README.md

In yoki5348/VT_SEM: Vertical-Transmission Structural Equation Modeling (VT-SEM)

VT_SEM

Our project aims to develop the structural equation model (SEM) for estimating the vertical transmitted environment factors via polygenic risk score (PRS) of Trio data

Installing VT-SEM

To install the package, use following codes in R:

require("devtools")

devtools::install_github("yoki5348/VT_SEM")

Example

library(VTSEM)

num.cvs <- 100 #The number of causal variants

am <- 0.25 # assortative mating parameter, mu

herit0 <- 0.5 #Inital heritability

vf0 <- 0.15 #Initial variance of vertical transmitted environment factor

Unequal_AM=TRUE # TRUE: Generating disequilibrium AM dataset, FALSE: Generating equilibrium AM dataset

prop.h2.latent <- 0.5 #Proportion of heritability

seed <- 1234567

max.cores <- 1 ##Number of used cores for OpenMx

num.gen <- 20 #Number of generating generation

pop.size <- 1000 #The number of independent trio dataset

num.its <- 1000

RUN.MARKERS <- FALSE #whether to only consider GRMs built from CVs (FALSE) or both CVs and SNPs (TRUE)

avoid.inb <- FALSE

save.covariances <- TRUE

save.history <- TRUE

MIN.MAF <- .1; MAX.MAF <- .50#AM Simulation wildcards

MAF.VECTOR <- runif(num.cvs,MIN.MAF,MAX.MAF) #Can change the distribution of MAFs here

GENTP.VAR <- MAF.VECTOR(1-MAF.VECTOR)2

ALPHA.VECTOR <- sample(c(-1,1),num.cvs,replace=TRUE)*sqrt(1/(num.cvs*GENTP.VAR)) #Can change the distribution of effect sizes here - fixed f'n of MAF

CV.INFO <- data.frame(MAF=MAF.VECTOR,alpha=ALPHA.VECTOR) #we'll use this for both the observed and latent

AM.DATA <- AM.SIMULATE(CV.INFO=CV.INFO, H2.T0=herit0, NUM.GENERATIONS=num.gen, POP.SIZE=pop.size*3, MATE.COR=am, AVOID.INB=avoid.inb, SAVE.EACH.GEN=save.history, SAVE.COVS=save.covariances, SEED=seed, VF.T0=vf0,PROP.H2.LATENT=prop.h2.latent,Unequal_AM=Unequal_AM) #Dataset generation

dat_init=as.data.frame(AM.DATA$PHEN)

dat_init=dat_init[sample(1:nrow(dat_init),pop.size),] ##independent dataset check

model0=perform_SEM_model0(dat_init,"NTMO","TMO","NTPO","TPO","Y")

model1_eq=perform_SEM_model1_eq(dat_init,"NTMO","TMO","NTPO","TPO","Y")

model1_dis=perform_SEM_model1_dis(dat_init,"NTMO","TMO","NTPO","TPO","Y")

model2_eq=perform_SEM_model2_eq(dat_init,"NTMO","TMO","NTPO","TPO","YM","YP","Y")

model2_dis=perform_SEM_model2_dis(dat_init,"NTMO","TMO","NTPO","TPO","YM","YP","Y")

model2_eq_NP=perform_SEM_model2_eq_NP(dat_init,"NTMO","TMO","NTPO","TPO","Y",herit0,prop.h2.latent)

model2_dis_NP=perform_SEM_model2_dis_NP(dat_init,"NTMO","TMO","NTPO","TPO","Y",herit0,prop.h2.latent)

results=rbind(model0$result_summary,model1_eq$result_summary,model1_dis$result_summary,model2_eq$result_summary,model2_dis$result_summary, model2_eq_NP$result_summary,model2_dis_NP$result_summary)

print(results) ##Print results

summary(model0$AFE.Fit) #Check the fitted results

Simulated confidence interval for several parameters

These functions gives user simulated confidence intervals for some parameters such as VF,r2PGS,r2LGS, and w.

n:number of trio

r2PGS:true R^2 of observed PGS

r2LGS: true R^2 of latent PGS

VF: true explained variance of familial shared environmental factors

w: true cov(F,Y)

percentile: probability ranges of confidence interval (e.g. percentile=95, 95% C.I will be returned)

c.i.compute_VF=function(n,r2PGS,VF,percentile) : Confidence interval of VF

c.i.compute_r2PGS=function(n,r2PGS,percentile): Confidence interval of R^2 of PGS

c.i.compute_r2LGS=function(n,r2PGS,r2LGS,percentile) : Confidence interval of R^2 of LGS

c.i.compute_w=function(n,r2PGS,w,percentile) : Confidence interval of w

yoki5348/VT_SEM documentation built on July 24, 2021, 5:10 p.m.