Summary statistics spec

The results of GWAS are represented as summary statistics, with the following columns:

• SNP - marker name, for example rs#.

• CHR - chromosome label

• BP - base-pair position

• A1 - effect allele for Z and BETA columns

• A2 - other allele

• N - sample size

• CaseN, ControlN - sample size for cases and controls (logistic regression only)

• FRQ - frequency of A1 allele

• Z - z-score (or t-score) of association

• BETA - effect size; for logistic regression, this contains log(OR)

• SE - standard error of the BETA column

• L95, U95 - lower and upper 95% confidence interval of the BETA.

• P - p-value

For SNP, CHR, BP, A1 and A2 columns the scripts/gwas/gwas.py script will simply copy over the information from the genetic file, i.e. from .bgen or .bim files. This means that SNP is likely to be dbSNP rs#, or some other form of identifyied such as CHR:BP:A1:A2. For CHR and BP, there we don’t enforce a specific genomic build - it all depends on what build was used by the genotype data. Finally, A1 and A2 are not guarantied to be minor or major alleles, but A1 will be used as an effect allele for signed summary statistics (i.e. Z and BETA columns).

The sample size N is as reported by the software (plink2 or regenie). For case-control traits, this appears to be a sum of cases and controls (not the effective sample size which would take into account imbalance between cases and controls).

L95 and U95 columns are only provided for plink2 results. CaseN and ControlN columns are only provided for plink2 results for logistic regression. If you need these columns for regenie analysis consider also running plink2 analysis, and copy over the columns into your regenie output.

Comparison of columns names

• CoMorMent: this file

• LDSC: https://github.com/precimed/ldsc/blob/master/munge_sumstats.py

• BioPsyk: https://github.com/BioPsyk/cleansumstats/blob/dev/assets/schemas/cleaned-sumstats.yaml

• NORMENT: https://github.com/precimed/python_convert/blob/master/sumstats_utils.py

CoMorMent	daner	LDSC	BioPsyk	NORMENT	Description
missing	?	missing	0	missing	good idea to provide this column and referencing a line in .bim file
CHR	CHR	CHR	CHR	CHR	OK
BP	BP	BP	POS	BP	keep BP which is more informative ( “POS” could also stand for genomic position )
SNP	SNP	SNP	RSID	SNP	keep SNP which makes more sense as we copy over marker name from genetic file
A1	A1	A2	EffectAllele	A1	keep A1 for consistency with LDSC even thought EffectAllele is more informative
A2	A2	A2	OtherAllele	A2	keep A2 for consistency with LDSC even though OtherAllele is more informative
P	P	P	P	PVAL	OK
SE	SE	SE	SE	SE	OK
L95	?	missing	ORL95	missing	keep “L95” as confidence interval may also be for the BETA or LOG(OR)
U95	?	missing	ORU95	missing	keep “U95”
N	?	N	N	N	OK
CaseN	Nca	N_CAS	CaseN	NCASE	OK
ControlN	Nco	N_CON	ControlN	NCONTROL	OK
INFO	INFO	INFO	INFO	INFO	OK
Direction	Direction	missing	Direction	DIRECTION	OK
BETA	BETA or OR	BETA	B	BETA or OR	keep “BETA” for consistency with LDSC (and also BETA is more informative)
Z	?	Z	Z	Z	OK
FRQ	FRQ_A_NNN	FRQ	EAF	FRQ	keep “FRQ” which makes more sense for non-EUR populations
missing	?	missing	EAF_1KG	missing	not needed

Change log

• v0.9 - first version of this document