Abstract
In the present study, we analyzed a large corpus of English-language online media articles covering genome-wide association studies (GWAS), exemplifying the use of computational methods to study science communication in biological sciences. We analyzed trends in media coverage, readability, themes, and mentions of ethical and social issues, in over 5,000 websites published from 2005 to 2018 from 3,555 GWAS publications on 1,943 different traits, identified via GWAS Catalog using a text-mining approach to inform the discussion about genetic literacy and media coverage. We found that 22.9% of GWAS papers received media attention but most were described in language too complex to be understood by the public. Ethical issues are rarely mentioned and mentions of translation are increasing over time. We predicted media attention based on year of publication, number of genetic associations identified, study sample size, and journal impact factor, using a regression model (r2 = 38.7%). We found that chronotype, educational attainment, alcohol and coffee consumption, sexual orientation, tanning, and hair color received substantially more attention than predicted by the regression model. We also evaluated the prevalence of the clickbait "one gene, one disease" headlines (e.g., "Scientists Say They've Found Gene That Causes Breast Cancer") and found that it is declining. In sum, online media coverage of GWAS should be more accessible, introduce more modern genetics terms, and when appropriate, ELSI should be mentioned. Science communication research can benefit from big data and text-mining techniques which allow us to study trends and changes in coverage trends across thousands of media outlets. Results can be explored interactively in a website we have built for this manuscript: https://jjmorosoli.shinyapps.io/newas/.