LeonDLotter
commented
2 years ago Hi Qiannan, Thanks for your question! And please excuse the late answer.
First, what you want to do is use data generated from a surface-based pipeline. I designed ABAnnotate for use with volumetric data and as such, all functions including the one that generates null maps from the input data expects a volumetric parcellation file (*.nii). The DK parcellation is surface-based and thus cannot used with ABAnnotate. However, there is a way to make it work: The "Mindboggle" parcellation is available in MNI space and is based on the DK atlas. It contains, however, only surface parcels. And you would have to check if the parcel indices in the MNI file correspond to your freesurfer-extracted ROI data (and, if not, relabel it).
Info on the leadDBS page: https://www.lead-dbs.org/helpsupport/knowledge-base/atlasesresources/cortical-atlas-parcellations-mni-space/ Mindboggle page: https://mindboggle.info/data.html OSF for the data: https://osf.io/kgdey/ You would want to use the file "OASIS-TRT-20_jointfusion_DKT31_CMA_labels_in_MNI152_2mm_v2.nii"
To use it with ABAnnotate you want to:
- Import your gene expression data. You already extracted the data with abagen, so you might want to take a look of the structure of the ABAnnotate atlas datasets. These are stored at
ABAnnotate/atlas/atlas_name. You would have to prepare a parcellation file (the mindboggle parcellation) and the expression matrix (e.g., stored as "desikan_expression.mat"). Take a look at the contents and variable names of the "*_expression.mat" files of the other atlases inABAnnotate/atlas/atlas_name(after you used one atlas once so that ABAnnotate actually downloads it) to create a file from your desikan abagen data. - You could then call ABAnnotate as follows:
opt.analysis_name = 'EXAMPLE_NAME';
opt.n_nulls = 1000;
opt.phenotype_data = ROI_DATA; % a vector with size (number of rois, x 1),
opt.dir_result = 'EXAMPLE_DIR';
opt.atlas = 'path/to/desikan.nii' % the mingboggle/desikan parcellation volume (likely modified to correspond to your data)
opt.aba_mat = 'path/to/desikan_expression.mat' % the abagen dataset *mat file
opt.GCEA.dataset = 'PsychEncode-cellTypesTPM-discrete'; % or another dataset
cTable = ABAnnotate(opt);That being said, there is another new toolbox very similar to ABAnnotate, that works with surface data and, I think, actually uses the DK atlas by default: https://doi.org/10.1016/j.celrep.2021.110173 https://github.com/alegiac95/Imaging-transcriptomics You might want to try this before you get into the possibly painful process I described above ;)
Feel free to ask any follow up question if you try to go with ABAnnotate!
Best Leon
sidchop
Hi Leon, Thank you for developing the ABAnnotate, which really helps me conduct GSEA analyses.
I got a problem when using ABAnnotate in my data. I performed the recon-all pipeline of FreeSurfer on patients and controls and extracted measures such as volumes from 68 neocortical regions and 14 subcortical regions parcellated with the D-K atlas. These regional brain volumes were further used in some machine learning and statistical testing procedures in R and Python, and I got a case-control statistic for each ROI. I also got the expression values of genes for ROIs of the D-K atlas using the abagen toolbox.
My question is, is it possible to conduct correlation analyses between this region*statistic value vector or matrix with expression values of genes, and then perform GSEA analyses for these genes using ABAnnotate? I found that ABAnnotate receives Nifti files rather than numeric matrices. I was trying to transform this numeric matrix into a Nifti file with a statistic value in each ROI, but I found the transformation achieved low quality and further procedures in ABAnnotate. Are there any recommended ways to transform such numeric matrices into imaging files that could be further used in ABAnnotate?
Any help will be appreciated. Qiannan