[Protein-analysis] Re: quantifying western blots
gillespie.amanda.spamfilter at gmail.com
Mon Jun 19 03:05:22 EST 2006
This is for western blots - how about for quantifying bands in ordinary
gels? Stained with a stain with a wide dynamic range like Sypro, not
On 6/17/06, DK <dk at no.email.thankstospam.net> wrote:
> In article <4ffph8F1isuitU1 at news.dfncis.de>, Joerg Schaber <
> schaber at molgen.mpg.de> wrote:
> >Thanks for your reply.
> >Of course, I should have mentioned that I do not want to determine
> >absolute protein quantities but only relative quantities for each gel,
> >by scaling the maximal value on each gel to one, for instance. Just to
> >get some numbers.
> That's the whole thing: unless you are in a subset of a very specific,
> rare and narrow range/situations, the numbers you will get will be wrong.
> Pixels =/= protein amount. As far as biology is concerned, it deals with
> protein amounts (or no amounts at all), NOT pixels. So if you count
> something, it is only meaningful until you count protein!
> To illustrate: Say, you take some sample and load some amoutn of it
> on a gel. Lets call it 100%. Then on other lanes you load 50,25,12,6,2,1%.
> Then do a Western, scan and count pixels. The function pixels(%loaded)
> will not be linear! Very crudely, it will be approximated by the sum
> of hyperbola and a straight line. The relative contribution from the two,
> half-saturation of the hyperbola and angle of the straight are all highly
> variable. Sooooo, it all means that if you just take two samples and
> count pixels that correspond to them then you will NOT be able to
> say anyhting more than "one sample has more pixels and that must
> means more protein". How much more will remain completely unknown
> in the absense of the signal-response calibration!
> >Moreover, I assume that I use the same loading for each lane and that
> >the response is not saturated.
> >When I understood you correctly you propose the following:
> >1) Quantify the band for a specific lane (e.g. by counting black pixels
> >in a rectangle around the band).
> Useless unless done with calibrations curve. Without calibration
> curve, it's a pretense of science, not science, a pure sham.
> >2) define the background for this lane (e.g. by counting black pixels in
> >a rectangle placed next to the one above but in the same lane)
> Wait, there might be a confusion here: you are not countng lanes
> - right? In gels - and Westerns in particular - it's the *band*, not
> lane counting that usually makes sense. Assuming that you really
> mean "band":
> It's a lot more complex. In the very crude first approximation, this is
> right (assuming the above implies subtracting background value that
> corresponds to the area defined as a "band"). Problem here is how
> to define what is band and what is background. In truth, the best
> results can only be obtained manually by someone with a clue and -
> as much as possible - without bias.
> Problem is, next to the band to be quantified there is frequently
> another band and they sometimes start merging. Measuring background
> outside of the lane is compeltely pointless because it will always
> be much lower than the background in the lane.
> So the objective is: a) define and area that covers ~ 100% of you band
> without incorporating any of the pixels from neighbor band(s), count total
> pixels inside that area; b) define at least ~ equal area of the
> background next to that band (minimally to the right and left of the
> band; and as much as possible on top and from bottom of it - all without
> incorporating pixels from other "real" bands; calculate average pixel
> intensity per/area; c) using that value, calculate the total background
> value in the experiment band area, and subtract it from the total.
> >4) do this for each lane.
> For each *band*, yes.
> Proteins mailing list
> Proteins at net.bio.net
If at first you don\'t succeed, destroy all evidence that you tried ---
Fight back spam! Download the Blue Frog.
More information about the Proteins