How was it made?
Simulator of cellulose acetate electrophoresis

Starting data

First of all, a warning: the simulator has a reasonable scientific basis, but it is empirically adjusted to resemble some real profiles I had access to. Since its purpose is illustrative, I have not intended it to be very accurate (which, in addition, is really hard since the details available for this technique are scarce).

Densitogram

The profile is made by adding up the individual curves for the 5 peaks of protein fractions (albumin and globulins alpha1, alpha2, beta and gamma, or the 5 LDH isoenzymes).

The shape of the individual peaks is achieved using a modified gaussian function, where the exponent is 1.8 rather than 2. This gives somewhat sharper peaks than the standard gaussian.
(I have lost the referece where I got this from. Other options ready to be used are the standard gaussian and a lorentzian, but this is the one that yields best result.)

y = a × exp( −0.5 × | (x − x₀) / b |^c )

It is construed that peak area must be proportional to the abundance of the respective protein fraction. The area is calculated as the product (peak height) × (peak width at half the height) —which is strictly correct for the normal gaussian.

The parameters used in the function are:

• x₀ defines peak position (mobility). I adjusted this empirically and it is fixed:

  Albumin	alpha1	alpha2	beta	gamma
  0.186	0.321	0.464	0.600	0.786

  LDH-1	LDH-2	LDH-3	LDH-4	LDH-5
  0.219	0.344	0.469	0.594	0.719

• b defines peak width. It is calculated as the width at half the height, divided by the square root of (8×ln2) —which is accurate for normal gaussian, and remained when I switched to modified gaussian.
• The width at half the height was empirically adjusted for each protein:

  Albumin	alpha1	alpha2	beta	gamma
  12	10	16	18	54

  LDH-1	LDH-2	LDH-3	LDH-4	LDH-5
  3.9	3.9	3.9	3.9	3.9

• a is the peak height. In the standard gaussian, area is (height) × (width at half height), so height is calculated as percent abundance (assuming this is proportional to area under the peak) divided by width "b", including also a proporcionality factor that I adjusted empirically to match the real profile.
• c is the parameter that makes this gaussian different to the original Gauss equation (where c=2). It changes peak sharpness. In the simulator, its value is fixed at 1.8.

All this provides a normal profile visually similar to the real profile that I had available.

Staining

Regarding the colour of the 'staining', I followed a visually attractive method, but again with a mathematical basis:

Colour is calculated for each horizontal position in the strip, from the height calculated in the densitogram:

1. Heigth at that position is made relative (in a 0--1 range) to the maximum in the graph.

2. A correcting function is applied that reinforces medium and low intensities more than high intensities. (I don't remember the reason for this, but it looked necessary for a better look.) It is a convex function (see the figure):

colour intensity = exp(8*(Max.Int. − Intensity) +1) / exp(8+1)

(I guess I adjusted that empirically)

3. The corrected intensity is multiplied by the maximum value of colour (separately for the three RGB components), which I define empirically to simulate the different staining methods. The mathematically cleanest one is 'autoradiography' which is based on pure black.

Webpage implementation:

JavaScript graphic libraries by Walter Zorn are used to achieve both plots: the drawing of the densitogram and the simulated stained strip.

• “JavaScript Vectorgraphics Library” (GNU LGPL) http://www.walterzorn.com/jsgraphics/jsgraphics_e.htm
• “JavaScript Online Function Grapher” (GNU GPL) http://www.walterzorn.com/grapher/grapher_e.htm

All the calculation needed for building the graphs (above described) is done using JavaScript.

Version history

September 2013: version 2.1:

• A deeper blue color for Coomassie stain.
• Practice activities were added.
• English translation added.

in Spanish only:

October 2009: version 2.0:

• Peaks are made wider and a little closer, for a more realistic aspect (see figures).
• Added more pathological profiles.
• The (internal) 100% reference is removed and replaced by total protein concentration (needed for some pathological profiles and for more realism).
• Interface adjustments (see figures). Among them, title changed from “plasma proteins ” to “serum proteins” since the profile does not include fibrinogen (it would be an extra peak in the beta zone).
• Added the simulator for LDH isoenzymes.
• Part of the JavaScript code separated to make it common to both simulators.

January 2008: small fixes, version 1.1

May 2007: first release, version 1.0

v. 1.1	v. 2.0
Plasma proteins	Serum proteins

Angel Herráez. Part of the Biomodel.uah.es website