KAML-package: KAML: Kinship Adjusted Multiple Loci Best Linear Unbiased...

In YinLiLin/KAML: Kinship Adjusted Multiple Loci Best Linear Unbiased Prediction

KAML: Kinship Adjusted Multiple Loci Best Linear Unbiased Prediction

Description

KAML is originally designed to predict phenotypic value using genome- or chromosome-wide SNPs for sample traits which are controled by limited major markers or complex traits that are influenced by many minor-polygene. In brief, KAML incorporates pseudo QTNs as fixed effects and a trait-specific K matrix as random effect in a mixed linear model. Both pseudo QTNs and trait-specific K matrix are optimized using a parallel-accelerated machine learning strategy.

Usage

KAML(pfile="", gfile="", kfile=NULL, dcovfile=NULL, qcovfile=NULL,
  pheno=1, SNP.weight=NULL, GWAS.model=c("MLM","GLM", "RR"), GWAS.npc=NULL,
  prior.QTN=NULL, prior.model=c("QTN+K", "QTN", "K"),
  vc.method=c("brent", "he", "emma"),
  Top.perc=c(1e-4, 1e-3, 1e-2, 1e-1), Top.num=15,
  Logx=c(1.01, 1.11, exp(1), 10), qtn.model=c("MR", "SR", "BF"),
  BF.threshold=NULL, binary=FALSE, bin.size=1000000, max.nQTN=TRUE,
  sample.num=2, SNP.filter=NULL, crv.num=5, cor.threshold=0.3,
  count.threshold=0.9, step=NULL,
  bisection.loop=10, ref.gwas=TRUE,
  theSeed=666666, file.output=TRUE, cpu=10
)

Arguments

pfile	phenotype file, one column for a trait, the name of each column must be provided(NA is allowed)
gfile	genotype files, including "gfile.geno.desc", "gfile.geno.bin" and "gfile.map"
kfile	n*n, optional, provided KINSHIP file for all individuals
dcovfile	n*x, optional, the provided discrete covariates file
qcovfile	n*x, optional, the provided quantitative covariates file
pheno	specify phenotype column in the phenotype file(default 1)
SNP.weight	provided weights of all SNPs
GWAS.model	which model will be used for GWAS(only "GLM" and "MLM" can be selected presently)
GWAS.npc	the number of PC that will be added as covariance to control population structure
prior.QTN	the prior QTNs which will be added as covariants, if provided prior QTNs, KAML will not optimize QTNs and model during cross-validation
prior.model	the prior Model for the prior.QTN that added as covariants
vc.method	method for variance components estimation("brent", "he", "emma", "ai")
Top.perc	a vector, a subset of top SNPs for each iteration are amplified when calculating KINSHIP
Top.num	a number, a subset of top SNPs for each iteration are used as covariants
Logx	a vector, the base for LOG
qtn.model	the strategy of selecting pseudo QTNs. c("MR", "SR", "BF")
BF.threshold	the threshold of BF method
binary	whether the phenotype is case-control
bin.size	the size of each bin
max.nQTN	whether limits the max number of Top.num
sample.num	the sample number of cross-validation
SNP.filter	the SNPs whose P-value below this threshold will be deleted
crv.num	the cross number of cross-validation
cor.threshold	if the top SNP which is used as covariant is in high correlation with others, it will be deleted
count.threshold	if the count of selected SNP for all iteration >= sample.num*crv.num*count.threshold, than it will be treated as covariance in final predicting model
step	to control the memory usage
bisection.loop	the max loop(iteration) number of bisection algorithm
ref.gwas	whether to do GWAS for reference population(if not, KAML will merge all GWAS results of cross-validation by mean)
theSeed	the random seed
file.output	whether to write the predicted values in file
cpu	the number of CPU for calculation

Details

Package:	KAML
Type:	Package
Version:	1.2.0
Date:	2021-11-04
License:	GPL(>=3)

Author(s)

Lilin Yin, Haohao Zhang and Xiaolei Liu
Maintainer:
Lilin Yin <[e-mail]>
Xiaolei Liu <[e-mail]>

Examples

Please see at: https://github.com/YinLiLin/KAML

YinLiLin/KAML documentation built on April 12, 2025, 5:49 a.m.