Shutterstock
Western blots are frequently included in pharmaceutical research papers to report results when analysing proteins in a sample. However, research shows that Western blot images are a common source of integrity issues, the publication of which can be harmful to the advancement of new treatments. Here, Dr Dror Kolodkin Gal, founder of image checking tool, Proofig AI, explores the challenges faced when reviewing Western blot images in pharmaceutical research. Then, he advises on how to more effectively identify issues before publication to maintain the integrity of research that can improve health outcomes.
Western blotting is commonly used in laboratories around the world to detect proteins within a cell or tissue lysate in order to gauge their expression levels. A search for ‘Western blot’ on PubMed returns over 400,000 results, 20,000 of which are from 2022. However, according to leading image data integrity analyst Jana Christopher, MA, the percentage of manuscripts flagged for image-related problems, including issues with Western blots, ranges from 20 per cent to 35 per cent.
Including these issues can have far reaching consequences for pharmaceutical researchers, including potential rejection at approvals stage, loss of investor trust, depreciation in company stock value or loss of public trust regarding treatments. So, identifying and resolving these issues pre-publication is crucial for all parties involved in pharmaceutical research.
Risks to Publishers
A serious consequence of publishing an article containing image integrity issues is retraction. Arguably, retractions have the biggest impact on the authors' careers and reputations, but it can also lead to a lack of confidence in the findings in question and scepticism about any subsequent studies.
Not only can a retraction damage the reputation of a journal and authors, but it could lead to funding for future research projects being withheld — at best slowing, or at worst preventing, the development and marketing of potentially life-saving treatments. On a public health scale, high profile retractions could cause mistrust, potentially reducing uptake for certain treatments.
Where Issues Arise
People reporting instances of image issues often assume that the only reason for altering an image like a Western blot is to falsify results or fraudulently increase chances of publication, but most issues are honest mistakes.
In my experience of working with researchers to check images in their manuscripts before submission, around one in four manuscripts includes at least one image integrity issue. Most of these issues are mistakes, like duplications, but even if a mistake is made innocently or because of disorganisation, it can still impact the conclusions of the research and be very problematic.
While rare, some alterations are a deliberate attempt by cheats to mislead the reader and increase the chances of publishing a manuscript. “Western blots are easy to manipulate in an unsophisticated way,” said Chris Graf, Research Integrity Director at Springer Nature. “Researchers or paper mills might do that to beautify the data … or to fake it in the first place.”
Volume of Submissions
The sheer number of submissions journals receive makes ensuring integrity and proactively identifying these issues a challenge. Many journals receive far more articles than they can publish. Science, for example, received over 10,000 submissions in 2022 and accepted just 6.1 per cent of them.
Say a manuscript contains images of 50 Western bands. Reviewing those 50 bands for simple duplications requires 2,500 separate side-by-side comparisons. Manuscripts often contain dozens of subimages, sometimes hundreds, so just reviewing the figures for duplication errors requires a substantial time investment, let alone reviewing the study’s content or checking for less obvious image issues.
The challenge is compounded by the delicate patterns and subtle variations in band intensities, which often make Western blots look very similar to one another when reviewing by eye.
Proactive checks before publication offer significant benefits to journals and editors compared to reactive investigations. Streamlining checks prior to publication helps prevent problematic figures entering the literature, saves time and money, and protects the integrity of important research, but doing this manually is often unsustainable.
Automation
Using AI tools can significantly improve pre-publication image integrity checks by screening images and flagging any that require more detailed attention by a human reviewer.
Once images are uploaded to the platform, the tool scans them all, checking each first against itself and then all other images. AI tools can identify a range of issues, including duplication, rotation, flipping, cropping, splicing, and manipulations including deletion and insertion of bands. Any potential issues are then highlighted and flagged for the user, who can then investigate the tool’s findings and make a final decision about whether or not an issue has occurred.
As well as reducing the time image checks take by selecting images most likely to contain issues, AI tools can also improve the investigation process. By using advanced filters to highlight anomalies that are invisible to the human eye, AI tools help catch errors that reviewers might have otherwise overlooked. This can enhance the overall accuracy of the image integrity assessment.
In a trial of image checking tool, Proofig AI, by the American Association for Cancer Research (AACR), tool-assisted review of manuscripts that had been provisionally accepted for publication identified more than twice the number of issues in just over half the time compared to manual review alone. Such tools help editors and reviewers work more efficiently and allocate more time to other critical aspects of the peer-review process.
“There’s the opportunity to use technology to interrogate [Western blot] images and to identify the overlapping regions within the image, the scrub marks that have been left by the use of an eraser tool in Photoshop, for example, and we’re able to spot the fact an image has been manipulated,” added Graf.
These tools are not designed to replace human reviewers or to pass judgement. They offer time savings and efficiency improvements by highlighting potentially problematic images for editors to check in more depth. Identifying these issues at an early stage ensures that any unintentional mistakes are removed before publication, helping to prevent delay into what is a lengthy road from research to market approval while also maintaining public confidence in research.
While researchers are responsible for submitting work that conforms to the highest standards of scientific rigour, the probability of including problematic figures in pharmaceutical research is extremely high, without prior examination using specialised software. Implementing AI tools for pre-publication image analysis can assist journals and pharmaceutical and biotech companies in upholding the credibility of scientific research, instil confidence in new and emerging treatments, and help conserve resources.