Image integrity best practice: the problem with altering western blots

Monday, 22 January, 2024

Image integrity best practice: the problem with altering western blots

Western blots are frequently used in scientific and medical research to detect proteins and semi-quantify protein amounts. Searching for the phrase on PubMed1 returns over 400,000 publications — with more than 21,000 from 2022 alone. Dr Dror Kolodkin-Gal, founder of image integrity tool Proofig AI, has found that many of the integrity issues AI tools recognise in scientific images exist within western blots. Here he explores how these image issues arise and what researchers and institutions can do to ensure the integrity of the work they submit to journals.

If someone reports an issue — such as an image integrity problem — in a published scientific article, an investigation may be conducted, which may result in a retraction. In a 2022 study of 1367 articles2, the American Association for Cancer Research (AACR) found that image issues were most likely to come from western blots (40.8%), so researchers and institutions must take care when using these images in their content.

Where issues can arise

Space is at a premium in journals. To keep a manuscript compact, researchers might crop or splice images to remove blank space. Splicing refers to removing unwanted empty space between lanes in a gel or combining lanes from different gels by digitally ‘stitching’ the lanes together (Figure 1). Generally, splicing is unacceptable, and cropping can only be performed if it is declared and why it was necessary is explained. Any cropped lanes must be clearly delineated, the molecular weights must be unchanged, and the unprocessed original image data must be available. This is especially important because splices are often invisible to the naked eye, so mislabelled images containing splices could go unnoticed by fellow researchers or a busy reviewer checking the manuscript after submission and end up being published.

The guidelines published by journals often differ on what they deem acceptable levels of image alteration for the purposes of ‘beautification’. Some journals discourage forming figures from different gels; others encourage it where it improves clarity. However, all are united on the idea that images should not be deliberately manipulated in order to show a result incongruous with the original data.

“Data must be reported directly, not through a filter based on what you think they ‘should’ illustrate to your audience,” write Rossner and Yamada in their seminal paper on the topic3. “For every adjustment that you make to a digital image, it is important to ask yourself, ‘Is the image that results from this adjustment still an accurate representation of the original data?’ If the answer to this question is ‘no’, your actions may be construed as misconduct.”

Limiting retraction risk

According to Bushra Khair4, a research integrity specialist at Frontiers, “While it is common practice to include cropped western blot images in manuscripts so as to include the relevant proteins, cases where there is questionable cropping, undeclared cropping or duplicate bands raise concerns for the team. In such cases, the authors will be requested to provide originally saved image files for assessment.”

An issue like this might be reported many years after the research, by which time the original gels or blots may be impossible to retrieve. If the authors are unable to address any issues or provide sufficient explanation, the article could be retracted. Retraction not only delays research and increases financial costs, it also causes potentially irreversible reputational damage that can affect the researchers’ careers and institution’s future.

Nature recommends “retaining unprocessed data and metadata files after publication, ideally archiving data in perpetuity”5. However, avoiding the situation in the first place is the way to go.

Under the radar

In a 2023 webinar on image integrity in biomedical research publication6, Jana Christopher, image data integrity analyst at FEBS Press, said, “Many journals only perform spot checks and their image integrity screening is often mostly reactive rather than preventative, meaning that if a published paper is flagged up to them then they might, eventually, take a closer look and correct or retract the paper, but they seldom check before they publish.”

Occasionally editors identify issues before publication, but some experts suggest that the real responsibility of ensuring that submitted manuscripts are worthy of publishing lies with researchers.

“Whilst it is absolutely imperative to react to concerns and to correct the record … it is equally important and it’s valuable that journals check figures pre-publication to avoid errors, misinformation and frauds entering the literature in the first place,” Christopher continued. “Authors and reviewers should contribute to this effort and pay attention to image data.”

In my own experience, western blots are the source of about 40% of image integrity issues. Another 40% comes from microscopy images and the remaining 20% covers everything else. I often encounter researchers with a ‘that would never happen to me’ mentality, but a vast proportion of biosciences papers contain these two types of images. The likelihood of missing an issue with a western blot is also not surprising as this form of image is time-consuming to produce and difficult to review by eye. There’s really no guarantee your images accurately represent your findings, even if you check them thoroughly.

While deliberate manipulations do occur, most issues across all types of images are the result of innocent mistakes. Because images can easily be mishandled and end up flipped, duplicated, rotated or cropped when added to a manuscript, it’s crucial to focus on finding these mistakes pre-submission.

The future of image integrity

To combat these issues, researchers and institutions have a responsibility to ensure best practices when working with western blots. Firstly, it’s essential to make it abundantly clear when cropping, splicing or altering western blots and doing so in a way that is acceptable to the intended publication. Authors must provide ample explanation and retain original files for as long as possible.

Secondly, using a software tool, such as Proofig AI, to check work for any image issues before submission has benefits over checking manually. AI significantly reduces the time image checks take by making thousands of comparisons in seconds and flagging what needs attention. Lead researchers can then use the AI tool to investigate potential issues, adding different filters to understand where splicing, cropping or deletion may have occurred. In projects that involve collaborating with researchers from different institutions, this provides an opportunity to get clarification on any potential issues before publication.

Figure 1: Before and after using Proofig AI.

Identifying image mistakes using AI tools can also highlight areas for improvement in the laboratory, providing an opportunity to improve mentoring and promote the adoption of best practices. For instance, if a professor’s students consistently make similar mistakes, it suggests a need for more focused instruction and guidance on those specific concepts or skills.

The scientific publishing landscape is increasingly competitive. Science, for example, rejects 84% of submissions before review and has an acceptance rate of just 6.1%7, so researchers must ensure the utmost rigour and integrity in their work. Researchers already use AI tools for spelling, grammar and plagiarism checks — adding image checking software to their suite will help protect their reputation, avoid potential retractions and enhance the credibility of their findings. To see how the software works, or for more information, visit








Top image credit:

Related Articles

CIA Labs embraces LIMS to drive efficiency, best practice

CIA Labs recently streamlined its approval process, reduced the burden of paperwork and improved...

New method for assessing red blood cell deformability

Using image-based deep learning and dynamic shape classification techniques, researchers have...

AI can detect COVID and other conditions from chest X-rays

As scientists compare different AI models to improve automated chest X-ray interpretation, a new...

  • All content Copyright © 2024 Westwick-Farrow Pty Ltd