Project 1 Abstract
Title: An Exploratory Analysis to Assess reproducibility of COVID-19 bulk RNA-seq Datasets
Authors: Geordan Bolden, M.S.
Abstract
Integrating public transcriptomic datasets can significantly increase statistical power, yet technical batch effects and biological heterogeneity frequently compromise downstream meta-analyses. In this study, we evaluated the requirements, limitations, and cross-dataset reproducibility of integrated bulk RNA-seq workflows using five publicly available COVID-19 datasets. Utilizing the ComBat-Seq algorithm alongside principal variance component analysis (PVCA), we demonstrated that while technical batch effects from sequencing platforms and baseline depths were successfully minimized, the post-correction landscape resolved into two distinct macro-clusters. This persistent divergence cuts across tissue types and sequencing platforms, indicating fundamental differences in experimental design or clinical disease stages rather than technical artifacts. Differential expression and over-representation analyses (ORA) revealed zero universally shared differentially expressed genes (DEGs) across all five datasets, though a robust consensus core of 60 DEGs and 15 ORA terms tied to cell-cycle suppression was identified. When restricting analysis tightly to identical clinical contrasts (Healthy vs. COVID-19), we identified a conserved downregulation of B-cell-mediated immunity markers (IGHG1, IGHG3). Our findings demonstrate that robust batch-correction algorithms are necessary but insufficient to reconcile deep-seated discrepancies in sample composition and clinical phenotypic phases. Consequently, establishing reproducible biomarkers requires future meta-analyses to strictly harmonize clinical contrast specificity and capture the continuous biological gradients inherent to heterogeneous cohorts.
Small Changes will be made: Feel free to leave sugestions in the Issues tab