COVID-19 has caused millions of deaths globally; however, the characterization of molecular biomarkers of severe disease remains of great scientific importance. The aim of this study was to capture the transcriptional differences of the whole blood gene expression between COVID-19 patients with mild and severe disease, using Next Generation Sequencing technologies, on admission and after 7 days. The genes which were differentially expressed in severe compared to mild patients were used for Gene Ontology (GO) enrichment analysis. Gene expression data were used to estimate the cell abundance of 22 immune cell types via digital cytometry. GO terms related to the response to molecules of bacterial origin, such as intestine-derived lipopolysaccharide (LPS), were enriched, among other dysregulated pathways, which are well described as paramount mechanisms of severe manifestations of COVID-19. The neutrophil population increased in patients with severe disease, whereas the monocyte, CD8⁺ T cell, and activated Natural Killer (NK) cell populations were depleted. These cell population dynamics are also indicative of severe COVID-19 and intestinal bacterial translocation. This study elucidates the molecular basis of severe COVID-19 and highlights intestinal bacterial translocation as a potential driver of severe disease.

Background: False-positive variant calls generated within low-complexity regions (LCRs) remain a persistent bottleneck in clinical genomics, complicating downstream analysis. This study evaluates a targeted spatial masking strategy designed to suppress deterministic artifacts in short-read sequencing data, while preserving clinically actionable variants residing outside LCRs. We implemented a selective masking protocol prior to variant calling across analytical reference standards (EQA, NA12878) and two independent breast cancer whole-exome sequencing cohorts (n = 25). Methods: Callsets were evaluated for diagnostic sensitivity, precision gains, mutational signatures, VAF behavior, pseudo-multiallelic noise and ClinVar/dbSNP annotation. Results: The protocol removed thousands of sequencing and alignment artifacts while maintaining the retained biological callset, with negligible disease-associated diagnostic variants detected in the excluded artifact fraction. LCR masking preserved physiological Ti/Tv and Ins/Del profiles in retained calls, resolved pseudo-multiallelic noise, and distinguished excluded artifact calls by distorted mutational and VAF signatures. dbSNP profiling showed cohort-dependent behavior: TCGA-BRCA reproduced an intriguing phenomenon, with excluded calls showing higher dbSNP annotation than retained calls, whereas AURORA showed the opposite direction. Conclusions: These findings demonstrate the potential vulnerability of one-dimensional database annotation for variant authentication and highlight targeted spatial filtration as a critical, early pipeline intervention for high-fidelity clinical genomics of non-LCR-associated germline variants using short reads.