This page contains some files that might be useful when running LDAK. Please note you can download the data used in the examples on this website from Test Datasets.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Repositories of GWAS summary statistics:
Here are some links to websites that provide GWAS summary statistics, which may be useful when running MegaPRS or GigaPRS. Please note that these websites generally performed bulk GWAS (i.e., analyzed 100s of phenotypes using automated pipelines). If you are interested in a specific trait, it is often better to Google for the latest bespoke GWAS (if published in a Nature journal, the article should include a Data Availability section, which explains how to access the summary statistics).
UK Biobank (click on the phenotype manifest that contains download links)
FinnGen (once you register, you are emailed an access link)
Million Veterans Project: click here for article; click here for FTP link
Taiwan Precision Medicine Initiative (note that it is faster to download PLINK 2 results)
Biobank Japan (there are four different papers, each containing different results)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Bulk summary statistics:
The file below provides results from 52 European GWAS of heritable quantitative traits (average sample size 280k) that can be used as bulk summary statistics with MegaPRS or GigaPRS.
European bulk summary statistics
Please note that the above file is gzipped and must be decompressed before use (on a UNIX operating system, you can use the command gunzip).
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Human Genome LD Breakpoints:
The files below can be used when applying MegaPRS to human data (specifically, they tell LDAK how to partition the genome when calculating SNP-SNP correlations). Note that these breakpoints were generated by Zheng et al, who merged the LD blocks previously identified by Berisa and Pickrell.
Berisa hg19/GRCh37 Assembly Breakpoints
Berisa hg38/GRCh38 Assembly Breakpoints
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Baseline LD Model SNP Annotations:
When applying MegaPRS to human data, we now recommend using the SNP annotations from the Baseline LD Model (version 2.2), developed by Marquez-Luna et al. (we have found these are slightly more accurately than the BLD-LDAK Model annotations). Download these by clicking on the link below.
You can specify these annotations when running MegaPRS by adding the options --annotation-prefix BaselineLD and --annotation-number 86. Note that the file above actually contains 96 annotations, but we only recommend using the first 86 (the final ten are MAF bins, which we find unnecessary, but include for completeness).
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Genetic Map:
The file below contain genetic distances for the human genome. I use the distances when constructing a Reference Panel from the 1000 Genomes Project. I believe I originally obtained the file from the (no longer updated) IMPUTE2 website.
Genetic Map (hg19 assembly)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
1000 Genome Project Reference Panels:
I generated European, American, South Asian, East Asian and African reference panels by following the scripts on Reference Panel (first for non-Finnish Europeans, then Americans, then South Asians, then East Asians, then African individuals). Each panel contains between 7.5M and 14.8M SNPs (all with MAF>0.01) and between 347 and 661 individuals. Genomic positions correspond to the hg19/GRCh37 assembly. Note that these sample sizes are relatively small; ideally your reference panel should contain at least 2000 individuals.
If you are using a UNIX operating system, you can download a file by typing wget followed by its web address (you can obtain the web address of the files below by right-clicking on the corresponding links). Having downloaded a set of files, you should extract it. On a UNIX operating system, you can extract a set of files by typing tar -xzvf followed by the file name.
404 non-Finnish European individuals
347 American individuals
489 South Asian individuals
504 East Asian individuals
661 African individuals
I have also created reduced versions of the above five datasets (each contains approximately 1.1M SNPs) by retaining only HapMap3 SNPs (I use these for the Short Courses).
Reduced version of 404 non-Finnish European individuals
Reduced version of 347 American individuals
Reduced version of 489 South Asian individuals
Reduced version of 504 East Asian individuals
Reduced version of 661 African individuals
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Gene Annotations:
RefSeq Annotations (Assembly hg19/GRCh37)
RefSeq Annotations (Assembly hg38/GRCh38)
I obtained these gene annotations from the UCSC Genome Browser Tables Browser. Having picked the species and genome assembly, I selected the group "Genes and Gene Predictions" and the track "NCBI RefSeq". I made sure "genome" was ticked (next to region), then chose "selected fields from primary and related tables" and entered the output filename "refseq". Having clicked "get output", I ticked the fields "chrom", "strand", "txStart", "txEnd" and "name2" (you may prefer to instead tick "chrom", "strand", "cdsStart", "cdsEnd" and "name2"). Finally, I clicked "get output" to download the file (its size was about 3Mb).
I then processed the downloaded file using the following unix commands (you can read more about AWK here).
rm refseq.clean
for j in {1..22}; do
awk -v j=$j '($1=="chr"j){print $5, j, $3, $4, $4-$3, $2}' refseq | sort -r -n -k 5,5 | awk '(!seen[$1]++){print $1, $2, $3, $4, $6}' | sort -n -k 3,3 >> refseq.clean
done
The file refseq.clean is now in the format required by LDAK. Note that this command excluded a few genes with unusual chromosome names (e.g., gene "LOC389831" has chromosome "chr7_gl000195_random"). Further, for genes with the same name, it only kept the longest. Finally, it ensured that genes were ordered by chromosome then start basepair.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Pathway Annotations:
MSigDB Annotations (all 16768)
MSigDB Annotations (subset of 6000)
Takiy created these files using gene pathways curated on the MSigDB website (either all 16,768 pathways, or a subset of 6000).
