Search

Abstract The Coronavirus Disease of 2019 (COVID-19) pandemic caused by SARS-CoV-2 led the Spanish government to impose a national lockdown in an attempt to control the spread of the infection. Mobility restrictions and the requirement of a medical prescription for serological testing for COVID-19 were included among the control measures. Under this scenario, between April 15th and June 15th, 2020, we performed an observational study including 449 individuals allowed to be tested according to the governmental restrictions, i.e. fulfilling the following prescription requirements: manifestation of COVID-19-compatible symptoms, contact with a confirmed COVID-19 patient, or employment as an essential worker, including health care workers, firefighters and public safety personnel such as police. Importantly, a relevant feature of the studied cohort was that none of the participants had been hospitalized. We analyzed SARS-CoV-2 IgG seropositivity in this specific cohort, uncovering intrinsic features of great demographic interest. The overall rate of IgG seropositivity was 33.69% (95% CI: 29.27–38.21). This frequency was comparable among the different participant occupations. A RT-PCR positive test, contact with a household member previously tested positive and the presence of COVID-19-compatible symptoms were positively associated with IgG + results. Among these symptoms, ageusia/anosmia was positively and independently associated with SARS-CoV-2 IgG seropositivity, while odynophagia was inversely associated. However, fever, ageusia/anosmia and asthenia were the most frequent symptoms described by IgG + subjects. Therefore, our data illustrate how specific cohorts display particular characteristics that should be taken into account when studying population-wide SARS-CoV-2 seroprevalence and key defining symptoms of COVID-19.

Abstract Machine learning is widely applied in drug discovery research to predict molecular properties and aid in the identification of active compounds. Herein, we introduce a new approach that uses model-internal information from compound activity predictions to uncover relationships between target proteins. On the basis of a large-scale analysis generating and comparing machine learning models for more than 200 proteins, feature importance correlation analysis is shown to detect similar compound binding characteristics. Furthermore, rather unexpectedly, the analysis also reveals functional relationships between proteins that are independent of active compounds and binding characteristics. Feature importance correlation analysis does not depend on specific representations, algorithms, or metrics and is generally applicable as long as predictive models can be derived. Moreover, the approach does not require or involve explainable or interpretable machine learning, but only access to feature weights or importance values. On the basis of our findings, the approach represents a new facet of machine learning in drug discovery with potential for practical applications.