Quantify a gel honestly, without uploading it

Open an SDS-PAGE or blot image and box the bands: draw a box around one band, or drag across a whole row and every band in it is found and boxed at the same size. Gel Bench subtracts the background and reports a volume for every box, normalised the way you choose. The image never leaves this tab.

Drop a gel image here, or click to choose TIFF (including 16-bit scanner files), PNG, JPEG — nothing is uploaded

1
Draw a box around a bandBox tool (B) — drag from corner to corner
2
Or drag across a whole rowRow tool (R) — each band inside is found and boxed, all the same size
3
Read the numbersbackground-subtracted volumes, normalised, in the table and chart

Gel

target (T) loading control (C) standards (S) other bands lane gel area

No box selected.

Image

dark bands on a light background

Tick for Coomassie, silver or a colorimetric blot. Untick for chemiluminescence or fluorescence, where bands are bright.

Background under each box

Boxes and lanes

keep every box in a row the same size (resize one, resize all)

Intensities are sums over an area, so bands can only be compared when the boxes are the same size. Leave this on unless you have a reason.

show each box's number and intensity on the image

Lanes — ladders (grey) are left out of every number. Click a lane to switch it:

Normalise

Normalise the target to Report relative to Lane names (comma-separated, optional) Standards, ng per lane (optional — for a standard curve)

One entry per lane, in order; “-” or blank for a sample lane. Load at least three amounts (e.g. 100, 250, 500 ng BSA). The standards are read from the row marked Standards, or from the target row if none is marked.

Results

Every box

Chart

the normalised target per lane

Lane profile

the selected lane, its background, and where each box sits

How the boxes were drawn, and what the numbers mean

1 · Signal. Each pixel becomes one number. A colour image uses the usual luminance weights (0.299 R + 0.587 G + 0.114 B); a 16-bit scanner file keeps its full range. With “dark bands on a light background” the values are inverted, so more protein = a larger number.

2 · Which gel. A figure from a paper is gels and blots on white paper, with labels, marker ticks, arrows, diagrams and charts around them. The paper is pure white while even a pale blot is not, so each gel shows up as a solid block of not-quite-white; text is only thin strokes, and a bar or a coloured box in a diagram is one flat colour where a gel always has grain. Each gel found is offered above the image (the largest is measured first), and only that gel is measured — never the labels around it. A thin black frame around a blot is trimmed off.

3 · Lanes. Each column of the gel is summarised twice — the signal standing above what runs slowly down it, and its mean above the gel between lanes — and the sum has one hump per lane. A hump is a lane when it clears the profile's own noise and holds some band; its centre is read at half its height. Two humps are one lane when a single band spans both and is barely wider than the gap between them (a band darker at its edges than its middle), but never when the gel between them falls back to background. The box is as wide as a typical band. Lanes can be unevenly spaced and some can be empty.

4 · Band rows. Each lane's profile (the signal averaged across the lane, row by row) has its slowly varying background removed by a one-dimensional rolling ball, and peaks that clear the noise — and sit on a real band, not a speck of dust — are bands. Bands at the same height across lanes are gathered into one row, and every box in a row gets the same width and height: the height of the biggest band in the row, so the strongest lanes are not cut off. An integrated density is a sum over an area, so two bands are only comparable when the same area was measured. The target and loading-control rows have a box in every lane (an empty lane is a real zero); other rows are boxed only where a band was found.

5 · Background. Three choices. Local ring takes the median of a frame of gel around each box — the usual densitometry correction (corrected intden = intden − mean background × area), which follows an uneven background without being pulled up by the band. Rolling ball uses the lane profile's background, like ImageJ's Subtract Background. Baseline under each peak is ImageJ's Gel Analyzer: a straight line from the valley just above the band to the valley just below it, and only the peak above that line counts.

6 · Numbers. Intden is the background-subtracted signal summed over the box. The target starts as the strongest row and the loading control as the most even one; change either by selecting a box. The lane total is all background-subtracted signal in the lane, for total-protein normalisation. The vs column divides every lane by the lane you pick.

7 · Standard curve. Load known amounts of a standard (at least three, e.g. 100, 250 and 500 ng BSA) and enter them under Standards. Their intden is fitted against ng by least squares (intden = a·ng + b, with R²), and every other lane's target is read back as ng. Amounts outside the range of the standards are extrapolated and marked. If the standard is a different protein from the target (mark its row with Set as standards), the amounts assume the stain binds both equally per ng — true only roughly for Coomassie. This follows Rinaldi (2025), Protein quantification in SDS‑PAGE gels by densitometry using ImageJ/Fiji, protocols.io.

What this cannot do. Numbers are proportional to protein only if the image is linear. A figure taken from a paper is 8-bit and already contrast-adjusted; a band whose peak sits at the top of the scale is clipped. Gel Bench warns about both but cannot undo them. Use the raw 16-bit file from the imager and treat everything as relative.