FunBurd is designed to test the association between variants aggregated across a gene set and a given trait. This repository contains all code and data required to reproduce the figures and statistical analyses presented in the project.
Developed and maintained by Sayeh Kazem (Jacquemont Lab).
Bird_Logo_Video_Genetic_Analysis.mp4
Determinants of functional burden pleiotropy and gene dosage responses across human traits
If you use this project or its code in your research or pipelines, please cite the paper. Your citation helps others discover the project and acknowledges our work.
Paper
Kazem, S., Kumar, K., Yang, J. et al. Determinants of functional burden pleiotropy and gene dosage responses across human traits. Nat Commun (2026). https://doi.org/10.1038/s41467-026-76676-0
@article{kazem_funburd_2026,
author = {Kazem, Sayeh and Kumar, Kuldeep and Yang, Jane and Beniti{\`e}re, Florian and Huguet, Guillaume and Mollon, Josephine and Renne, Thomas and Schultz, Laura M. and Knowles, Emma E. M. and Engchuan, Worrawat and Shanta, Omar and Thiruvahindrapuram, Bhooma and MacDonald, Jeffrey R. and Greenwood, Celia M. T. and Scherer, Stephen W. and Almasy, Laura and Sebat, Jonathan and Glahn, David C. and Dumas, Guillaume and Jacquemont, S{\'e}bastien},
title = {Determinants of functional burden pleiotropy and gene dosage responses across human traits},
journal = {Nature Communications},
year = {2026},
doi = {10.1038/s41467-026-76676-0},
url = {https://doi.org/10.1038/s41467-026-76676-0}
}Software
To cite this specific version of the code:
Kazem, S. (2026). FunBurd: Functional Burden Association Test (v1.1.0). Zenodo. https://doi.org/10.5281/zenodo.17369429
@software{kazem_funburd_software_2026,
author = {Kazem, Sayeh},
title = {FunBurd: Functional Burden Association Test},
version = {1.1.0},
year = {2026},
publisher = {Zenodo},
doi = {10.5281/zenodo.17369429},
url = {https://doi.org/10.5281/zenodo.17369429}
}Functional_Burden_Association_Analysis/— FunBurd: FunBurd is designed to test the association between variants aggregated across a gene set and a given trait. The traits of interest were considered as a function of the number of genes within the gene set disrupted by CNVs. To avoid effect size inflation, due to multigenic CNVs, we adjusted for the number of genes (not members of the gene set) disrupted by the same CNV. We also adjusted for age, sex, and ancestry:
Provided as both a notebook (FunBurd_Multitrait_EffectSizes_Computation_Script_On_ToyDatasets.ipynb) and an equivalent command-line script (.py), so the same code can be run interactively or launched as a batch job. Since UK Biobank / All of Us genotypes can't be redistributed, the folder ships with a synthetic toy dataset (Genotypic_info_Toy/, Phenotypic_info_Toy/, Outcome_Toy/) so the full pipeline can be run end-to-end, plus the three functional gene set collections used throughout the paper (Gene_Sets/): 60 whole-body tissue sets, 81 whole-body cell-type sets, and 31 whole-brain cell-type (Siletti) sets. Slurm_FunBurd.sh and Args_File_FunBurd.txt show how to launch the script as a SLURM/glost array job across many trait × gene-set combinations.
The following scripts and data (Figure_Generation/) generate the main figures for the study:
- Figure 2: Tissue-specific associations of CNVs with complex traits.
- Figure 3: Whole-body cell type-specific associations of CNVs with complex traits.
- Figure 4: Dissecting pleiotropy, gene function, and genetic constraint.
- Figure 5: Rare and common variant architectures across complex traits.
- Figure 6: Gene dosage responses across traits and functional gene sets.
Each FigureN.Rmd / FigureN.md pair is a self-contained, rendered analysis (data loading → statistics → plotting), with output plots saved alongside in the matching FigureN_files/ folder.
Permutation_PreservingJaccardDistance_P_Jaccard/— P-Jaccard: A permutation-based method for assessing gene set associations that conditions on the degree of overlap between gene sets using the Jaccard index, avoiding inflated significance for gene sets that share many genes. For each trait, the method generates 1,000 Jaccard-preserving surrogate gene sets (viaBrainSMASH) to build a null distribution for the deletion–duplication burden correlation, producing two p-values per trait — one fixing deletions and surrogating duplications, one fixing duplications and surrogating deletions — which are then FDR-corrected (PJaccard/,JaccardDist/,surrogate_maps_DelDup/).
- Python (
*.ipynb,*.py): FunBurd effect-size computation and P-Jaccard computation. - R Markdown / Markdown (
*.Rmd,*.md): Downstream statistical analyses and figure generation. - R Data Files (
*.RData): Processed data underlying each figure. - Tabular / text data (
*.tsv,*.csv,*.txt): Toy genotype/phenotype inputs, gene set definitions, and effect-size / P-Jaccard outputs. - NumPy arrays (
*.npy): Jaccard-preserving surrogate maps used to build null distributions. - Cluster scripts (
*.sh+ args files): SLURM/glosttemplates for running FunBurd across many traits and gene sets in parallel.
- Clone the repository:
git clone https://github.com/JacquemontLab/FunBurd.git- Try FunBurd on the provided toy dataset:
cd FunBurd/Functional_Burden_Association_Analysis
python3 FunBurd_Multitrait_EffectSizes_Computation_Script_On_ToyDatasets.py \
Phenotypic_info_Toy/phenotype1_toy.tsv phenotype1 phenotype1 \
Gene_Sets/GeneSet_HPA_tissue_fantom_nTPM HPA_tissue_fantom_nTPMThe script takes five positional arguments — phenotype file, phenotype column, phenotype name, gene-set directory, and gene-set name — and writes effect sizes to Outcome_Toy/. See Args_File_FunBurd.txt for the full list of trait × gene-set combinations, and Slurm_FunBurd.sh for running them all as a parallel batch job.
- Python (FunBurd):
numpy,pandas,statsmodels,patsy,firthlogist - Python (P-Jaccard):
numpy,pandas,scipy,statsmodels,matplotlib,nibabel,brainsmash - R (Figure Generation):
ggplot2,ggpubr,ggprism,ggcorrplot,ggraph,ggrepel,ComplexHeatmap,circlize,igraph,vegan,dplyr,tidyr,tibble,forcats,stringr,reshape2,patchwork,gridExtra,openxlsx,svglite,knitr,MASS,Rmpfr
This project is released under the MIT License.