Blots, gels and micrographs are data, not illustrations. When a journal flags one, review can stop until the authors produce the original, unprocessed files.
Why journals screen figures at all
A western blot, a gel or a micrograph in a research paper is data. Readers treat each panel as a record of what was on the membrane or under the objective, so a change to what the panel shows is a change to the results. The ICMJE Recommendations list "data falsification including deceptive manipulation of images" among the forms of scientific misconduct.
That does not make every processed image suspect. Nature Portfolio's image integrity policy says a certain degree of processing is acceptable, and for some techniques unavoidable, but that the final image must correctly represent the original data. The same policy says editors may use software to screen images for manipulation.
So the question a screener asks is not "was this image edited?" It is "does the figure still show what was acquired, and can the authors prove it?" A flagged figure may turn out to be an honest assembly error or something more serious. Either way, it gets resolved with the original file, which is why this is a lab and institutional issue as much as an author's.
What screeners flag
| Problem | What it looks like |
|---|---|
| Duplicated panel | The same image used for two different conditions, in one figure or across figures |
| Repositioned duplicate | Two panels that overlap once one is shifted, rotated, flipped or resized |
| Cloned region | Bands, cells or patches of background repeated within one image |
| Unmarked splice | Lanes from different positions, gels or exposures joined with no dividing line |
| Selective adjustment | Contrast changed in one region, or in experimental panels but not controls |
| Cleaned background | A flat, noise-free background around bands |
| Composite field | Cells from several microscope fields shown as one field |
| Overcropping | Bands cut so tightly that background and non-specific bands are lost |
Most of this comes straight from published policy. The PLOS ONE figure guidelines call "the introduction, enhancement, movement, or removal of specific feature(s) within an image" unacceptable, and say images must keep the background noise they originally contained. Nature Portfolio says cloning and healing tools, or any feature that deliberately obscures manipulations, are unacceptable.
How duplication gets caught
Screening usually has two layers. Software compares panels across the whole manuscript, including copies that have been rotated, flipped or rescaled, and commercial screening tools advertise detection of copy-move and splicing edits, comparison against large collections of published figures, and detection of AI-generated images.
The second layer is a person. A screener adjusts brightness, contrast or colour on a copy of the figure, which can expose seams, pasted areas and repeated background texture that look normal at ordinary settings. Anything left unexplained becomes a request for the originals.
Before submission, ask someone who did not build the figures to lay every panel, main and supplementary, side by side and confirm that no two panels meant to show different conditions share an image. If one loading control genuinely belongs to two panels from the same membrane, say so in both legends, so the reuse reads as disclosed.
Acceptable adjustments and where the line sits
| Adjustment | What journals expect |
|---|---|
| Brightness and contrast | Applied to the whole image and equally to controls; no data may disappear |
| Cropping | Keep all important bands and some background above and below them |
| Gamma changes and other nonlinear adjustments | Disclosed in the legend |
| Pseudo-colouring | Disclosed in the legend |
| Changes to individual channels in a merged image | Noted in the legend |
| Reducing image size | Acceptable; increasing the pixel count without shrinking the image creates data |
| Images from different times or locations in one picture | Borders clearly marked and explained in the legend |
| Clone, healing or erase tools | Not acceptable |
| Thresholds chosen to emphasise one region | Not acceptable |
Most rows follow Nature Portfolio's wording; the cropping margin and the pixel rule come from PLOS ONE. Nature Portfolio also asks authors to name image acquisition tools, processing software and key settings in the Methods.
A decision rule for anyone preparing figures: would a reader of the adjusted image reach a different conclusion about the data, the background or the controls than a reader of the raw file? If yes, do not make the change. If no, but the change is nonlinear or affects one channel, disclose it.
Western blots and gels: the most detailed rules
Blot and gel rules differ between publishers, so check the target journal before assembling these figures.
- Splicing within one image. PLOS ONE accepts removing a lane and joining the remaining lanes if the pieces come from the same original image, with a thin white or black line at the join and a description in the legend. Nature Portfolio requires rearranged non-adjacent lanes to be clearly delineated and the rearrangement stated in the legend.
- Bands from different blots. PLOS ONE does not allow composite panels built from different blots, gels or exposures; they must appear as separate panels with a legend saying so. Nature Portfolio strongly discourages quantitative comparison across different gels or blots; where it is unavoidable, the legend must say the samples come from the same or parallel experiments processed in parallel.
- Controls. Nature Portfolio requires loading controls such as GAPDH or actin to be run on the same blot, and sample processing controls run on different gels to be identified in the legend.
- Exposure. Nature Portfolio discourages high-contrast gels and blots because overexposure may mask additional bands. PLOS ONE asks for background above and below bands and for non-specific bands to be kept and explained.
A legend that pre-empts the query reads like this:
Lanes 1 to 3 and 4 to 6 are from the same membrane and exposure. Two intervening lanes were removed and the join is marked with a black line. The uncropped image is provided in the supplementary raw image file.
The raw image file editors ask for
Nature Portfolio requires unprocessed original images of gels and western blots for all life science papers with the final accepted version, and publishes them in the Supplementary Information. PLOS ONE requires original, uncropped and minimally adjusted images for every blot and gel result; they are not needed at first submission but are required during peer review or before acceptance.
PLOS ONE's instructions are a useful standard to build to. It asks for one PDF in which each original image is labelled with the loading order, sample identity, capture method and the figure panel it supports, with molecular weight markers included or indicated and unused lanes marked with an "X". If there are concerns about the availability of these images, the submission is placed on hold.
Build this file when the figure is made, not when a journal asks for it.
When a journal queries a figure
Nature Portfolio says editors may request unprocessed data files during review, and that if the files are unavailable, peer review may be halted until the issues are resolved. PLOS ONE says it may ask for original images in their original file format and resolution, may decline the manuscript if the response is not satisfactory, and may contact the authors' institutions to assist with an investigation.
A response that resolves a query quickly:
- Send the original acquisition files in their native format with metadata, not exported JPEGs or screenshots of the figure.
- Map each questioned panel to its source file and list every step in between: crop, rotation, adjustment and software.
- If the query exposes an honest mistake, such as a wrong panel pasted in, say so directly and supply the correct image from the original data.
- If an original cannot be found, say that plainly. Never send a newly acquired image as though it were the original.
- Tell all co-authors, and the lab lead or research office if local policy requires it.
For institutions, this means a journal may contact you, so have a named contact and a way to secure original data quickly. Our post on the research integrity responsibilities of a lab lead covers the data custody side.
After publication
The ICMJE Recommendations on corrections say honest errors require a published correction, and that errors serious enough to invalidate a paper's results may require retraction. Where misconduct is alleged, editors should follow procedures such as those from COPE, consider informing institutions and funders, and may publish an expression of concern until the outcome is known; if an investigation proves misconduct, the article should be retracted.
Here, too, the outcome can hinge on the original files. Nature Portfolio notes that missing unprocessed data can make post-publication issues hard to resolve, and PLOS says the unavailability of original data may affect editorial outcomes. For how each type of notice works, see when and how to correct the published record.
A pre-submission image check for labs and research offices
- Original acquisition files stored read-only, separate from working copies
- Instruments, software and key settings recorded and named in the Methods
- Every panel traced to a named source file
- Brightness and contrast changes applied to whole images and equally to controls
- Nonlinear adjustments, pseudo-colouring and channel changes disclosed in legends
- Splices marked with a line and described in the legend
- Loading controls on the same blot as the samples they normalise
- All panels, including supplementary figures, checked side by side for duplication
- Uncropped, labelled blot and gel images assembled in one file that matches the figures
- Retention period for originals at least as long as the longest institutional or funder requirement
Retention is the item institutions control most directly. Nature Portfolio recommends keeping unprocessed data and metadata after publication, ideally in perpetuity, and PLOS notes that its data policy sets no time limit even where institutions or funders do. Put a retention rule for original images into lab data policy and your research integrity office checklist.
Directive Publications collects its author guidelines and research integrity policies on its For Authors page. Read them alongside the target journal's figure instructions before the first panel is assembled, because the original files a screener asks for can only be kept at the moment they are created.
Frequently asked questions
What image adjustments are acceptable in scientific figures?
Brightness and contrast changes are generally acceptable when they are applied equally to the whole image and to its controls, and no data or background disappears as a result. Cropping is usually fine if the important bands and some surrounding background remain. Nonlinear changes such as gamma adjustments, pseudo-colouring and changes to individual colour channels should be disclosed in the figure legend. Cloning, healing, erasing or enhancing a single feature is not acceptable.
How do journals detect duplicated or reused images?
Editorial staff and screening software compare panels across the manuscript for identical or overlapping images, including copies that have been shifted, rotated, flipped or resized. Some commercial tools also advertise comparison against large collections of published figures. When something looks wrong, a screener adjusts contrast or colour on a copy of the figure to expose seams, pasted areas and repeated background, and the journal then asks the authors for the original files.
Is splicing western blot lanes considered image manipulation?
Splicing is not automatically manipulation, but it must be visible and explained. PLOS ONE, for example, accepts removing lanes and joining the rest when every piece comes from the same original image, provided a thin line marks the join and the legend describes it. Joining bands from different blots, gels or exposures into one panel so they look like a single experiment is not acceptable; show them as separate panels instead.
What happens if image problems are found after an article is published?
Journals such as PLOS ONE and the Nature Portfolio titles state that they may request the original, unprocessed data, and the editor may contact the authors' institution. An honest error that leaves the conclusions intact normally leads to a correction. Where misconduct is alleged, the ICMJE Recommendations say editors may publish an expression of concern while an investigation runs and should retract the article if misconduct is proven.
How long should researchers keep original unprocessed image files?
There is no single rule, because institutions and funders set different retention periods. Some publishers set no end date: PLOS says its data availability policy has no time limit, and Nature Portfolio recommends keeping unprocessed data and metadata after publication, ideally in perpetuity. A workable rule is to keep originals and their metadata for as long as the article is in the literature, and never for less than your institution's or funder's minimum.