Q1: What are the symbols (e.g. "C", "F", "R[E]", "R[G]") in model equation of LOG file?
The meanings of the symbols are below:
C: quantitative covariates
F: discrete covariates, also is called as fixed effect
F[ExE]: environmental by environmental fixed effect
R[E]: environmental random effect
R[ExE]: environmental by environmental random effect
R[G]: genetic random effect
R[GxG]: genetic by genetic random effect
R[GxE]: genetic by environmental random effect
PA: pedigree based additive relationship matrix
PD: pedigree based dominant relationship matrix
GA: genome based additive relationship matrix
GD: genome based dominant relationship matrix
HA: pedigree and genome based additive relationship matrix
HD: pedigree and genome based dominant relationship matrix
Q2: How to decrease the memory cost for analysis?
There are two strategies:
1. assign lower steps to the option --step when genotype is used for analysis.
2. add flag --float-prec in the command, it can nearly save half of memory cost, but it should be noted that mathematical operation in float precision may lose the accuracy of estimated parameters for specific cases.
Q3: Can options "--extract", "--exclude", "--remove", "--keep" work for all functions in HIBLUP?
Yes, if the analysis involves individuals and SNPs, all above options can work effectively.
Q4: Can HIBLUP estimate genetic correlation for environmental random effect?
Yes, users can firstly construct XRM for this environmental random effect (see Construct relationship matrix), then assign it to flag --xrm when fitting the multiple traits model.
Q5: Can users assign the relationship matrix in text format to HIBLUP directly?
By now, HIBLUP can only accept relationship matrix in binary format (*.bin and *.id), relationship matrix in text format is unacceptable, users can convert it to binary format by the function --trans-xrm (see Format conversion of XRM).
Q6: How to compute the proportion of phenotypic variance explained (PVE) for significant SNPs by HIBLUP?
It’s easy to be achieved by HIBLUP. Firstly, please write the id of significant SNPs into a file snp.txt as a column without header, then calculate the variance components by following steps:
#step1: construct GRM1 using the significant SNPs
./hiblup --make-xrm
--bfile demo
--extract snp.txt
--out grm1
--threads 32 # grm1.GA.id, grm1.GA.bin will be generated
#step2: construct GRM2 using the SNPs excluding the significant SNPs
./hiblup --make-xrm
--bfile demo
--exclude snp.txt
--out grm2
--threads 32 # grm2.GA.id, grm2.GA.bin will be generated
#step3: compute the variance components for GRM1 and GRM2 by single trait model
./hiblup --single-trait
--pheno demo.phe
--pheno-pos X
--xrm grm1.GA,grm2.GA
--out vc
--threads 32
#step4: calculate the PVE for significant SNPs by following equation:
pve = V(grm1) / ( V(grm1)+V(grm2)+V(e) )