Screenshot

Training

In the two toy-data examples the target population was a non-specific African (AFR) population and the model population was European (EUR). In these examples each population had a corresponding:

  • LD Reference Panel (Used for Clumping)
  • GWAS Summary Stat File
  • Genotype/Phenotype Data

This example may be unrealistic given real-world constraints and the realities of data sharing. Here we will consider will consider using BridgePRS in more realistic scenarios. In each of these challeges the task is to create valid configuration given the described scenario.

Challenge 1:

I have access to genotype/phenotype data (<1000 samples, continuous phenotype) from a diverse population in Central Africa (1,2). I would like ot run PRS as accurately as possible, but my dataset is not large enough to conduct GWAS, however, I do have access to a moderately sized GWAS in a related West African population (3) and access to large-scale GWAS data from a European population (4). I also have access to the 1000G reference panels, how can I analyze my data?

Available Data:

  1. caf/genotypes/caf_genotype.bed, caf_genotype.bim, caf_genotype.fam
  2. caf/phenotypes/caf_test.dat, ukb/phenotypes/caf_validation.dat
  3. yoruba/sumstats/yoruba.chr1.glm.gz,...,yoruba.chr22.glm.gz
  4. ukb/sumstats/ukb.sumstats.gz
  5. 1000G Reference Panel

Can you fill out the configuration files to carry this out? 

POP=
LDPOP=
SUMSTATS_PREFIX=
GENOTYPE_PREFIX=
PHENOTYPE_FILE=
VALIDATION_FILE=

POP=CAF
LDPOP=AFR 
SUMSTATS_PREFIX=yoruba/sumstats/yoruba.chr 
SUMSTATS_SUFFIX=.glm.gz 
GENOTYPE_PREFIX=caf/genotypes/caf_genotype 
PHENOTYPE_FILE=caf/phenotypes/caf_test.dat  
VALIDATION_FILE=caf/phenotypes/caf_validation.dat 

POP=
LDPOP= 
SUMSTATS_PREFIX=
GENOTYPE_PREFIX=
PHENOTYPE_FILES=

POP=EUR
LDPOP=EUR
SUMSTATS_PREFIX=ukb/sumstats/ukb.sumstats
GENOTYPE_PREFIX=caf/genotypes/caf_genotype 
PHENOTYPE_FILE=caf/phenotypes/caf_test.dat  
VALIDATION_FILE=caf/phenotypes/caf_validation.dat

Challenge 2:

**I have access to GWAS data (1) and genotype/phenotype (2,3) data (>2000 samples, binary phenotype) from a moderate sized population in Eastern Europe. I have reason to believe that this population has unique LD structure and would like to incorporate this into my model.
I also have GWAS (4) and genotype/phenotype data from the UKB biobank (5,6) that I wish to use to improve my results, how should I do this?

  1. ukr/sumstats/ukr.sumstats.out
  2. ukr/genotypes/chr1.bed,bim,fam...chr22.bed,bin,fam,
  3. ukr/phenotypes/ukr_test.dat, ukr/phenotypes/ukr_validation.dat
  4. ukb/sumstats/ukb.sumstats.gz
  5. ukb/genotypes/chr1.bed,bim,fam...chr22.bed,bin,fam
  6. ukb/phenotypes/ukb_test.dat, ukb/phenotypes/ukb_validation.dat 
POP=
LDPOP=
SUMSTATS_PREFIX=
GENOTYPE_PREFIX=
PHENOTYPE_FILE=
VALIDATION_FILE=

POP=ukr
LDPOP=ukr*
SUMSTATS_PREFIX=ukr/sumstats/ukr.sumstats  
GENOTYPE_PREFIX=ukr/phenotypes/chr
PHENOTYPE_FILE=ukr/phenotypes/ukr_test.dat 
VALIDATION_FILE=ukr/phenotypes/ukr_validation.dat

POP=
LDPOP=
SUMSTATS_PREFIX=
GENOTYPE_PREFIX=
PHENOTYPE_FILE=
VALIDATION_FILE=

POP=UKB
LDPOP=EUR 
SUMSTATS_PREFIX=ukb/sumstats/ukb.sumstats
GENOTYPE_PREFIX=ukb/genotypes/chr
PHENOTYPE_FILE=ukb/phenotypes/ukb_test.dat 
VALIDATION_FILE=ukb/phenotypes/ukb_validation.dat