<html>
<body>
John,<br><br>
just a small note to add to your comment - it is actually possible to fit
negative signals in the newest version of sedphat (4.02) in the hybrid
discrete/continuous model, if you check the box entitled
"neg." in the field of the discrete species #1. In this
case, this particular Lamm equation solution will be multiplied with
minus 1. This option was introduced precisely to deal better with
situations you describe, where we have excess buffer components in the
reference. We encountered this problem with samples that were not
dialyzed properly (or somebody picked the wrong buffer). <br><br>
I think you're absolutely correct that this can create false peaks, if
not accounted for.<br><br>
You can currently download this version from the sedphat site.
Originally, I wanted to wait with the official announcement a few days
for the concurrent update of the website, and for the new release of a
slight further upgrade to version 4.04, which has a few more bugs fixed
and slightly revised interface for ITC data analysis. This will be
out in a matter of days. <br><br>
Peter<br><br>
<br><br>
<br>
At 07:22 PM 9/26/2005, you wrote:<br>
<blockquote type=cite class=cite cite="">
<font face="arial" size=2 color="#0000FF">Qin,<br>
</font> <br>
<font face="arial" size=2 color="#0000FF">I agree with Peter's response,
but I wanted to point out some additional issues. It is quite difficult
to get good cancellation of the micelle signals by having Tween in the
reference buffer, particularly for absorbance studies. Most, if not all,
of the absorbance of Tween is due to impurities, and the absorbance
varies markedly depending on the manufacturer and even lot-to-lot. Unless
your reference buffer is made from the very same bottle of Tween, it is
likely it won't match the Tween in the sample. Further, Tween also makes
the solutions very good at picking up UV-absorbing extractables from
plastic tubes, syringes, etc. which creates another source of
sample/reference differences. Even if everything does match perfectly,
you won't get perfect cancellation of the micelle signals unless you
match the meniscus positions in sample and reference too.<br>
</font> <br>
<font face="arial" size=2 color="#0000FF">This has important consequences
on the software side, because it is all too easy to end up with NEGATIVE
signals from the micelles (more detergent, or at least more absorbance,
on the reference side). The <i>c(s)</i> algorithms cannot handle negative
signals (all concentrations are forced to be positive). Thus if you do
have any negative signal (which you always will if the meniscus positions
are not matched) then the fitting algorithm cannot possibly accurately
represent the data and you will very likely get some sort of false peak.
Because of these issues I often use a buffer without the Tween as the
reference sample, which ensures the micelle signals are positive. <br>
</font> <br>
<font face="arial" size=2 color="#0000FF">That 2 S value for the micelles
you cited is in fact experimental, not calculated, but as Peter said the
actual micelle size varies with solvent conditions in addition to the
usual density/viscosity effects. Definitely you should run your buffer
versus water to establish the exact micelle sedimentation coefficient for
your conditions. <br>
</font> <br>
<font face="arial" size=2 color="#0000FF">Overall though when your major
protein component has a sedimentation coefficient close to that of the
micelles you likely to always have significant interference from
them.<br>
</font> <br>
<font face="arial" size=2 color="#0000FF">John<br>
</font> <br>
<br>
<font face="tahoma" size=2>-----Original Message-----<br>
<b>From:</b> rasmb-admin@server1.bbri.org
[<a href="mailto:rasmb-admin@server1.bbri.org" eudora="autourl">
mailto:rasmb-admin@server1.bbri.org</a>] <b>On Behalf Of </b>Qin
"Chin" Zou<br>
<b>Sent:</b> Wednesday, September 21, 2005 7:22 PM<br>
<b>To:</b> RASMB@server1.bbri.org<br>
<b>Subject:</b> [Rasmb] Tween 80 effect on c(s)<br><br>
</font>
<dl>
<dd><font face="Courier New, Courier" size=2>Hi all,<br>
</font><br>
<dd><font face="Courier New, Courier" size=2>I wonder if anyone has
looked at the effect of Tween 80 on the s distribution. I remember that
John Philo once asked the similar question and had calculated s for Tween
80 around 2s. <br>
</font><br>
<dd><font face="Courier New, Courier" size=2>I have an 18KD protein that
runs at about 1.7s when there is no Tween 80 in the buffer. However, in
the presence of Tween 80, there were two species 1.2 and 1.5s (without
regularization. If F=.68, then two peaks are overlapped).<br>
</font><br>
<dd><font face="Courier New, Courier" size=2>The experiment was run at
60K rpm, absorbance at 276nm. Also, the fit was not good with rmsd above
0.01. In addition, I tried to scan at 289nm, where the absorbance of
Tween 80 is minimal, with higher protein concentration (5mg/ml). Although
the 1.2s peak is not there anymore, the main peak is at 1.4s instead of
the expected 1.7s. Also the fit was quite bad with rmsd around 0.03. For
both experiments, I had 0.02% Tween in the reference buffer.<br>
</font><br>
<dd><font face="Courier New, Courier" size=2>Is it possible that the 1.2s
peak is due to the Tween absorbance at the region between the meniscuses
of reference and sample sectors? I have not tried with the same column
height in both sectors. It is hard to match them exactly. Would it be
better to run this kind of experiment without including Tween 80 in
reference buffer and note the potential 2s peak as the Tween peak? For my
protein of 1.7s, it could be still difficult. Any input will be
appreciated.<br>
</font><br>
<dd><font face="Courier New, Courier" size=2> <br>
</font><br>
<dd><font face="Courier New, Courier" size=2>Qin "Chin"
Zou<br>
</font><br>
<dd><font face="Courier New, Courier" size=2>Eli Lilly and Co.<br>
</font><br>
<dd><font face="Courier New, Courier" size=2> <br>
</font><br>
<dd><font face="arial" size=2> <br>
</font>
</dl></blockquote></body>
</html>