Manipulating GRM

--grm test
or
--grm-bin test
Input the GRM generated by --make-grm option. This option actually tells GCTA to read three files, e.g. test.grm.bin, test.grm.N.bin and test.grm.id (See the option --make-grm). GCTA automatically adds suffix “.grm.bin”, “.grm.N.bin” or “.grm.id” to the specified root filename. If the test.grm.N.bin file (which contains the number of SNPs used to calculate GRM) is missing, the program will still be running because all the analysis except --grm do not actually need the the number of SNPs used to calculate the GRM.

--grm-gz test
To be compatible with the previous version of GCTA. Same as --grm but read the GRM files in compressed text format generated by --make-grm-gz option. This option actually tells GCTA to read two files, e.g. test.grm.gz and test.grm.id (See the option --make-grm-gz). GCTA automatically adds suffix “.grm.gz” and “.grm.id” to the specified root filename.

Examples: converting the two formats from each other
# From *.grm.gz to *.grm.bin
gcta64  --grm-gz test  --make-grm  --out test
# From *.grm.bin to *.grm.gz
gcta64  --grm test  --make-grm-gz --out test

--mgrm multi_grm.txt
or
--mgrm-bin multi_grm.txt
Input multiple GRMs in binary format (See the option --make-grm). The root filenames of multiple GRMs are given in a file, e.g. multi_grm.txt
Input file format
multi_grm.txt (full paths can be specified if the GRM files are in different directories)
test_chr1
test_chr2
test_chr3
……
test_chr22

--mgrm-gz multi_grm.txt
To be compatible with the previous version of GCTA. Same as --mgrm but read the GRM files in compressed text format generated by --make-grm-gz.

Examples
# This option is very useful to deal with large dataset. You can firstly run the jobs (split one job into 22 pieces)
gcta64  --bfile test  --chr 1  --make-grm  --out test_chr1
gcta64  --bfile test  --chr 2  --make-grm --out test_chr2
…
gcta64  --bfile test  --chr 22  --make-grm  --out test_chr22

# To estimate the GRMs from the SNPs on each chromosome, then merge them by the command
gcta64  --mgrm multi_grm.txt  --make-grm  --out test

--grm-cutoff 0.025
Remove one of a pair of individuals with estimated relatedness larger than the specified cut-off value (e.g. 0.025). GCTA selectively removes individuals to maximize the remaining sample size rather than doing it at random. NOTE: When merging multiple GRMs, this option does not apply to each single GRM but to the final merged GRM.

--grm-adj 0
When using the SNPs to predict the genetic relationship at causal loci, we have to adjust the prediction errors due to imperfect LD because of two reasons: 1) the use of only a finite number of SNPs; 2) causal loci tend to have lower MAF than the genotyped SNPs (input 0 if you assume that the causal loci have similar distribution of allele frequencies as the genotyped SNPs) (see Yang et al. 2010 Nat Genet for details).

--dc 1
By default, the GRM, especially for the X-chromosome, is parameterized under the assumption of equal variance for males and females, unless the option --dc is specified (1 and 0 for full and no dosage compensation, respectively). You need to use the option --update-sex to read sex information of the individuals from a file (see the --update-sex option above).

NOTE: you can add the option --make-grm or --make-grm-gz afterwards to save the modified GRM. You can also use the option --keep and/or --remove in combination with these five commands. It is also possible to use these five commands in the REML analysis (see the section below).

Examples
# Prune the GRM by a cutoff of 0.025 and adjust for prediction errors assuming the causal variants have similar distribution of allele frequencies as the genotyped SNPs)
gcta64  --grm test  --grm-adj  0  --grm-cutoff  0.025  --make-grm  --out test_adj

# Use --keep or --remove option
gcta64  --grm test  --keep test.indi.list  --grm-cutoff  0.025  --make-grm  --out test_adj
gcta64  --grm test  --remove test.indi.list  --grm-adj 0  --make-grm  --out test_adj

# Assume full and no dosage compensation for the X chromosome
gcta64  --grm test_xchr  --dosage-compen 1  --update-sex test.indi.sex.list  --make-grm  --out test_xchr_fdc
gcta64  --grm test_xchr  --dosage-compen 0  --update-sex test.indi.sex.list  --make-grm  --out test_xchr_ndc


References

Method for estimating the GRM: Yang et al. (2010) Common SNPs explain a large proportion of the heritability for human height. Nat Genet. 42(7): 565-9. [PubMed ID: 20562875]

Method for estimating the GRM for the X chromosome and GCTA software: Yang J, Lee SH, Goddard ME and Visscher PM. GCTA: a tool for Genome-wide Complex Trait Analysis. Am J Hum Genet. 2011 Jan 88(1): 76-82. [PubMed ID: 21167468]

A demonstration of estimating variance explained by the X chromosome for height and BMI: Yang et al. (2011) Genome partitioning of genetic variation for complex traits using common SNPs. Nat Genet. 43(6): 519-525. [PubMed ID: 21552263]