GEPHAST

GEnotype-PHenotype ASsociation Test

(Amos & Acevedo-Whitehouse, 2009)

# GEPHAST ####
gephast <- gtype_dfg2[65:174, 1:13]

gephast$age  <- mscpt_data[65:174, ]$age
gephast$mass <- mscpt_data[65:174, ]$mass
gephast$svl  <- mscpt_data[65:174, ]$svl
gephast$hw   <- mscpt_data[65:174, ]$hw
gephast$hl   <- mscpt_data[65:174, ]$hl
gephast$sex  <- mscpt_data[65:174, ]$Sex

gephast <- gephast[which(rowSums(is.na(gephast[, c(1:13)]), dims = 1) == 0), ]

gephast_M <- gephast %>% 
  filter(sex == "M")
gephast_F <- gephast %>% 
  filter(sex == "F")

gephast_M <- gephast_M[which(!is.na(gephast_M$hw)), ]
gephast_F <- gephast_F[which(!is.na(gephast_F$hw)), ]

gephast_M$hw_z <- z_score(gephast_M$hw)
gephast_F$hw_z <- z_score(gephast_F$hw)

gephast_corr <- rbind(gephast_M, gephast_F)

gephast_corr$average_hom <- (13-apply(gephast_corr[, 1:13], 1, sum))/13
gephast_corr$risk <- NA
gephast_corr$risk[gephast_corr$average_hom <= as.numeric(adu_Ho[3])] <- "low"
gephast_corr$risk[gephast_corr$average_hom > as.numeric(adu_Ho[3])] <- "high"

or_chi <- apply(gephast_corr[,1:13], 2, function(x){chisq.test(table(gephast_corr$risk, x))})
or_chi <- unlist(lapply(or_chi, function(x){x$statistic}))


# scramble data nrep times
scrumble_genotypes <- function(genotypes_zero_one,
                               nloci, nreps, loci_names){
  r <- vector(mode = "list", length = nreps)
  for (i in 1:nreps) {
    rand_gen <- as.data.frame(t(apply(genotypes_zero_one, 1, sample,
                                      nloci, replace = F)))
    names(rand_gen) <- loci_names
    r[[i]] <- rand_gen
  }
  return(r)
}

scrumbled <- scrumble_genotypes(gephast_corr[, 1:13], 13, 100000, lnames[1:13])
scrumbled <- lapply(scrumbled,
       function(x){apply(x, 2, function(x){chisq.test(table(gephast_corr$risk, x))})})
scrumbled <- lapply(scrumbled, function(x){unlist(lapply(x, function(x){x$statistic}))})
scrumbled <- as.data.frame(do.call(rbind, scrumbled))

Reference

Amos, W., & Acevedo-Whitehouse, K. (2009). A new test for genotype-fitness associations reveals a single microsatellite allele that strongly predicts the nature of tuberculosis infections in wild boar. Molecular Ecology Resources, 9(4), 1102–1111. Wiley. Retrieved from https://doi.org/10.1111%2Fj.1755-0998.2009.02560.x