Abstract
Meta-analyses increase statistical power by combining statistics from multiple studies. Meta-analysis methods have mostly been evaluated under the condition that all the data in each study have an association with the given phenotype. However, specific experimental conditions in each study or genetic heterogeneity can result in “unassociated statistics” that are derived from the null distribution. Here, we show that power of conventional meta-analysis methods rapidly decreases as an increasing number of unassociated statistics are included, whereas the classical Fisher’s method and its weighted variant (wFisher) exhibit relatively high power that is robust to addition of unassociated statistics. We also propose another robust method based on joint distribution of ordered p-values (ordmeta). Simulation analyses for t-test, RNA-seq, and microarray data demonstrated that wFisher and ordmeta, when only a small number of studies have an association, outperformed existing meta-analysis methods. We performed meta-analyses of nine microarray datasets (prostate cancer) and four association summary datasets (body mass index), where our methods exhibited high biological relevance and were able to detect genes that the-state-of-the-art methods missed. The metapro R package that implements the proposed methods is available from both CRAN and GitHub (http://github.com/unistbig/metapro).
Highlights
Meta-analyses increase statistical power by combining statistics from multiple studies
As p0-values were included, the power of Lancaster, Z-method, and weighted Z-method rapidly decreased, while Fisher, Weighted Fisher’s method (wFisher), and ordmeta exhibited a relatively slow decline. These results demonstrate (1) Fisher and Z-method are quite different methods, the latter uses Z-scores transformed from p-values and (2) wFisher is superior to the original Fisher’s method, both with or without p0-values. Another weighted Fisher’s method, Lancaster showed a rapid decline as opposed to Fisher or wFisher
We focused on the hypothesis of meta-analysis that one or more studies involved are associated, whereas conventional meta-analyses have assumed all or most of the studies are associated when testing their performance
Summary
Meta-analyses increase statistical power by combining statistics from multiple studies. We show that power of conventional meta-analysis methods rapidly decreases as an increasing number of unassociated statistics are included, whereas the classical Fisher’s method and its weighted variant (wFisher) exhibit relatively high power that is robust to addition of unassociated statistics. We propose another robust method based on joint distribution of ordered p-values (ordmeta). Many p-value or Z-score combining methods including Fisher’s method take the null hypothesis that the true effect in each of the combined datasets is zero[2] This suggests a high sensitivity of the methods even when only a subgroup of the combined datasets have a nonzero effect size. With a positive (or negative) effect, and unassociated p-value (or p0-value), the uniformly distributed under the null hypothesis
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
