SCOURGE project

The SCOURGE Consortium was established as a collaboration between Spain and some countries in Latin-America, aimed to find genetic variants associated with COVID-19 infection and severity. It was a major effort that went on for 4 years, with two main research lines and several sub-projects, that was also part of the COVID-19 Host Genetics Initiative (HGI Consortium). I joined this project as a research technician while finishing my MSc, and then developed a PhD thesis based on it.

The two main investigations focused on Spanish and Latin-American patients, sharing first authorship with Dr. Cruz in both of them (Github of the project).

Spanish cohort study

In the first part of the research, we recruited over ~12.000 patients infected with the virus around 25 cities in Spain, obtaining their clinical information (comorbidities, COVID-19 outcomes, complications…) and genetic information. Then, we conducted a genome-wide association study (GWAS), which sweeps the entire genome in search of variants associated with the tested phenotype (case/control study). One of the key findings was a positive association between a variant in the gene codifying for the Aquaporin 3 (AQP3) and hospitalization due to COVID, suggesting a potential therapeutic target for the disease. We also conducted sex-stratified analyses, which revealed an absence of genetic associations in females, likely driven by lower effect sizes. In fact, after meta-analyzing several studies, the widely replicated signal in chromosome 3 emerged in this group. Lastly, we built a genetic risk score using the main associated variants and fit a multinomial regression for a COVID-19 severity scale, obtaining a significant differentiation (in terms of genetic risk) between classes asymptomatic/mild, moderate+severe disease, and critical disease.

Latin-American cohort study

We recruited more data in collaboration with Brazil, Colombia, Ecuador, Mexico and Paraguay. Then, we retrieved the patients recruited in Spain with a strong evidence of origin in Latin-America, resulting in over ~3.500 patients and conforming the largest Latin-American cohort to study the genetics of the disease. A meta-analysis was conducted with all data available from other studies. Two interesting associations emerged, one within the gene DDIAS, with a low-frequency variant which was mostly absent in other population groups, and another one within the gene BAZ2B. Following our previous research, we built a genetic score with the variants associated with COVID-19 severity and tested its performance in this novel cohort.

Polygenic scores

In the final phase of my PhD thesis, I focused on constructing more sophisticated polygenic risk scores (PRS) and evaluating their performance and clinical utility across two distinct cohorts. I developed a total of 130 unique polygenic risk scores for each cohort, leveraging various combinations of genetic data, and identified the top-performing models based on their improvement in pseudo-R2 values. Contrary to what’s tipically observed when applying polygenic risk scores to diverse population groups, the COVID-19 PRS showed similar performance in both cohorts, highlighting its potential generalizability.

Following this finding, I imagined a clinical scenario where an individual wanted to understand their position within the genetic risk distribution for COVID-19. To address this, I developed a single PRS and evaluated three distinct training models for risk prediction, using a new cohort of Spanish individuals as the test set. The three approaches included: (1) training the PRS regression model exclusively on the Spanish cohort, (2) training it on the entire combined cohort, and (3) a novel approach where the training cohort was selected based on the Euclidean distance between the principal genetic components (PCs) of the test set and those of the training set (best-fit model).

Lastly, I explored the behaviour of the genetic score in relation to key clinical variables, such as age, sex, and relevant comorbidities, and its association with other COVID-19-related complications. I also assessed the utility of incorporating the score into a prediction model for COVID-19 hospitalization, particularly when pre-existing comorbidities were included, by utilizing decision curve analysis (DCA) and net reclassification improvement (NRI) metrics.

This research was not published beyond my thesis dissertation, but here you can download the slides for a presentation I gave at the Health Research Institute of Santiago de Compostela (IDIS) (research was not finished yet).