#Explanation of what scripts do

• eQTLs_scripts/create_eQTL_summary.R creates a summary file by running the snpStats analysis on all identified cis-eQTL loci based on matrixeQTL, then pulls the best SNP based on P-value and reports that best SNP in the summary table.
• Pickrell/create_Pickrell_input.R creates the "locus based" summary file with P-values, alleles, effect size... in one genomic region

List of required packages

• MatrixEQTL
• snpStats
• ggplot2

It may be useful to run:

scl enable devtoolset-1.1 'bash'

#Explanation of eQTL pipeline

Explanation of what folders contain

expression_data folder

one file for each "condition", which can be a cell type, a time point, an activation for format of each file must be:

• expression_[condition].RData where [condition] will vary depending on the data.
• Each of these RData file must contain two R objects: [condition] and support.[condition] Condition is a matrix of numbers, with row.names that match the ProbeID column of the support file and colnames that match the different individuals in the dataset.

The support.[condition] includes the following columns:

• row names must be a unique identified, typically the probe (mandatory)
• ensemblID (mandatory)
• Gene.name (mandatory, human readable version of ensemblID)
• gene.chromosome (optional, if missing inferred from ensemblID)
• gene.position.start (optional, if missing inferred from ensemblID)
• gene.position.end (optional, if missing inferred from ensemblID)

The scripts assume the expression data has already been normalised.

what if I use exon instead of gene based probes?

ensemblID must remain to mark the gene but Gene.name becomes something like SORT1_ex5 to identify the exon of interest.

genotypes folder

One file per chromosome, stored in snpStats format, called

chr1
chr2
...
chr22
chrX
chrY
(no RData suffix, for no good reason)

The name of the object must be genotypes in each file. The genotypes R object is a list that contains:

• Mandatory: genotypes$genotypes which is a snpStats object, snps ideally in rsid format
• Mandatory: genotypes$map file, with columns: SNP, allele.1, allele.2, position. Make sure position is typed as integer and that the rownames of the genotype$map data frame are the unique snp.names
• Optional: a ped file, genotypes$fam

phenotypes folder

Two potential files:

• binary.RData contain a single object named binary.pheno
• continuous.RData contains a single object named continuous.pheno

numeric matrix, columns are measurements and rows are sample IDs that MUST match the genotypes Cut out columns to leave only metabolites Don't forget the row and column names.

covariates folder

basic is a file called *covariates.tab tab delimited, must contain the colum "id" that matches the sample names

######## Summary stats file, choice of headers:

• SNPID, CHR, POS, then BETA.biom, SE.biom, PVAL.biom, N.biom in biom_chrXX.RData where biom is a code (ht), df
• SNPID, CHR, POS, then BETA.cc, SE.cc, PVAL.cc, N.cc, Ncases.cc in cc_chrXX.RData where cc is a code (SCZ...).df name of dataframes: biomarker or cc

automated generation of these files from : genotypes, phenotypes, and covariates