Genome-wide Association Study of Long COVID (preprint)

Infections can lead to persistent or long-term symptoms and diseases such as shingles after varicella zoster, cancers after human papillomavirus, or rheumatic fever after streptococcal infections(1,2). Similarly, infection by SARS-CoV-2 can result in Long COVID, a condition characterized by symptoms of fatigue and pulmonary and cognitive dysfunction(3-5). The biological mechanisms that contribute to the development of Long COVID remain to be clarified. We leveraged the COVID-19 Host Genetics Initiative(6,7) to perform a genome-wide association study for Long COVID including up to 6,450 Long COVID cases and 1,093,995 population controls from 24 studies across 16 countries. We identified the first genome-wide significant association for Long COVID at the FOXP4 locus. FOXP4 has been previously associated with COVID-19 severity(6), lung function(8), and cancers(9), suggesting a broader role for lung function in the pathophysiology of Long COVID. While we identify COVID-19 severity as a causal risk factor for Long COVID, the impact of the genetic risk factor located in the FOXP4 locus could not be solely explained by its association to severe COVID-19. Our findings further support the role of pulmonary dysfunction and COVID-19 severity in the development of Long COVID.

Health, socioeconomic and genetic predictors of COVID-19 vaccination uptake: a nationwide machine-learning study (preprint)

Reduced participation in COVID-19 vaccination programs is a key societal concern. Understanding factors associated with vaccination uptake can help in planning effective immunization programs. We considered 2,890 health, socioeconomic, familial, and demographic factors measured on the entire Finnish population aged 30 to 80 (N=3,192,505) and genome-wide information for a subset of 273,765 individuals. Risk factors were further classified into 12 thematic categories and a machine learning model was trained for each category. The main outcome was uptaking the first COVID-19 vaccination dose by 31.10.2021, which has occurred for 90.3% of the individuals. The strongest predictor category was labor income in 2019 (AUC evaluated in a separate test set = 0.710, 95% CI: 0.708-0.712), while drug purchase history, including 376 drug classes, achieved a similar prediction performance (AUC = 0.706, 95% CI: 0.704-0.708). Higher relative risks of being unvaccinated were observed for some mental health diagnoses (e.g. dissocial personality disorder, OR=1.26, 95% CI : 1.24-1.27 ) and when considering vaccination status of first-degree relatives (OR=1.31, 95% CI:1.31-1.32 for unvaccinated mothers) We derived a prediction model for vaccination uptake by combining all the predictors and achieved good discrimination (AUC = 0.801, 95% CI: 0.799-0.803). The 1% of individuals with the highest risk of not vaccinating according to the model predictions had an average observed vaccination rate of only 18.8%. We identified 8 genetic loci associated with vaccination uptake and derived a polygenic score, which was a weak predictor of vaccination status in an independent subset (AUC=0.612, 95% CI: 0.601-0.623). Genetic effects were replicated in an additional 145,615 individuals from Estonia (genetic correlation=0.80, 95% CI: 0.66-0.95) and, similarly to data from Finland, correlated with mental health and propensity to participate in scientific studies. Individuals at higher genetic risk for severe COVID-19 were less likely to get vaccinated (OR=1.03, 95% CI: 1.02-1.05). Our results, while highlighting the importance of harmonized nationwide information, not limited to health, suggest that individuals at higher risk of suffering the worst consequences of COVID-19 are also those less likely to uptake COVID-19 vaccination. The results can support evidence-informed actions for COVID-19 and other areas of national immunization programs.