Editorial
General Science

Electrophoresis Images Diminish Credibility of Life Science Research An Examination of Image Manipulation in Academic Papers

Shinichiro Takezawa

Abstract

We investigated the frequency and tendency of “image manipulation” in electrophoresis experiments in the field of Life Sciences which includes Medicine, with the goal of improving the credibility of research papers. We reviewed image manipulation in 374 papers published in the journals Nature and PNAS. Of the 302 papers in Nature between 2000 and 2006, 20 (6.6%) papers were suspected to have a high likelihood of image manipulation. Combining those with the papers with indications of image manipulation totaled 72 (23.8%) complying papers. In PNAS in 2003 vol. 100(1) to (3), the comparison of papers with Japanese named authors and papers without Japanese named authors revealed that the rate of papers with Japanese names that required further investigation of image manipulation which means some falsification is suspected was 42.9%, significantly higher (P<0.05) than that without Japanese names which was 17.2%. The authors of these papers wherein image manipulation is indicated should reveal the actual data, replace graphs and images, or withdraw said papers in order to clarify the truth. Additionally, university posts obtained as a result of personnel evaluation focusing on impact factor should be vacated to honest researchers, if their publications were fabrication. The authors should be negatively evaluated by university authorities utilizing a fraud database as well as positively evaluated through impact factor.

Author and Article Information

Author infoGeneral healthcare Inc.

PublishedJun 6 2014

CitationTakezawa S (2014) Electrophoresis Images Diminish Credibility of Life Science Research—An Examination of Image Manipulation in Academic Papers. Science Postprint 1(1): e00024. doi: 10.14340/spp.2014.06E0001

Keyword

Introduction

General Healthcare Inc., which is the publisher of Open Access Journal “Science Postprint,” has created mechanisms to analyze image manipulations and provides contract service to defend against these types of fabrication in order to improve the credibility of academic papers. This service is targeted for the image recycling and image manipulation, both of them often seen in electrophoresis experiment papers in the field of Life Sciences including Medicine. Although there are a number of known methods of paper fabrication 1, image manipulation is considered the most frequent form of them and it can convey the great malignancy with which scientific facts are distorted. Therefore, we investigated the actual amount of published papers with image manipulation, collaborating with outside researchers. As a first step, we focused on the papers with Japanese named authors.

Methods

Acquired Data

We investigated the 302 papers from Japanese named authors containing electrophoresis images published in Nature from 2000-2006 and 72 papers containing those images published in volumes 100(1) – 100(3) of the Proceedings of the National Academy of Sciences of the United States of America (PNAS) from January to February 2003. The electrophoresis experiment images were extracted from each paper and evaluated for image manipulation potential.

Evaluation of Image Manipulation

The papers were acquired as PDF files, images were copied and pasted into the analysis software, and changes to the image were evaluated by varying contrast and gamma correction. The image manipulation level was evaluated in three point scales. A rating of 0 was defined as no image manipulation being detected. A rating of 1 meant the data suggested the possibility of image manipulation, but was not conclusive or the image manipulation had been confirmed, but the manipulated data did not affect the conclusions since they do not have any concern with the negative or positive control. Papers with high-contrast images and those wherein the background had been whited-out and could not be legitimately evaluated received point 1. In addition, papers wherein there was some possibility of fabrication, but with no clear evidence were evaluated at 1. An evaluation of 2 was designated for papers possessed a high likelihood of image manipulation. For example, this evaluation was presented if the cut and paste were seen visually by varying contrast within the same image.

Results

Table 1 indicates the results of the evaluation of academic papers by Japanese named authors published in Nature between 2000 and 2006 that include electrophoresis related. The result of the investigation of these 302 papers showed that 72 (23.8%) of them are suggested to have a possibility of image manipulation, and 20 (6.6%) out of those 72 cases possessed a high likelihood of image manipulation. Of the 20 papers to receive an evaluation of 2, the corresponding authors of 6 belonged to research institutions in Japan and 11 to research institutions in the United States. On the other hand, the first author of 10 of those papers had Japanese names. We suspect the Japanese authors of 4 papers to have been immigrants or studying in the United States. Further, 2 of those 4 currently hold positions as professors in Japan as of 2014. In addition, of those 20 papers to receive an evaluation of 2, there were 2 cases in which 2 papers with the same first author is detected. These two authors belonged to institutions in the United States.

Table 1 Overview of the papers with Japanese named researchers concerning electrophoresis Image manipulation published in Nature between 2000 and 2006.

Table1

We inspected PNAS to see whether or not possibility of fabrication via image manipulation was affected by the journal. An evaluation was performed of all images in the 72 papers with electrophoresis-related image data published in PNAS in issues 100(1)–100(3) in 2003. As the results show in Table 2, 56 (77.8%) were evaluated at 0, 13 (18%) at 1, and 3 (4.2%) at 2. The rate of papers in PNAS for that period that required further investigation into image manipulation was 22.2%. The papers were then tabulated separately between papers with Japanese named authors and those without. A significant difference (P<0.05) was found between the numbers of the papers with Japanese named authors evaluated at 1 or 2 and that without Japanese named authors. 42.9% of the papers by Japanese named authors required further investigation being evaluated at point 1 or 2, and on the other hand, 17.2% of papers by non-Japanese named authors required it. There was no significant difference seen in the amount of papers that requires further investigation into image manipulation between papers in Nature and that in PNAS.

Table 2 Overview of the all papers concerning electrophoresis image manipulation published in issues 100(1)–100(3) of PNAS in 2003.

Table2

Discussion

The Reality of Image manipulation

In this study, we evaluated the falsification of electrophoresis images published in international academic journals since 2000. As a result, we concluded that 6.6% of the papers published in Nature possessed a high likelihood of image manipulation occurrence, and that 23.8% should be further examined for image manipulation. These results show that the amount of paper fabrications published in the media and elsewhere are only the tip of the iceberg, and that there are numerous researchers out there that have yet to be discovered despite fabricating results similarly. Although there would be some cases among papers evaluated at point 1 where image contrast had been already high for raw data and backgrounds had been already whited out, we suspect that in many cases the backgrounds had been intentionally erased using image software, and evaluated those cases at 1. Therefore, we do not believe that all papers with a rating of 1 were the result of malicious image manipulation. Additionally, there were several papers that pasted a separate image just for control together with data images to appear as one photo. Such cases received a rating of 1. Strictly speaking, if the control is clearly expressed in a separate frame, there was no problem. However, in quantitative experiments, since we consider control as a standard for quantitative comparison, it is impossible to discuss quantitative properties when images from other experiments are included. Such cases were rated at 2.

A Request for Authors of Papers Wherein Our Investigation Indicates Image Manipulation

An evaluation of 2 was given to papers where it was suggested that the likelihood of image manipulation was high. If the authors of these papers were to present their raw data, correct data, or clear data, the scientific value of their papers would not be called into question. Similarly, this survey does not mean to disparage the credibility of those papers rated at 1. However, it may be difficult to produce the raw data for some studies where 10 years or more have passed since the original experiments. In those cases, it would be preferred if the experiments were performed again, and similar photographs were presented. In cases where the experiments cannot be performed again, the authors should explicitly show in another paper what results could be reproduced and to what extent therein. If results cannot be reproduced or fabrication is admitted, the only option for the paper is to be withdrawn. In our opinion, the important aspect here in regard to science is not to condemn fabrication or the researchers, but to clarify the data can be trusted and the data that cannot. As described above, we hope the authors of these papers wherein there is a likelihood of image manipulation clarify whether our claims of image manipulation are groundless or whether fraud really occurred, via scientific discussion based on the data. What is believed to be nonsense in paper examination back then might be revealed to be possible. True science may be hiding within phenomena overlooked in these papers.

The Background

Although situations varied for each author of the 20 cases in Nature evaluated at 2, from the analysis of all the cases in PNAS, the difference in rate of image manipulation indications when comparing Japanese named researchers to others suggests this issue may be a particular characteristic of Japanese researchers. Additionally, although we hypothesized the image manipulation rate might increase as the impact factor of the journal increased, there was no significant difference between Nature and PNAS at this time.
Considering the context in which Japanese first authors manipulate their images, there may have been some pressure. For example, he or she could have been a researcher studying abroad, to produce results in order to obtain a post in Japan. Or perhaps the author was focused on his or her own self-interest in obtaining honor and status. In addition, he or she may just have wanted to produce better looking images in order to pass the paper examination quickly. It can be associated with the severe Japanese research environment, in which the employment demand for researchers kept low against the increase in the number of Ph.Ds which is the results of national policy.

It is also possible that a corresponding author involved in the image manipulation as a group. Of the 20 cases with a rating of 2 from our examination, image manipulation was detected in 2 cases each from 2 laboratories. Both of those labs were in the United States. Further investigation is recommended including the potential of organizational fabrication.

Conclusions

Doubt regarding the credibility of these papers due to fabrication like image manipulation needs to be clarified through scientific discussion and the presentation of actual data. It is the responsibility of these authors to disclose the facts. Further, posts obtained through academic fraud should be given up to more honest scientists. Without justifying paper fabrication, the authorities of the universities and research institutions with the authority to manage personnel affairs of these researchers should keep a database of researcher fraud at hand, and consider discounting fraudulent data even if the evaluation by impact factor is positive.

Reference

  1. Takezawa S (2014) Aiming to Improve on the Reliability of Research Papers for the Good of Scientific Progress. Science Postprint 1(1): e00023. doi: 10.14340/spp.2014.05E0005.
Editorial is an opinion article on a topical issue. Editorials are not peer reviewed and only receive editing from internal editors. Editorial office may solicit Editorial article from an appropriate member of SPP Editorial Board.