Problems in Harvard Medical School studies include images taken from other researchers’ papers and vendor websites

Over the last week I’ve been analyzing a set of papers from a research group at Brigham and Women’s Hospital, which falls under the Harvard Medical School. The papers have the usual image problems we unfortunately encounter so often, such as duplicated photos of mice and overlapping Western blots.

But this set also includes a 2022 paper that appears to have copied/pasted several figure panels from other researchers and even from scientific vendors. Most of these problems were found with ImageTwin or reverse image searches.

Quick links [the spreadsheet and slide have been updated with an additional problem found on Feb 2, 2024]:

• spreadsheet with the 28 29 papers found with PubPeer links [Excel file]
• slide deck with the 58 59 image problems found in the 28 29 papers [PDF].

A tip about questionable research practices at a Harvard Medical School lab

Two months ago I received a tip about sloppy and potentially fraudulent work at a research group at Brigham and Women’s Hospital, which is part of the Harvard Medical School. It was a story I have heard too many times before, about sloppiness in storing samples, contaminated cell lines, and bullying and threatening of PhD students and post-docs on visas to cut corners and to produce desired results. The tipster did not have a particular paper or problem for me to look at, and often such questionable research practices might not be visible by just looking at the papers produced by a lab.

But the tipster’s story was alarming enough for me to investigate some recent papers by the principal investigator (PI) of the lab, Dr. Khalid Shah. Dr. Shah has an impressive number of simultaneous jobs, any one of which might be enough to fill his day.

According to this website, ‘Dr. Khalid Shah is the Vice Chair of Research for the Department of Neurosurgery at BWH and a Professor at Harvard Medical School. He directs The Center for Stem and Translational Immunotherapy and also a joint Center of Excellence in Biomedicine. He is also the Principal Faculty at Harvard Stem Cell Institute in Boston.’ In addition, he also received a grant for ‘research and training of students from Saudi Arabia’ and ‘leads the collaborative project with Sunway University from Malaysia’. Finally, he has founded two biotech companies, one of which is AMASA Therapeutics.

He certainly sounds a busy man, with very little time left to attend to his research lab.

The Shah lab investigates gene editing as a way to treat cancer, where the genome of cancer cells or stem cells is altered to express therapeutics. The cells are then put back into the patient to kill other cancer cells, and perhaps even prevent future cancer growth. His work has been featured by BBC News, The Scientist, and WCVB news.

Dr. Shah has received several awards, including the ‘Harvard Young Mentor Award’ and a 2021 ‘Pillars of Excellence Award’ from Mass General Brigham in the area of “Integrating Diversity, Equity & Inclusion.” He also got the ‘young investigator award from Alliance for cancer gene therapy (ACGT), Research fellow award from American Cancer Society (ACS), Distinguished research award from Academy of Radiology and Innovation awards from James McDonnell Foundation, American Brain Tumor Association (ABTA) and Goldhirsh foundation’.

Funding for his lab has been mainly through the National Institutes of Health’s National Cancer Institute, amounting to millions of dollars.

A surprise finding in a 2022 Nature Communications paper

One of the first papers I checked for image problems was a 2022 Nature Communications paper by Deepak Bhere et al., Target receptor identification and subsequent treatment of resected brain tumors with encapsulated and engineered allogeneic stem cells, DOI: 10.1038/s41467-022-30558-3 [PubPeer post].

This paper, with 33 (!) authors, was featured on EurekAlert with the headline ‘Scientists develop ‘off the shelf’ engineered stem cells to treat aggressive brain cancer’. The article stated that mice with cancer receiving the treatment had a much longer survival than untreated mice, without any signs of toxicity.

I did my usual ImageTwin scan to see if any of the figure panels were duplicated or overlapping. What I found was a huge surprise.

Several images showing mouse tissues matched those of older papers from completely different groups.

ImageTwin compares each image in an uploaded scientific manuscript to a large database of Open Access images from scientific papers. It sometimes finds matches with older papers from the same group, perhaps recycled old experiments (not great) or presented as different experiments (bad), and it occasionally finds matches between two unrelated papers (very bad). Usually these unrelated papers with matching images are from paper mills, which might sell the same image or overlapping images to different buyers. It was completely unexpected to find images from a Harvard lab matching images published by a different set of authors working in a different lab.

Images copied from older, unrelated papers?

Two glioblastoma tissue panels in Figure 1e of the 2022 Nature Comms paper matched those of a 2010 PLOS ONE paper by Elrod et al., from researchers at Emory University School of Medicine, where they represented head and neck squamous cell carcinoma tissues. These panels are shown below marked with red and green boxes. The older paper is on the left, and the 2022 paper is on the right. There are no overlapping authors or affiliations, and the older images are of higher resolution and are less cropped than the 2022 images, suggesting that they were the originals.

One panel in Figure 1f matched another in a 2016 Journal of Translational Medicine paper published by researchers at the University of South Florida and the Medical University of South Carolina.

Another Figure 1f panel was taken from a 2018 Neuro-Oncology paper from the same Harvard Medical School researchers, but again appeared to be showing a different experiment.

Supplemental Figure 4b has nine (!) panels that ImageTwin flagged for duplications. These panels came from the five following papers:

• Xuejiao Liu et al., Journal of Experimental & Clinical Cancer Research (2019), DOI: 10.1186/s13046-019-1235-7, from researchers at Xuzhou Medical University in China (two panels)
• Jinling Wang et al., International Journal of Nanomedicine (2016), DOI: 10.2147/IJN.S113882, from Beijing University of Chinese Medicine (four panels)
• Peng Miao et al., Neural Plasticity (2016), DOI: 10.1155/2016/3258494, from Shanghai University (one panel, Ischemic Injury in mice)
• Musi Ji et al., Experimental and Therapeutic Medicine (2017), DOI: 10.3892/etm.2017.5327, from Guangdong Medical University (one panel; representing aging in dogs)
• Yuanyuan Xu et al., G3 (2018), DOI: 10.1534/g3.118.300448, from Jilin University (one panel; representing a diabetes model in rabbits)

Note that the images in these older papers are higher-resolution; published years before the Bhere (2022) paper came out; published in mostly Open Access journals (meaning the images were included in the ImageTwin database); and not always representing mice.

Images copied from laboratory suppliers?

Now it got weirder. I started to suspect that other panels in the 2022 Bhere et al. paper may also have been copied from other papers. Perhaps some were copied from articles that were not included in the open-access database, so ImageTwin would have been unable to find them. How could I find more of these matches?

I decided to do a Google reverse image search for each of the remaining panels. This did not yield scientific papers, but there were new surprises.

Three panels matched images of tissues and cells featuring products sold by suppliers of laboratory materials and chemicals.

One panel in Figure 1F matched an image on the Fisher Scientific website, featuring a high-resolution photo of a tissue sample stained with an antibody for Butyrylcholinesterase, a different protein than the DR5 that it represents in the 2022 paper. Fisher Scientific is an Amazon-like vendor for laboratories, selling antibodies, cell lines, glassware, labels, racks, chemicals, and equipment.

Two panels in Supplemental Figure 2a yielded matches with photos of human bone marrow-derived mesenchymal stem cells, a product advertised on the website of ScienCell (for $845, in case you wondered).

In summary, at least 12 panels in this paper appear to have been copied from unrelated papers; one panel from one of their own older papers; and three panels from scientific vendors. There was also an overlap between panels in Supplemental Figure 5a (shown on PubPeer, not here), bringing the total number of problematic images in this paper to 17.

Other problematic papers from the Khalid Shah lab

I did not find other papers from the Shah lab that appeared to contain copied images from vendors or random other papers. But I did discover 27 28 additional papers containing the regular range of image problems, such as duplicated and overlapping photos, or photos reused from papers by the same authors.

One of the oldest papers in this set is Shah et al., Role of Threonines in the Arabidopsis thaliana Somatic Embryogenesis Receptor Kinase 1 Activation Loop in Phosphorylation, Journal of Biological Chemistry (2001), DOI: 10.1074/jbc.M102381200, when a young Shah worked at Wageningen University in the Netherlands on plant research.

Figure 4 appears to show a set of duplicated lanes in the bottom panel blot.

There are several papers in which photos of mice injected with tumor cells appear to have been reused. This suggests sloppiness in the correct labeling of experiments and photos, which is particularly bad in my opinion when it involves work with animals.

In one 2018 paper showing glioblastoma cell lines, two panels appear to have been reused from a 2017 paper where they represented colorectal cancer and lung cancer cell lines, clearly different experiments.

There are several more problematic papers from this group, which I have compiled in these links (I will update them if additional problems come to light).

• spreadsheet with the 28 29 papers found with PubPeer links [Excel file]
• slide deck with the 58 image problems found in the 28 29 papers [PDF].

Media coverage

• Top Harvard Medical School Neuroscientist Accused of Research Misconduct – Veronica H Paulus and Akshaya Ravi – The Harvard Crimson
• Harvard Professor’s Papers Contain Copied Images, Science Sleuth Claims – Nidhi Subbaraman – Wall Street Journal