A practical guide to linking brain-wide gene expression and neuroimaging data

Abstract

The recent availability of comprehensive, brain-wide gene expression atlases such as the Allen Human Brain Atlas (AHBA) has opened new opportunities for understanding how spatial variations on the molecular scale relate to the macroscopic neuroimaging phenotypes. A rapidly growing body of literature is demonstrating relationships between gene expression and diverse properties of brain structure and function, but approaches for combining expression atlas data with neuroimaging are highly inconsistent, with substantial variations in how the expression data are processed. The degree to which these methodological variations affect findings is unclear. Here, we outline a seven-step analysis pipeline for relating brain-wide transcriptomic and neuroimaging data and compare how different processing choices influence the resulting data. We suggest that studies using AHBA should work towards a unified data processing pipeline to ensure consistent and reproducible results in this burgeoning field.

Generative modeling of brain maps with spatial autocorrelation

“…Next, SA is reintroduced by smoothing the permuted map with a distance-dependent kernel. Motivated by previous work Romero-Garcia et al, 2018;Markov et al, 2011;ArnatkevicIūtė et al, 2019), here we use a smoothing kernel with weights which fall off exponentially 295 with distance. However, we found that null distributions are largely insensitive to the functional form of the kernel ( Supplementary Fig.…”

“…In network neuroscience, distancedependent wiring rules have been incorporated into generative null network models to establish expectations for graph-theoretic measures (Song et al, 2014;Betzel et al, 2016). Here we have incorporated SA into our null hypothesis in an adaptive (i.e., target brain-map specific) manner due to its prominence and ubiquity in brain maps Markov et al, 2011;ArnatkevicIūtė et al, 2019;Romero-Garcia et al, 2018), its variation across brain maps Gryglewski et al, 2018), and because of its profound impact on 630 statistical measures of interest. 5).…”

Relating quantitative 7T MRI across cortical depths to cytoarchitectonics, gene expression and connectomics: a framework for tracking neurodegenerative disease

“…Values were then averaged across participants and only the left hemisphere was used in order to match the data from the AHBA, since it contains data from 6 left hemispheres and only 2 right hemispheres. Thus, this analysis was in keeping with previous analyses using only the left hemisphere (Arnatkevic lute et al, 2019). This resulted in matrices of 8×43 for von Economo and 8×180 for HCP-MMP 1.0 atlases.…”

“…For this analysis lists of genes specific to cortical layers were obtained (Bernard et al, 2012) and brain region expression data was taken from the AHBA for each ROI in the left hemisphere of the HCP-MMP 1.0 atlas, as the complete data set is not available for the right hemisphere (Glasser et al, 2016; Arnatkevic lute et al, 2019). R2* values were averaged for each ROI.…”

“…Big Brain cortical layers were defined using a machine learning approach, as previously described (Wagstyl et al, 2019). Gene expression data for the AHBA was extracted for 180 left-hemisphere regions of the HCP-MMP 1.0 atlas as detailed by (Arnatkevic lute et al, 2019). Data was available from 6 neurotypical human brains (6 left-hemisphere and 2 right hemisphere).…”