Circulating MicroRNA ratios as diagnostic biomarkers for long COVID with an independent validation cohort

Abstract Background Long COVID is a multisystemic condition with heterogeneous clinical presentations lacking diagnostic biomarkers with discriminatory capacity. This study screened circulating microRNAs as candidate biomarkers for long COVID detection by miRNA sequencing, followed by targeted qPCR quantification focusing on miRNA ratios to reduce technical variability. Methods Overall, 206 long COVID patients ( > 12 weeks after acute infection) and 71 uninfected, unvaccinated healthy controls were recruited through our prospective POSTCOV registry. Candidate microRNAs were screened by microRNA sequencing and quantified by targeted quantitative polymerase chain reaction, focusing on microRNA ratios expressed as cycle-threshold differences. In a training cohort, we developed a logistic regression model based on two microRNA ratios with age and sex as covariates. Internal validation was evaluated using cross validation and bootstraping. We then test the model in an independent external validation cohort processed with the same normalization and ratio definition. Results Here we show both microRNA ratios (hsa-miR-484/hsa-miR-126-3p, hsa-miR-345-5p/hsa-miR-223-3p) are inversely associated with long COVID. Discrimination in the training cohort is moderate (AUC: 0.760), with acceptable calibration (intercept ~0, slope ~1, Brier score 0.131). In external validation, discrimination decreases (AUC: 0.676) and calibration reveals systematic risk overestimation (intercept −1.88, slope 0.77). Nevertheless, both ratios are consistently lower in long COVID compared to controls. Conclusion Consistent between-group ratio differences support these microRNAs as candidate biomarkers, whereas limited external model performance highlights the challenges of transportability in heterogeneous long COVID cohorts. Target enrichment indicates inflammatory signaling pathways and integration with drug perturbation profiles suggests compounds for future validation and repurposing studies.