Statistical testing in transcriptomic-neuroimaging studies: A how-to and evaluation of methods assessing spatial and gene specificity

Multiscale integration of gene transcriptomic and neuroimaging data is becoming a widely used approach for exploring the molecular underpinnings of large-scale brain organization in health and disease. Proper statistical evaluation of determined associations between imaging-based phenotypic and transcriptomic data is key in these explorations, in particular to establish whether observed associations exceed “chance level” of random, nonspecific effects. Recent approaches have shown the importance of statistical models that can correct for spatial autocorrelation effects in the data to avoid inflation of reported statistics. Here, we discuss the need for examination of a second category of statistical models in transcriptomic-neuroimaging analyses, namely those that can provide “gene specificity.” By means of a couple of simple examples of commonly performed transcriptomic-neuroimaging analyses, we illustrate some of the potentials and challenges of transcriptomic-imaging analyses, showing that providing gene specificity on observed transcriptomic-neuroimaging effects is of high importance to avoid reports of nonspecific effects. Through means of simulations we show that the rate of reported nonspecific effects (i.e., effects that cannot be specifically linked to a specific gene or gene-set) can run as high as 60%, with only less than 5% of transcriptomic-neuroimaging associations observed through ordinary linear regression analyses showing both spatial and gene specificity. We provide a discussion, a tutorial, and an easy-to-use toolbox for the different options of null models in transcriptomic-neuroimaging analyses.