Spatial accuracy of Allen Human Brain Atlas tissue samples and its effects on neuroimaging studies
Presented During:
Wednesday, June 25, 2025: 5:45 PM - 7:00 PM
Brisbane Convention & Exhibition Centre
Room:
M2 (Mezzanine Level)
Poster No:
704
Submission Type:
Abstract Submission
Authors:
Yohan Yee1, Yuhan Liu2, Leon French3, Yashar Zeighami1, Gabriel Devenyi2, Mallar Chakravarty1
Institutions:
1McGill University, Montréal, Canada, 2Douglas Mental Health University Institute, Montréal, Canada, 3University of Toronto, Toronto, Canada
First Author:
Co-Author(s):
Introduction:
The Allen Human Brain Atlas (AHBA) consists of gene expression profiled within 3702 tissue samples from six donor brains. The provision of sample coordinates within MNI-space has made comparing neuroimaging findings to normative gene expression feasible and popular; the AHBA has been featured in at least 202 studies over the last five years (via PubMed) and ~44 OHBM 2024 abstracts. In fact, there exist multiple sets of coordinates describing these sample locations: 1) "original" coordinates from the AHBA (Hawrylycz, 2012), 2) updated coordinates from the "alleninf" package (Gorgolewski, 2014) and used in the abagen software (Markello, 2021), and 3) "CIC" coordinates derived at the Cerebral Imaging Centre via multispectral image registrations to a newer MNI template (Devenyi, 2018). Surprisingly, these coordinates place many tissue samples in dramatically different anatomical locations (Figure 1a,b). Here, we test the accuracy of these three coordinate sets through multiple types of tests of location accuracy, and then show that inaccuracies in coordinates can result in improper inferences in neuroimaging studies.
Methods:
We defined several tests of coordinate accuracy. For each set of coordinates, we:
1) correlated donor MR image intensity values at true sample locations with MNI template intensity values (Figure 1c), separately for each donor brain. Higher accuracy of coordinates would be reflected as higher intensity correlations (a reflection of donor-template alignment quality).
2) used expert annotations of anatomical structure from which samples were obtained along with an MNI-space anatomical atlas (both from the Allen Institute), and examined the concordance between true dissection structure and the structure in the coordinate-space.
3) examined the performance of coordinates on an interpolation task. We interpolated the expression of parvalbumin (PVALB) and calbindin (CALB1) genes within the thalamus, and compared this to known expression patterns within thalamic nuclei (Jones, 1998).
After quantifying accuracy, we assessed the effects of using inaccurate coordinates on imaging-transcriptomics studies. First, we correlated gene expression to a brain map representing metabolic differences in Alzheimer's disease (Patel, 2020), only modifying the AHBA coordinate set. We obtained ranked gene lists, and examined the extent that the ranks of genes differed between coordinate sets. To generalize this, we applied a similar analysis to 66 NeuroQuery brain maps and examined coordinate-related differences in gene ranks and (using the U-test for enrichment) in downstream inferences on biological pathways and processes.
1) correlated donor MR image intensity values at true sample locations with MNI template intensity values (Figure 1c), separately for each donor brain. Higher accuracy of coordinates would be reflected as higher intensity correlations (a reflection of donor-template alignment quality).
2) used expert annotations of anatomical structure from which samples were obtained along with an MNI-space anatomical atlas (both from the Allen Institute), and examined the concordance between true dissection structure and the structure in the coordinate-space.
3) examined the performance of coordinates on an interpolation task. We interpolated the expression of parvalbumin (PVALB) and calbindin (CALB1) genes within the thalamus, and compared this to known expression patterns within thalamic nuclei (Jones, 1998).
After quantifying accuracy, we assessed the effects of using inaccurate coordinates on imaging-transcriptomics studies. First, we correlated gene expression to a brain map representing metabolic differences in Alzheimer's disease (Patel, 2020), only modifying the AHBA coordinate set. We obtained ranked gene lists, and examined the extent that the ranks of genes differed between coordinate sets. To generalize this, we applied a similar analysis to 66 NeuroQuery brain maps and examined coordinate-related differences in gene ranks and (using the U-test for enrichment) in downstream inferences on biological pathways and processes.
Results:
Intensity correlations were highest for CIC coordinates, for both T1- and T2-weighted image intensities (Figure 1d). Concordance between true and atlas-identified structures was highest for CIC coordinates (Figure 1e). Spatial interpolation of PVALB and CALB1 resulted in the most expected expression pattern when using CIC coordinates, both visually and as quantified by an ROC analysis (Figure 1f). Thus, CIC coordinates are most optimal.
Switching coordinates in the Patel (2020) study results in large rank differences, with at least 1 in 10 genes having a rank difference of more than 4100 places when compared to original or alleninf coordinates (Figure 2a). Similar rank differences are seen with NeuroQuery maps, which when subject to enrichment analyses, results in up to 21% of gene ontology and 43% of pathway terms being falsely identified or rejected when considering CIC coordinates as truth (Figure 2b).
Switching coordinates in the Patel (2020) study results in large rank differences, with at least 1 in 10 genes having a rank difference of more than 4100 places when compared to original or alleninf coordinates (Figure 2a). Similar rank differences are seen with NeuroQuery maps, which when subject to enrichment analyses, results in up to 21% of gene ontology and 43% of pathway terms being falsely identified or rejected when considering CIC coordinates as truth (Figure 2b).
·Figure 1: anatomical accuracy of original, alleninf, and CIC coordinates for AHBA tissue samples. High resolution version available at: https://shorturl.at/0RotX
·Figure 2: measured and predicted bias resulting from the previous less accurate original and alleninf coordinates. High resolution version available at: https://shorturl.at/OhBlL
Conclusions:
CIC coordinates best reflect the true anatomical locations of AHBA samples. Using less accurate original or alleninf coordinates to compare gene expression to neuroimaging findings results in biased gene lists and improper identification/rejection of biological pathways/processes. Past inferences made with previous coordinates may be compromised, and future studies could be improved using CIC coordinates.
Genetics:
Transcriptomics 1
Modeling and Analysis Methods:
Image Registration and Computational Anatomy
Neuroinformatics and Data Sharing:
Brain Atlases 2
Informatics Other
Keywords:
Atlasing
Computational Neuroscience
Data analysis
Informatics
STRUCTURAL MRI
Structures
Other - Allen Human Brain Atlas; transcriptomics; gene expression; coordinates; bias; NeuroQuery; localization
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
Which processing packages did you use for your study?
Provide references using APA citation style.
Gorgolewski, K. J., Fox, A. S., Chang, L., Schäfer, A., Arélin, K., Burmann, I., ... & Margulies, D. S. (2014). Tight fitting genes: finding relations between statistical maps and gene expression patterns. F1000Posters, 5(1607), 10-7490.
Hawrylycz, M. J., Lein, E. S., Guillozet-Bongaarts, A. L., Shen, E. H., Ng, L., Miller, J. A., ... & Jones, A. R. (2012). An anatomically comprehensive atlas of the adult human brain transcriptome. Nature, 489(7416), 391-399.
Jones, E. G. (1998). The core and matrix of thalamic organization. Neuroscience, 85(2), 331-345.
Markello, R. D., Arnatkeviciute, A., Poline, J. B., Fulcher, B. D., Fornito, A., & Misic, B. (2021). Standardizing workflows in imaging transcriptomics with the abagen toolbox. elife, 10, e72129.
Patel, S., Howard, D., Man, A., Schwartz, D., Jee, J., Felsky, D., ... & French, L. (2020). Donor-Specific transcriptomic analysis of Alzheimer's Disease-Associated hypometabolism highlights a unique donor, ribosomal proteins and microglia. eneuro, 7(6).