Marco Korcak
Professor Vejdemo-Johansson
MHC 223
October 18, 2021
Identifying Human Infecting Viruses
Health is an important aspect of life that is constantly being observed and advanced as time goes on. The medical field relating to the human body is constantly changing and evolving with new information being discovered and published. With the recent outbreak of the Coronavirus, many people globally have been impacted, showing the world that dangers such as viral infections cannot be overlooked. Although the pandemic had numerous negative impacts on the world and its economies, it also resulted in new research being conducted and findings published. The New York Post published an article titled “AI used to predict which animal viruses are likely to infect humans: study” which was written by Julia Musto. The article referred to the peer reviewed research article titled “Identifying and prioritizing potential human-infecting viruses from their genome sequences”. This article was published by a group of researchers based in Glasgow, Scotland. The article stated that the researchers developed machine learning models to single out candidate zoonotic viruses. A zoonosis is an infectious disease that is transmitted between species from animals to humans. The article stated that a dataset of 861 viral species was used to collect a single representative genome sequence from the hundreds of RNA and DNA virus species. The article reported that the findings from the research article was that 72% of the viruses that predominantly or exclusively infect humans were correctly identified. The article also stated that the research separated results by differing levels of zoonotic potential. 77.2% of the viruses were predicted to have a very high zoonotic potential. When examining a second set of data that was comprised of 645 animal-associated viruses data showed that “results are consistent with the expectation that the relatively close phylogenetic proximity of nonhuman primates may facilitate virus sharing with humans” (Mollentze, 11). Overall, this article summarized the study done and explained that this information will evolve as time passes and more viruses are characterized.
“Identifying and prioritizing potential human-infecting viruses from their genome sequences” is a research article published by a team of researchers based in Glasgow, Scotland. The authors included Nardus Mollentze, Simon A. Babayan, Daniel G. Streicker and many more. The research article was published on September 28, 2021, after numerous peer reviews and has received media coverage from 53 news outlets and 2 blogs. The abstract of the research states that the goal was to determine which animal viruses may be capable of infecting humans to allow for outbreak preparedness and early investigation. The researchers developed a machine learning model to that can identify candidate zoonoses solely using signatures of host range encoded in viral genomes since there is an increasing use of genomics in virus discovery. The researchers collected a single representative genome sequence from 861 RNA and DNA virus species spanning 36 viral families. This was then used with three published datasets that reported human infectivity at the virus species level. Using a dataset of 861 viral species, the machine learning model outperformed other models based on the phylogenetic relatedness of viruses and the researchers also identified a new high-risk coronavirus strain. On a second set of data with 645 animal associated viruses the machine learning model was able to identify 272 high and 41 very-high risk candidate zoonoses. There were two additional cases studies done during this research to illustrate the utility of the prediction framework. In essence, this machine learning model was significant because it provides a rapid, low-cost approach to enable evidence-driven virus observation.
After reading the New York Post article and research article itself, it is evident that there are numerous similarities, but many discrepancies present as well. In terms of statistics and data, the article matches up with the research paper and overall research. The article stated some of the results of the research paper but failed to explain the process taken to get to the results as well as other significant findings. In order to properly get results the researchers needed to account for the fact that viruses change over time and adapt. The research article greatly emphasized the process that was taken in order to account for this change which was done by reconstructing taxonomy more accurately. This was done because genome composition features only partly tracked evolutionary history so much more detail was needed than available. Once this was done the researchers also carried out two case studies to illustrate the utility of the prediction framework. This was done by using 758 virus species that were not present in the training data and they also used a beta regression model to explore how predictions of zoonotic potential varied among host and viral groups. This process was very significant in the research and should have been emphasized in the article published by the New York Post but it was not. The post also failed to mention other valuable statistics and processes such as the SHapley Additive exPlanations (SHAP) algorithm. This computed the Shapley value for each feature and is increasingly used to improve the interpretability for the decisions made by the machine learning models. This was also very influential to the research article and was not mentioned in the New York Post article. In essence, there were similarities between the research article and the New York Post article, but the article did fail to mention very important aspects of the research process.
Works Cited
Mollentze, Nardus, et al. “Identifying and Prioritizing Potential Human-Infecting Viruses from Their Genome Sequences.” PLoS Biology, vol. 19, no. 9, Sept. 2021, pp. 1–25. EBSCOhost, doi:10.1371/journal.pbio.3001390.
News, F. (2021, October 4). AI used to predict which animal viruses are likely to infect humans: Study. New York Post. Retrieved October 12, 2021, from https://nypost.com/2021/10 /04/ai-used-to-predict-which-animal-viruses-are-likely-to-infect-humans-study/.
