONLIN in this comparison.

Summary:  4 set of the edit data from raw data 00008.ra1

Edit Equilibrium curve

Boundary radius region used

MW obtained from Origin / variance

MW obtained from Segal / correlation

Discrepancy@ %

Edit 1

6.913-7.148

67 KD / 17.1

59 KD / 0.999458

12

Edit 2

6.942-7.139

66KD / 18.3

60 KD / 0.999504

9

Edit 3

6.920-7.139

65 KD / 17.8

56 KD / 0.999322

14

Edit 4

6.897-7.118

56 KD / 13.1

54 KD / 0.999435

4

@ ( MW obtained from Origin  -  MW obtained from Segal) / MW obtained from
Origin  

With best regards,

Chris

-------------------------------------------------------------------------- 
Christopher Chin 
Manager, XLA-Analytical Ultracentrifugation facility 
Sealy Center for Structural Biology 
HBC&G, 5.134 MRB.UTMB, Galveston,Tx 77555-1055 
cchin at utmb.edu, 409-772-1693, efax 708-585-1920 
--------------------------------------------------------------------------- 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20030311/48c2d51d/attachment.htm>