Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.1101/759027

Test-statistic correlation and data-row correlation

Di, Y.·Jiang, D.·Zhuo, B.
View original source ↗Export metadata ↓

Abstract

When a statistical test is repeatedly applied to rows of a data matrix--such as in differential-expression analysis of gene expression data, correlations among data rows will give rise to correlations among corresponding test statistic values. Correlations among test statistic values create many inferential challenges in false-discovery-rate control procedures, gene-set enrichment analysis, or other procedures aiming to summarize the collection of test results. To tackle these challenges, researchers sometimes will--explicitly or implicitly--use the correlations (e.g., as measured by the Pearson correlation coefficients) among the data rows to approximate the correlations among the corresponding test statistic values. We show that, however, such approximations are only valid under limited settings. We investigate the relationship between the correlation coefficient between a pair of test statistics (test-statistic correlation) and the correlation coefficient between the two corresponding data rows (data-row correlation). We derive an analytical formula for the test-statistic correlation as a function of the data-row correlation for a general class of test statistics: in particular, two-sample t-test is a special case. The analytical formula implies that the test-statistic correlation is generally weaker than the corresponding data-row correlation, and in general, the latter will not well approximate the former when the involved null hypotheses are false. We verify our analytical results through simulations.

Source connections

Explore related subjects

geneticsPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Di, Y., Jiang, D., Zhuo, B.. 2019-09-08. Test-statistic correlation and data-row correlation. https://doi.org/10.1101/759027

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

SARDINE unlocks EWAS summary statistics for large-scale epigenetic biomarker analysis

Epigenetic clocks are increasingly used to study biological aging, yet their interpretation and utility remain uncertain. Meanwhile, the growing wealth of epigenome-wide association study (EWAS) summary statistics remains largely untapped for evaluating these biomarkers. Here, we introduce SARDINE, a statistical framework that estimates epigenetic clock-phenotype associations using EWAS summary statistics, clock weights, and reference methylation covariance, without individual-level data. SARDINE accurately reproduced individual-level association estimates across 38 epigenetic biomarkers and four phenotypes in 15,677 individuals. Applying SARDINE to diverse EWAS datasets revealed substantial heterogeneity in biomarker associations with aging, diseases, and lifestyle factors. By transforming CpG-level EWAS results into interpretable biomarker-level associations, SARDINE unlocks a largely underutilized data resource and establishes a scalable paradigm for systematic biomarker evaluation and discovery in aging research.

genetics↗
bioRxiv2026

Deep transcriptomic analysis of molecular mediators of complex traits in heterogeneous stock (HS) rats

Rats are a powerful model organism for genome-wide association studies (GWAS) of physiological and behavioral traits. Heterogeneous Stock (HS) rats, which were created by intercrossing 8 inbred rat strains, are heavily used for this purpose because of their genetic diversity and accumulated recombinations due to over 100 generations of outbreeding, which improves mapping precision. However, functional genomics resources for post-GWAS interpretation of genomic findings have been limited. We processed 3,160 RNA-seq samples, 1,932 of which were previously unpublished, from 13 tissues as part of RatGTEx. We used Pantry to quantify six modalities of transcriptome regulation (expression, isoform ratio, intron excision ratio, alternative transcription start site (TSS), alternative poly(A), and RNA stability) in each sample, which allowed us to identify 144,876 independent quantitative trait loci (QTL) for 22,599 genes. We performed a transcriptome-wide association study (TWAS) with 19,796 unique genes across tissues and transcriptional modalities and 308 traits collected from 14 studies using FUSION, which identified 7,473 significant gene-trait associations. This work presents a significant expansion of the RatGTEx resource in terms of samples, modes of transcriptome regulation, and genetic analyses. Our findings provide molecular evidence for the involvement of specific genes in mediating GWAS signals through a range of transcriptome regulatory mechanisms. These gene-to-phenotype relationships can be easily applied to genetic analyses in humans and other species. Further experiments can use this evidence to confirm these molecular mediators and target them for the treatment of behavioral and physiological traits and disorders. To facilitate use of these results, we have updated results on RatGTEx.org and provided a new interactive TWAS web portal (twas.ratgtex.org).

genetics↗
bioRxiv2026

Genomic and Epigenomic Relationships Capture Different Patterns of Genotype-by-Environment Interaction in Grapevine Clones

Genotype-by-environment interaction (GEI) challenges selection in clonally propagated perennial crops by reducing the consistency of clone performance across years. In perennial crops, both genetic and epigenetic variation are likely to contribute to complex trait variation among clones. However, the extent to which each molecular source contribute to GEI remains poorly understood. Here, we used UMR-seq, an assay that recovers SNPs and unmethylated-region (UMR; DNA methylation) profiles from a single library, to generate paired genomic and epigenomic profiles for 206 Riesling (Vitis vinifera L.) clones phenotyped for yield and juice quality traits across nine growing seasons. We then partitioned the contribution of each omic layer to clonal variation and GEI, and examined the genomic distribution of their effects. In models without explicit GEI, environmental and residual effects were the largest source of phenotypic variance and the clonal main effect was modest. Genomic relationships accounted for most of the clonal variance in four of five traits, and adding an epigenomic relationship matrix (ERM) to the genomic relationship matrix (GRM) had little effect. In contrast, when GEI effects were modelled using factor-analytic covariance structures, the epigenomic layer captured a substantial proportion of GEI variance and exceeded the genomic contribution in four of five traits. The two omic layers differed not only in the amount of GEI captured, but also in the pattern of clone re-ranking across years. When back-solved effects were aggregated into genomic windows, the largest genomic and epigenomic effects tended to fall in different regions, with overlap no greater than expected by chance. Together, these results show that genomic and epigenomic relationships capture complementary components of year-to-year variation in clone performance, with the epigenomic contribution expressed primarily through GEI. These findings provide a basis for investigating the molecular drivers of GEI and for testing whether genomic and epigenomic information can inform selection for stable clonal performance across variable environments.

genetics↗