Peer review is a valuable component in the research process, but it also lengthens the time to publish research. The need to rapidly communicate scientific findings has been especially apparent during the COVID-19 pandemic, which has led to an increase in the number of publications disseminated via preprint servers. With the lack of traditional peer review, the quality of these publications can be questionable. Associate Professor Halil Kilicoglu and the Automated Screening Working Group are working to assess COVID-19 preprints for rigor and transparency in their reporting. The broad goal of this group of scientists and software engineers is "to process every manuscript in the biomedical sciences as it is being submitted for publication, and provide customized feedback to improve that manuscript." The preprint screening leads to faster assessment of important medical research, such as findings related to COVID-19.
According to Kilicoglu, the screening process is currently limited to BioRXiv and MedRXiv, two primary preprint servers for biomedical science.
"Most COVID-19 related research first appeared in one of these servers, including some controversial claims like the one that proposed hydroxychloroquine to treat COVID-19," he said. "The process is to monitor these sources for new publications daily, download them, extract text or images needed by the individual tools in the screening pipeline, run each tool on the extracted data, and consolidate their results in a report, which is tweeted out as a comment. Also, a browser plugin called Hypothes.is displays the report on the preprint page, assuming the user has the plugin installed. This is all done automatically."
Kilicoglu is a coauthor on the article, "Automated screening of COVID-19 preprints: Can we help authors to improve transparency and reproducibility?," which was published in Nature Medicine (January 2021).
His contribution to the Automated Screening Working Group is the development of a tool that identifies whether authors are acknowledging the limitations or weaknesses of their work. He is also involved in an effort to assist the group in better recognizing the study types of the preprints (e.g., modeling study, randomized controlled trial, systematic review).
"Some tools in the pipeline may not be appropriate for a particular type of study, so we're trying to get better at only using the relevant tools to screen a particular publication," he said.
Kilicoglu's research interests include biomedical informatics, natural language processing, computational semantics, literature-based knowledge discovery, scholarly communication, science of science, and scientific reproducibility.
"In my own research, we are currently developing methods to check randomized controlled trial (RCT) manuscripts/publications for reporting quality and moving toward extracting fine-grained information that can help us make better sense of the methodological quality of the study," he said. "Since RCTs are considered to provide the most reliable kind of clinical evidence, understanding their quality is particularly important."
Prior to joining the iSchool faculty, Kilicoglu worked as a staff scientist at the U.S. National Library of Medicine, National Institutes of Health, where he led the Semantic Knowledge Representation project. He holds a PhD in computer science from Concordia University.