Significant sQTLs and splicing TWAS reference panels (AMP-AD brain and EADB Belgian LCL cohorts)

This dataset is part of the manuscript "New insights into the genetic etiology of Alzheimer’s disease and related dementias" by Bellenguez, Küçükali, et al. Nature Genetics 2022.

Publication link: https://www.nature.com/articles/s41588-022-01024-z

GitHub repository of all QTL/TWAS data shared for this study: https://github.com/SleegersLab-VIBCMN/EADB_GWAS_NatureGenetics_QTL_TWAS

For details, please see the publication. For any questions, please contact Fahri Küçükali (fahri.kucukali@uantwerpen.vib.be) and Kristel Sleegers (Kristel.Sleegers@uantwerpen.vib.be).

Significant sQTL catalogues are compressed with gzip and tar achieve of splicing TWAS reference panels are compressed with bzip2.

sQTL catalogues

The files show significant sQTL - splice junction pairs mapped in AMP-AD brain and EADB Belgian LCL cohorts. The catalogues are in hg38/GRCh38 human genome build. Most of the columns in the files are based on FastQTL output (http://fastqtl.sourceforge.net/).

sQTL file columns:

1. variant_id - ID of the significant sQTL variant based on the dbSNPv151 rsID annotation or hg38/GRCh38 CHR_POS_REF_ALT ID if rsID not available.
2. junc_id - sJunction ID assigned by Regtools/Leafcutter pipeline. For stranded datasets (ROSMAP and EADB Belgian) strand info is provided with "+" or "-" symbols, and if not stranded, "?" symbol is used.
3. junc_distance - Genomic distance between sQTL variant and splice junction start
4. ma_samples - Number of samples carrying the minor allele
5. ma_count - Total count of minor alleles
6. maf - Minor allele frequency
7. pval_nominal - Nominal P-value of the association
8. slope - Slope of the association with respect to alternative (ALT) allele indicated on column 14
9. slope_se - Standard error of the slope
10. pval_nominal_threshold - Nominal P-value significant threshold for thissJunction
11. min_pval_nominal - Most significant P-value observed for this sJunction
12. pval_beta - permutation P-value obtained via beta approximation and later used to calculate Storey q-values
13. junction - sJunction in chr:start-end splice junction format
14. cluster - The splice cluster of this sJunction
15. genes - Genes overlapping with this sJunction (if any), based on GENCODEv24 (AMP-AD) and GENCODEv32 (EADB Belgian)
16. GRCh38_chr_pos - Genomic position of the variant, separated by underscore
17. ref_alt - Reference (REF) and alternative (ALT) allele of the variant, separated by ">" sign. ALT is the tested (A1) allele

Of note, we also mapped the significant eQTLs in the same datasets (please see the data availability section of the manuscript or the GitHub repository).

Splicing TWAS reference panels

Custom splicing TWAS reference panels prepared using FUSION pipeline (http://gusevlab.org/projects/fusion/) in AMP-AD brain and EADB Belgian LCL cohorts. All data in hg38/GRCh38 genome build. In each directory, you will find ".pos", ."profile", and ".profile.err" files. These are explained in the FUSION website as well, but briefly these are:

1. .pos: This is a position file that describes the 1Mb extended splice junction start and end coordinates for each calculated weight file for splice junction phenotype. Used for scanning the variants in those coordinates for TWAS.
2. .profile: This informs about all prediction weights calculated, in terms of number of variants in the model, heritability information, and R2 info for each prediction model used (top1, blup, enet, bslmm, lasso; bslmm was not used therefore has NA values).
3. .profile.err: This summarizes the reference panel in terms of average hsq (with SD), and which model is the best performing.

Each TWAS weight is provided in a .RDat file under All_Splicing_Weights, and in this data we included all calculated functional weights independent of the fact that they are heritable features or not. In our TWAS analyses, we included the heritable functional weights at a hsq P-value ≤ 0.05 level.

Please also see the expression TWAS reference panels we prepared in the same datasets (see the data availability section of the manuscript or the GitHub repository). If you need an LD reference data in hg38/GRCh38 genome build based on 1000 Genomes NFE samples (whose variant ID annotation are matching to these functional weights), suitable for running the TWAS/FUSION pipeline, please contact us.