Band Morphology
Examines the shape of the protein bands in a western blot. A real band is irregular: electrophoresis and diffusion give it ragged edges, an asymmetric profile, and a size that varies from lane to lane with the amount of protein. A band that is unnaturally smooth, circular, convex, and identical to its neighbours in shape and size is a sign of generation or cloning rather than measurement. The indicator measures each band's circularity, convexity, and edge smoothness, and the variation in shape and size across bands, and scores how artificial the morphology looks. It works on the pixels alone, with no model.
Technical description
W2 is a deterministic, generator-agnostic screen for artificial band shape. Genuine western blot bands inherit the irregularity of the physical process that makes them: proteins migrate and diffuse through a gel, so bands have fuzzy edges, slight asymmetry, and natural variation in width, height, and shape across lanes. Synthetic or cloned bands tend to be too perfect, smooth-edged, near-circular, fully convex, and suspiciously uniform from band to band. W2 detects the bands, recovers their contours, and computes three shape descriptors per band, circularity, convexity, and edge smoothness, then aggregates the mean circularity and smoothness and the variation in shape and size across bands. A combination of high circularity, high smoothness, low shape variation, and low size variation raises an artificial-morphology score. The image must be at least 32 pixels on a side and contain at least one detected band.
How it works
Bands are detected by adaptive thresholding and contour analysis. For each contour above a minimum area, three descriptors are computed: the circularity is the isoperimetric shape factor 4 pi area over perimeter squared, equal to one for a circle and lower for an elongated or ragged shape; the convexity is the area divided by its convex-hull area, near one for a solid shape with no concavities; and the edge smoothness is the convex-hull perimeter divided by the contour perimeter, near one when the boundary is smooth and lower when it is ragged.
The mean circularity and mean smoothness are taken across bands, along with the coefficient of variation of the circularities and, separately, of the band areas. Artificial signals are accumulated: a mean circularity above 0.75, or above 0.60, each add weight, as real bands are elongated rather than round; a mean smoothness above 0.92, or above 0.85, add weight; a circularity coefficient of variation below 0.10 with more than one band adds weight, because real bands vary in shape; and a band-area coefficient of variation below 0.10 with more than one band adds weight, because real lanes carry different protein amounts and so produce bands of different sizes, while cloned or generated bands are often identical. The summed signals are scaled to a score capped at 5.0. Bands with notably high circularity or smoothness become findings with their locations. The metadata records the band count, the mean circularity and smoothness, and the shape and size coefficients of variation.
Score thresholds
| Score | Meaning |
|---|---|
| 0 to 1 | Bands have the irregular, variable morphology of real electrophoresis. |
| 2 to 3 | Some bands are unusually smooth or circular, or the bands are somewhat uniform. |
| 4 to 5 | Bands are smooth, circular, and uniform in shape and size. Consistent with generated or cloned bands. |
Why this matters
The morphology of a band is a fingerprint of how it was made. A real band is the smeared, diffuse trace of migrating protein, so its shape is irregular and its size reflects the loaded amount, varying across lanes; a band that is geometrically clean and uniform did not pass through a gel. The ethical guidance on scientific images stresses that bands must not be beautified or synthesised into idealised shapes, and that manipulation often shows up as edges and contours that are too perfect [3]. Detecting this needs a quantitative notion of shape, and the circularity that W2 uses is the classical isoperimetric shape factor introduced by Cox to grade roundness, equal to one only for a circle [1]. Uniformity is the other half of the signal: large surveys of biomedical figures show that fabricated blots frequently reuse a single band, which makes the bands identical in shape and size, exactly the low variation W2 measures [2]. By combining how perfect each band is with how alike the bands are, W2 separates the natural variability of real electrophoresis from the artificial regularity of generation and cloning.
Limitations
Shape analysis depends on clean band detection and segmentation, so faint, smeared, or touching bands can be merged or missed, distorting the descriptors. Circularity computed from a digital contour is biased by the staircasing of the boundary, which can lower it for small bands. The edge-smoothness measure reads boundary convexity rather than fine edge texture, so a solid band with a clean rectangular outline reads as smooth even when it is genuine, which is why no single descriptor decides the score. The uniformity signals require more than one band and can be triggered by a genuine blot that happens to have similar bands. The thresholds are directional rather than exact. Band-to-band duplication by pixel content is the separate duplicate-bands indicator W1, and background fabrication is indicators W3 and W4, so W2 stays on the shape and uniformity of the bands themselves.
Theoretical background
W2 rests on the physics of electrophoresis. A protein band forms as molecules of similar mass migrate together through a gel and diffuse laterally, so the resulting mark is intrinsically irregular: its edges are gradients rather than lines, its profile is rarely symmetric, and its extent depends on the quantity loaded, which differs between lanes. These properties make a real band low in circularity, less than fully smooth, and variable across a blot. A synthetic band drawn or generated to look clean, and a cloned band copied across lanes, lack this stochastic origin: the drawn band is too round and too smooth, and the cloned band is identical to its source in both shape and size. The shape descriptors quantify the first kind of artifice, and the coefficients of variation quantify the second, so the score rises when a blot shows both the unnatural perfection of individual bands and the unnatural sameness across bands. Reading morphology therefore tests whether the bands carry the irregularity that the physical process necessarily imparts.
References
- Cox EP. A method of assigning numerical and percentage values to the degree of roundness of sand grains. Journal of Paleontology. 1927;1(3):179-183. https://www.jstor.org/stable/1298056
- Bik EM, Casadevall A, Fang FC. The Prevalence of Inappropriate Image Duplication in Biomedical Research Publications. mBio. 2016;7(3):e00809-16. DOI: 10.1128/mBio.00809-16
- Cromey DW. Avoiding twisted pixels: ethical guidelines for the appropriate use and manipulation of scientific digital images. Science and Engineering Ethics. 2010;16(4):639-667. DOI: 10.1007/s11948-010-9201-y