NeuroimaGene

Background

This is the resource page for the NeuroimaGene resource described in our paper "A Transcriptomic Atlas of the Human Brain Reveals Genetically Determined Aspects of Neuropsychiatric Health." This resource serves as a publicly accessible atlas detailing the associations between endogenous gene expression and brain anatomy and physiology. As detailed in the paper, we conduct Joint Tissue Imputation (JTI) informed Transcriptome Wide Association Studies (TWAS) via summary statistics based S-PrediXcan for the >3,400 genome wide association studies (GWAS) conducted by the UK BioBank (UKB) for MRI-derived measures of the brain. We identify genetically regulated gene expression (GReX) associated with neurologic measures observed on MRI imaging. The patients comprising the UKB neuroimaging study are 40-69 and were screened for overt neurologic pathology. They generally represent an adult population without neurologic disease.

As such, NeuroimaGENE catalogues the neurologic consequences of lifelong exposure to increases or decreases in gene expression.

Before continuing, a little terminology.

GWAS  - Genome Wide Association Study
TWAS  - Transcriptome Wide Association Study
GReX  - Genetically regulated gene expression
JTI   - Joint Tissue Imputation (link)
NIDP  - Neuroimaging Derived Phenotype
MRI   - Magnetic resonance imaging
T1    - MRI modality classically used for structural characterization of the brain
dMRI  - diffusion weighted MRI (used in our data for white matter tractography)
fMRi  - functional MRI used for examining coordinated activity across regions of the brain 
UKB   - United Kingdom Biobank
eQTL  - expression Quantitative Trait Locus
GTEx  - Genotype Tissue Expression Consortium

R package in CRAN

neuroimaGene: Transcriptomic Atlas of Neuroimaging Derived Phenotypes. Download.

Associated papers for further reading

Barbeira, Alvaro N., et al. "Exploring the phenotypic consequences of tissue specific gene expression variation inferred from GWAS summary statistics." Nature communications 9.1 (2018): 1-20.

Zhou, Dan, et al. "A unified framework for joint-tissue transcriptome-wide association and Mendelian randomization analysis." Nature genetics 52.11 (2020): 1239-1246.

Miller, Karla L., et al. "Multimodal population brain imaging in the UK Biobank prospective epidemiological study." Nature neuroscience 19.11 (2016): 1523-1536.

Elliott, Lloyd T., et al. "Genome-wide association studies of brain imaging phenotypes in UK Biobank." Nature 562.7726 (2018): 210-216.

Gamazon, Eric R., et al. "Multi-tissue transcriptome analyses identify genetic mechanisms underlying neuropsychiatric traits." Nature genetics 51.6 (2019): 933-940.

Using the resource

Accessing the resource

The full neuroimaGene resource consists of commandline interface located here on Github as well as a SQL database containing the associations which are the heart of the project. Prior to using the tools here, download the NeuroimaGene resource from Zenodo and add the neuroimaGene.db file to the neuroimaGene directory.

Basic command line usage

Having downloaded the NeuroimaGene SQL database, the commandline tool, hosted here on github, is designed to enable the user end functionality. The scripts in this repository allow the user to query the NeuroimaGENE for NIDPs associated with GReX for one or more genes of interest through the commands detailed below. The output from the commands will return a txt file with the following data:

gene: the Ensmble Gene ID
gene_name: the HUGO gene name 
gwas_phenotype: the Neuroimaging Derived Phenotype as detailed by the UKB neuroimaging GWAS 
zscore: The normalized effect size of GReX on the NIDP (most appropriate for comparison)
effect_size: the raw value of the predicted effect size
training_model: the JTI-enriched tissue specific model in which the association is found
atl_bhpval: the atlas-specific benjamini-hochberg false-discovery rate corrected association p-value
		or
atl_bfpval: the atlas-specific bonferroni corrected p-value
		or
 	mod_bhpval: the modality-specific benjamini-hochberg false-discovery rate corrected association p-value
		or
mod_bfpval: the modality-specific bonferroni corrected p-value
		or
 	pvalue: the nominal, uncorrected association p-value

NIDP details can be found here and here

Analytic Tools and built-in NIDP prioritization

Included in the NeuroimaGene repository is a commandline tool for analysis of multiple genes (get_nidps.sh). This program takes as input a file of genes (HUGO gene names or ensembl id's). The script generates .txt files containing the NIDPs implicated by the provided list of genes as well as graphs providing visual representations of the association data.

This script and the data it generates are designed to identify instances in which dysexpression of multiple genes of interest converge upon related neurologic aspects. For example, one might expect multiple genes associated with distractiability to converge upon the executive network of the brain. Running get_nidps.sh on a set of genes associated with distractability will display the set of NIDPs predicted to be most different from baseline in the presence of altered expression of the input genes. The text file will indicate the number and identity of trait-associated genes associated with each NIDP. It will also detail the number and identity of the training models in which these associations were found to be significant (according to the provided multiple testing threshold).

NIDP selection for gene set analysis

In the process of using the tool, the user is responsible for selecting a subset of NIDPs from the resource for analysis. These NIDPs represent different types of brain measures such as hippocampal subfields, area and thickness of named cortical regions, fractional anisotropy of named white matter tracts etc. It is recommended to identify the type of brain measure one is interested in prior to performing the gene set analysis. The options provided in the commandline tool are detailed in the dropdown table below.

NIDP atlas descriptions and source links

MRI modality	atlas name	Number of NIDPs	Description	source
T1	all	1319	All measures recorded by the UKB neuroimaging study derived from T1 imaging	see note*
T1	Destrieux	444	Destrieux atlas parcellation of cortical sulci and gyri	Destrieux
T1	AmygNuclei	20	morphology of Nuclei of the amygdala	Amygdala nuclei
T1	Subcortex	52	subcortical volumetric segmentation	Subcortex
T1	Broadmann	84	cortical morphology via Broadmann Areas	Broadmann
T1	Desikan	202	Desikan Killiany atlas parcellation of cortical morphology	Desikan
T1	DKT	186	DKT atlas parcellation of cortical morphology	DKT
T1	FAST	139	cortical morphology via FMRIB's Automatic Segmentation Tool	FAST
T1	FIRST	15	Subcortical morphologogy via FIRST	FIRST
T1	HippSubfield	44	morphology of Hippocampal subfields	HippSubfield
T1	pial	66	structure: Desikan Killiany atlas of the pial surface	Desikan
T1	Brainstem	5	structure: Freesurfer brainstem parcellation	Brainstem
T1	SIENAX	10	structure: Structural Image Evaluation of whole brain measures	SIENAX
T1	ThalamNuclei	52	morphology of the Nuclei of the thalamus	ThalamNuclei
dMRI	all	675	All measures recorded by the UKB neuroimaging study derived from DWI imaging	see note*
dMRI	ProbtrackX	243	white matter mapping obtained via probabilistic tractography	ProbtrackX*
dMRI	TBSS	432	white matter mapping obtained via tract-based spatial statistics	TBSS*
rfMRI	ICA100	1485	functional connectivity using 100 cortical seeds	see note*
rfMRI	ICA25	210	functional connectivity using 25 cortical seeds	see note*
rfMRI	ICA-features	6	summary of functional connectivity components	see note*
T2_FLAIR	BIANCA	1	white matter hyperintensity classification algorithm	BIANCA
T2star	SWI	14	susceptibility-weighted imaging: microhemorrhage and hemosiderin deposits	see note*

* see original publication for details here (Alfaro-Almagro, Fidel, et al. "Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank." Neuroimage 166 (2018): 400-424.)

Selecting an appropriate Multiple Testing Correction for statistical significance.

In addition to selecting the set of NIDPs, get_nidps.sh requires a multiple testing threshold correction. Each imaging modality contains a different number of NIDPs (see table above). The Bonferroni correction treats each of these NIDPs as independent even though we know through significant data analyses that this is not true. This is a highly conservative threshold that will yield high confidence associations but is likely to generate many false negatives. Recognizing the interrelatedness of brain measures from the same modality and atlas, we recommend using the less stringent Benjamini Hochberg False discovery rate for discovery analyses.

The full set of associations in sql database form is a little cumbersome (~1 GB) and can be found on Zenodo."

Using the commandline tools

We include two commandline tools in the provided resource. The first is a simple script named get.sh. This can be used to preview neuroimaging measures that are associated with a single gene at a user-provided significance level. It is not intended as a primary research tool and does not return any interactable files. Instead, it is intended as a means to check the resource for a gene of interest and to familiarize oneself with the resource prior to more in-depth analyses. To run the script, navigate to the resource directory in the commandline and run the following command:

bash get.sh

The script will prompt the user for four pieces of information.

(1) the name of the gene (ensembl ID or HUGO gene name); 
(2) the subset of Neuroimaging features to be queried, 
(3) the multiple testing correction by which p-values should be adjusted for significance filtering 
	(Only associations with adjusted p-values less than 0.05 will be returned.) 
(4) The number of results the user wishes to preview in raw tabular form. (3-10 recommended)

Upon receipt of these data, the program will select the subset of significant GReX-neuroimaging associations that fit the user's criteria for the gene in question. It will print a preview of these associations to the terminal in accordance with the number of requested lines from prompt 4. below the data preview, the script provides a number of descriptive statistics about the gene in question and it's associations with the queried neuroimaging features.

The second script is designed for a richer analytic approach. Using similar commandline prompts, it takes as input a list of genes and identifies the neuroimaging measures most strongly associated with that set of genes. It returns a text file with the identified associations as well as a visual representation of the NIDPs most heavily associated with the gene set, annotated by brain regions.

To use get_nidps.sh, run the following command in terminal:

bash /PATH/get_nidps.sh

The script will provide five prompts in sequence.

"Enter file containing Genes or Ensmbl IDs_ " paste or write path and filename of genes for query
"Enter output directory_ " paste or write path to output directory
"Enter analysis tag_ " select short descriptive tag for analysis
"Choose imaging modality:" select imaging modality and atlas from dropdown menu
"Choose multiple testing threshold:" select preferred multiple testing correction from dropdown menu

The script will feed the provided data into an R analysis pipeline and deposit the resulting data into a file titled "[your tag here]NIDPs" in the provided output directory.

Alternatively, you may run the R-script directly providing the following parameters

INPUT_GENES.txt: A .txt file containing a single column of HUGO gene names or ensembl ids with no header
RESOURCE: The path and name of the NeuroimaGenefast.db file (Downloaded from Zenodo)
OUTPUT_DIR: a directory path to which the output from the analysis should be deposited
NAME: a short descriptive name to mark the analysis (eg. parknsn_genes if studying Parkinson's)
MODALITY: the modality of the queried neuroimaging set (T1, dMRI, rfMRI etc.) [required if type not 'all']
ATLAS: the atlas of the queried neuroimaging set [required if type = 'atl']
TYPE: the type of data subset ('atl' for an atlas-defined subset, 'mod' for modality-defined subset, or 'all')
PVALUE: P-value Multiple Testing Correction ("BH" for Benjamini Hochberg FDR, "BF" for Bonferroni, "nom" for nominal)
PATH: path to the downloaded NeuroimaGENE resource directory.

Run the script with the following commands customized for your genes of interest and directories.

Rscript PATH/get_NIDPs_vis.r \
	-f INPUT_GENES.txt \
	-r PATH/NeuroimaGenefast.db
	-o OUTPUT_DIR \
	-n NAME \
	-m MODALITY \
	-a ATLAS \
	-t TYPE \
	-p PVALUE \
	-s /PATH/BIG40-IDPs_v4_discovery2_anno.tsv \
	-b /PATH/fs_anno.txt

An additional flag is the -g flag for genes. include -g y in the Rscript command, you will receive a text file and png figure for each individual gene detailing the top associated NIDPs for that gene in addition to typical the full analysis for the aggregate set of NIDPs.

TUTORIAL

Within the NeuroimaGene directory is a tutorial directory for practice running the script. The data are derived from the following paper Mishra et al, Nature 2022 in which the authors use TWAS to identify several genes whose GReX is associated with stroke. Here we assess for structural MRI phenotypes from the Desikan Atlas associated with dysexpression of the prioritized genes. You can run the tutorial via the following commands requiring only the PATH of the downloaded directory.

bash /PATH/get_nidps.sh
	=> "Enter file containing Genes or Ensmbl IDs_ " PATH/tutorial/tutorial_gns.tx
	=> "Enter output directory_ " PATH/tutorial/
	=> "Enter analysis tag_ " stroke_gns
	=> "Choose imaging modality:" (1)
	=> "Choose multiple testing threshold:" (1)

Results should be generated and deposited in the following directory: PATH/tutorial/stroke_gns_NIDPs

Alternatively, you may run the program directly from the Rscript as shown below.

Rscript PATH/get_NIDPs_vis.r \
-f /PATH/tutorial/tutorial_gns.txt
-r /PATH/NeuroimaGenefast.db
-o /PATH/tutorial/
-n stroke_gns
-m 'T1' \
-a 'Desikan' \
-t 'atl' \
-p 'BH' \
-s /PATH/BIG40-IDPs_v4_discovery2_anno.tsv \
-b /PATH/fs_anno.txt

Amongst the other data generated, this should generate the following figure detailing NIDPs on the y axis and the mean normalized effect size magnitude on the x axis with color and shape detailing brain region descriptors and direction of effect respectively. There are 3 additional plots showing the mean normalized effect size of the genes on the brain regions characterized by the Desikan atlas divided into volume, surface area, and thickness. As stated above, detailed information concerning the naming of the NIDPs is available the the UKB online neuroimaging portal.