Presented During:
Monday, June 24, 2024: 5:45 PM - 7:00 PM
COEX
Room:
Grand Ballroom 101-102
Poster No:
2274
Submission Type:
Abstract Submission
Authors:
Ankush Shetty1, Ravi Bhatt2, Alyssa Zhu3, Shruti Gadewar3, Iyad Ba Gari1, Paul Thompson3, Neda Jahanshad4
Institutions:
1University of Southern California, Marina Del Rey, CA, 2USC, Los Angeles, CA, 3USC, Marina Del Rey, CA, 4Imaging Genetics Center, Keck School of Medicine of University of Southern California, Los Angeles, California
First Author:
Ankush Shetty
University of Southern California
Marina Del Rey, CA
Co-Author(s):
Iyad Ba Gari
University of Southern California
Marina Del Rey, CA
Neda Jahanshad, PhD
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, California
Introduction:
Identifying the degree to which the genetic architecture of brain structure and function derived from brain imaging features overlap with other brain-based traits is important for understanding overlapping neurobiological mechanisms. Pre-assembled pipelines for post-GWAS analyses, such as CTG-VL[1] can help with efficient processing, but have not been developed to specifically test for genetic relationships and enrichment against brain-related traits. Toolboxes from abagen[2] and ENIGMA[3] integrate brain gene-expression across different atlases but do not take GWAS summary statistics as input. Here we present GiNi (Genetic interrogator for Neuroimaging) a Python-based command line tool, to streamline a comprehensive collection of statistical genetic relationships between multiple brain-based traits. We include heritability estimates, global and local genetic correlation, and causal genetic associations and enrichment across various brain regions, tissues and cell types.
Methods:
We created a post-GWAS processing and analysis pipeline based on four existing packages (METAL, LDSC, LAVA, GSMR), and external brain gene-expression and -splicing data resources, combining the Python, R, and command line tools into a single Python package. Each analysis first prepares the required inputs, accounting for different file types and execution methods. All post-GWAS analyses can be run independently or together, and are focused on brain-related traits. The pipeline can either be run with the default parameters, or fine-tuned by providing preferred parameters specific to each software. We have incorporated parallel executions through either multi-threading or the use of a high-performance cluster for computational efficiency, as neuroimaging GWAS often include multiple traits [4, 5]. A meta-analysis can be performed to prepare combined summary statistics from multiple individual GWAS inputs (ie, same trait, different datasets) using METAL [6]. We have written the post-GWAS analyses to be specific to brain-related phenotypes (Fig 1). In addition to the analysis outputs, the pipeline will also provide a log for provenance and version tracking of the software involved. We tested our framework on the midsagittal corpus callosum (midCC), where we performed meta-analysis, followed by post-GWAS analyses on area and mean thickness of all phenotypes in the UK Biobank and ABCD study. For each analysis, we benchmarked the run time and required memory.
![Supporting Image: OHBM_Fig_1_Sub_3.jpg](https://files.aievolution.com/prd/hbm2401/abstracts/abs_3920/OHBM_Fig_1_Sub_3.jpg)
Results:
We combined all our wrappers into a Python-based command line interface called GiNI, (Fig 1a). For handling multiple GWAS inputs, the type of meta-analysis approach can be specified using the "--meta" flag, and the user can specify the reference panel with the "--ethnicity" flag. Output plots can also be customized using the flag "--plot_feature".
We applied GiNi to a GWAS meta-analysis of 12 neuroimaging features across UKB and ABCD (Fig 2). The run time for global genetic correlation, Mendelian randomization, heritability estimation and local genetic correlation which were 15 seconds, 30 minutes, 3 minutes-12 hours, and 12 hours-24 hours, respectively. The maximum memory requirement needed for the pipeline to perform is up to 128GB, which is for TWAS across tissue types. Depending on the user's study preferences and available resources, the pipeline can also be run sequentially using the "--computing_options" flag or only in specified parts using the "--analysis_list" flag.
![Supporting Image: OHBM_Fig_2_Sub_2.jpg](https://files.aievolution.com/prd/hbm2401/abstracts/abs_3920/OHBM_Fig_2_Sub_2.jpg)
Conclusions:
We have created GiNi, a user-friendly pipeline that streamlines post-GWAS analyses for pinpointing plausible genetically informed, neurobiological mechanisms. Users will be able to derive heritability estimates, partitioned by multiple tissue and cell types, as well as determine the genetic overlap and causal associations on a global and local genetic level with cerebral cortex and subcortical structures. The pipeline is easily adaptable and will include more brain phenotypes, analysis types, and state-of-the-art tools developed in the future.
Genetics:
Genetic Modeling and Analysis Methods
Transcriptomics
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Cortical Anatomy and Brain Mapping
Neuroinformatics and Data Sharing:
Workflows 1
Informatics Other 2
Keywords:
Computational Neuroscience
Computing
Data analysis
Data Organization
Design and Analysis
Development
Informatics
Meta- Analysis
Phenotype-Genotype
Workflows
1|2Indicates the priority used for review
Provide references using author date format
[1] Cuellar-Partida, G et all. (2019), Complex-Traits Genetics Virtual Lab: A community-driven web platform for post-GWAS analyses: Supplementary Figures. 10.1101/518027.
[2] Markello, RD,et al. (2021), Standardizing workflows in imaging transcriptomics with the abagen toolbox. Biorxiv. doi:10.1101/2021.07.08.451635
[3]Larivière, S.,et al.(2021), The ENIGMA Toolbox: multiscale neural contextualization of multisite neuroimaging datasets. Nat Methods 18, 698–700 . https://doi.org/10.1038/s41592-021-01186-4
[4] Grasby, K. L. et al. (2020), The genetic architecture of the human cerebral cortex. Science 367.
[5] Hibar, D., et al. Common genetic variants influence human subcortical brain structures.(2015), Nature 520, 224–229 . https://doi.org/10.1038/nature14101
[6] Willer, C.J., et al. (2010). METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26, 2190–2191.
[7] Bulik-Sullivan, B. et al. (2015). LD score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat. Genet. 47, 291–295.
[8] Werme, J., et al (2022). An integrated framework for local genetic correlation analysis. Nat. Genet. 54, 274–282.
[9] Zhu, Z., et al. (2018). Causal associations between risk factors and common diseases inferred from GWAS summary data. Nat. Comm
[10]Finucane, H.K. et al.(2015), Partitioning heritability by functional annotation using genome-wide association summary statistics. . Nat. Genet. 47, 1228–1235.
[11]Finucane, H.K., et al. (2018). Heritability enrichment of specifically expressed genes identifies disease-relevant tissues and cell types. Nat. Genet. 50, 621–629.
[12] de Leeuw, C.,et al. (2023). On the interpretation of transcriptome-wide association studies. PLoS Genet. 19, e1010921.
Acknowledgements
This work was supported by the National Institutes of Health (Grant Nos. R01 MH1340004 and R01 AG059874 [to NJ], National Science Foundation Graduate Research Fellowship Program (Grant No. 2020290241 [to RRB], the the Adolescent Brain Cognitive Development (ABCD) Study (https://abcdstudy.org), and UK Biobank (Resource Application No. 11559), and P41EB015922 (PI: Toga/Thompson).