This page explains the approach behind automated band detection and quantification in the Blot Analyzer, and where it has known limits. We'd rather you understand exactly what the software is doing than treat it as a black box.
Band detection
When you run auto-detect, AutoBlot Studio processes each lane independently:
Local background subtraction — the large-scale background intensity gradient is estimated and subtracted from the lane, correcting for uneven illumination or staining across the membrane.
Otsu thresholding — after background correction, Otsu's method is used to automatically find the intensity threshold that best separates band signal from background, based on the histogram of the corrected image.
Connected-component labelling — pixels above threshold are grouped into discrete regions. Each region becomes a candidate band, and a tight ROI box is fitted around it.
Lane ordering — detected bands are labelled left-to-right within each lane for consistent referencing.
Detection is a starting point, not a final answer. You can drag, resize, add, delete, or rename any ROI box after auto-detect runs — every measurement is fully editable.
Saturation detection
Saturated (overexposed) pixels carry no quantitative information — once a pixel hits the sensor or display ceiling, increasing band intensity no longer changes the recorded value, which can flatten or distort densitometry.
AutoBlot Studio checks each lane for pixels at or near the maximum value (≥253 of 255 in each channel). If more than 5% of a lane's pixels are saturated, it's flagged so you can re-expose the image or interpret that lane's quantification with caution.
Saturation flags are a guide, not a guarantee. Heavily compressed JPEGs or images that have already been brightness-adjusted before upload may not reflect the original sensor data.
Outlier flagging
In the Quantification Suite, lane means can be checked against the rest of the group using Grubbs' test (two-sided, critical value ≈ 1.89 for typical group sizes). A flagged lane is a prompt to look more closely — repeat the lane, check for loading errors, or document a reason for exclusion. It is not, by itself, justification for removing data.
Known limitations
Very faint bands close to background noise may not be detected automatically. Use the manual ROI tool to add them.
Touching or overlapping bands in the same lane can be merged into a single connected component by the detector. Split them manually using the ROI tools.
Curved or skewed lanes ("smiling" gels) are not automatically corrected — straighten or crop the image before upload for best results.
Heavily compressed images (e.g. low-quality JPEGs) introduce compression artefacts that can affect both detection and densitometry. TIFF or PNG is recommended for quantitative work.
Background subtraction assumes a smooth, large-scale gradient. Sharp local artefacts — dust, scratches, edge effects from cropping — are not treated as background and may be picked up as false bands.
Quantification is relative, not absolute. As with all densitometry, band intensities should be interpreted relative to appropriate loading controls and normalised samples, not as absolute protein quantities.
The Integrity Checker (Lab plan) is a separate tool that screens images for signs of duplication or manipulation using perceptual hashing and similarity detection. It flags areas for investigation and, like all automated screening, is not a substitute for editorial or expert review.
Have a blot that AutoBlot handles poorly, or a methods question we haven't covered? Tell us at support@autoblotstudio.com — real examples help us improve detection for everyone.
Try it on your own blot
Upload an image and see exactly which bands are detected and why.
