hStouffer: the enhanced meta-analysis method for the comprehensive analysis of large-scale RNA-seq data

With the exponential growth of public RNA-seq datasets, meta-analysis has become a crucial tool for integrating studies to increase statistical power and identify consistent biological patterns. However, conventional p value combination methods were not designed for large-scale integration and exhibit a critical flaw: as more datasets are added, the rate of false positives increases dramatically. This issue stems from the disproportionate influence of extremely low p values from just a few individual studies, which can create a misleading signal of overall significance and undermine the reliability of the findings. Here, we introduce a robust meta-analysis framework that incorporates p value capping, cutoff thresholding, and bagging to alleviate this issue. We implemented this framework using Stouffer’s method, termed the hybrid Stouffer (hStouffer) method. The proposed method demonstrates a reduced false positive rate while preserving high sensitivity. Furthermore, the validation showed that the DEGs identified by hStouffer accurately reflect the underlying biological phenomena, making it an essential tool for leveraging the full potential of expanding genomic databases to understand complex biological processes. The hStouffer method provides a statistically robust and biologically coherent solution for large-scale RNA-seq meta-analysis. By effectively controlling for technical artifacts and false discoveries, it enables researchers to extract more reliable and meaningful insights from complex, aggregated transcriptomic data.