exercise 8

cont.table <- matrix(c(5, 1, 2, 12), nrow = 2, ncol = 2)
fisher.test(cont.table, alternative = "greater")

## 
##  Fisher's Exact Test for Count Data
## 
## data:  cont.table
## p-value = 0.007224
## alternative hypothesis: true odds ratio is greater than 1
## 95 percent confidence interval:
##  2.138   Inf
## sample estimates:
## odds ratio 
##      22.96

load("/Users/tobias/Dropbox/Uni/Master/HT_Course/Module_8_EnrichmentAnalysis/toyFisher.RData")

rownames(toyExpr)

##  [1] "gene_1"  "gene_2"  "gene_3"  "gene_4"  "gene_5"  "gene_6"  "gene_7" 
##  [8] "gene_8"  "gene_9"  "gene_10" "gene_11" "gene_12" "gene_13" "gene_14"
## [15] "gene_15" "gene_16" "gene_17" "gene_18" "gene_19" "gene_20" "gene_21"
## [22] "gene_22" "gene_23" "gene_24" "gene_25" "gene_26" "gene_27" "gene_28"
## [29] "gene_29" "gene_30" "gene_31" "gene_32" "gene_33" "gene_34" "gene_35"
## [36] "gene_36" "gene_37" "gene_38" "gene_39" "gene_40" "gene_41" "gene_42"
## [43] "gene_43" "gene_44" "gene_45" "gene_46" "gene_47" "gene_48" "gene_49"
## [50] "gene_50"

colnames(toyExpr)

##  [1] "normal_1"  "normal_2"  "normal_3"  "normal_4"  "normal_5" 
##  [6] "normal_6"  "normal_7"  "normal_8"  "normal_9"  "normal_10"
## [11] "cancer_1"  "cancer_2"  "cancer_3"  "cancer_4"  "cancer_5" 
## [16] "cancer_6"  "cancer_7"  "cancer_8"  "cancer_9"  "cancer_10"

givesP <- function(x) {
    x$p.value
}
t_Parted <- function(x) {
    t.test(x[1:10], x[11:20], var.equal = T)
}

result <- apply(toyExpr, 1, t_Parted)
x <- lapply(result, givesP)

pvalue <- unlist(x)



fdr_v <- p.adjust(pvalue, method = "fdr")

Goterm 1

table(names(which(fdr_v < 0.01)) %in% toyGOterm$GO_1)

## 
## FALSE  TRUE 
##     1     9

table(names(which(fdr_v > 0.01)) %in% toyGOterm$GO_1)

## 
## FALSE  TRUE 
##    38     2

# for GO_1
ct.go1 <- matrix(c(9, 1, 2, 38), 2, 2)
fisher.test(ct.go1)

## 
##  Fisher's Exact Test for Count Data
## 
## data:  ct.go1
## p-value = 2.099e-07
## alternative hypothesis: true odds ratio is not equal to 1
## 95 percent confidence interval:
##    11.11 7143.45
## sample estimates:
## odds ratio 
##      126.6

Man sollte hier der alternativhypothese folgen

Goterm 2

table(names(which(fdr_v < 0.01)) %in% toyGOterm$GO_2)

## 
## FALSE  TRUE 
##     8     2

table(names(which(fdr_v > 0.01)) %in% toyGOterm$GO_2)

## 
## FALSE  TRUE 
##    29    11

# for GO_2
ct.go2 <- matrix(c(2, 8, 11, 29), 2, 2)
fisher.test(ct.go2)

## 
##  Fisher's Exact Test for Count Data
## 
## data:  ct.go2
## p-value = 1
## alternative hypothesis: true odds ratio is not equal to 1
## 95 percent confidence interval:
##  0.0598 4.1269
## sample estimates:
## odds ratio 
##     0.6643

Es ist zufall -> Go-Term sollte für die verschiedenen Gene abgehlehnt werden