Class 12 - Population Scale Analysis

Karolina Navarro (PID: A19106745)

##Section 4: Population Scale Analysis

One sample is obviously not enough to know what is happening in a population. You are interested in assessing genetic differences on a population scale. So, you processed about ~230 samples and did the normalization on a genome level. Now, you want to find whether there is any association of the 4 asthma-associated SNPs (rs8067378…) on ORMDL3 expression.

How many samples do we have?

expr <- read.table("rs8067378_ENSG00000172057.6.txt",
                   header = TRUE,
                   stringsAsFactors = FALSE)
head(expr)

   sample geno      exp
1 HG00367  A/G 28.96038
2 NA20768  A/G 20.24449
3 HG00361  A/A 31.32628
4 HG00135  A/A 34.11169
5 NA18870  G/G 18.25141
6 NA11993  A/A 32.89721

nrow(expr)

[1] 462

table(expr$geno)

A/A A/G G/G 
108 233 121 

library(dplyr)

Attaching package: 'dplyr'

The following objects are masked from 'package:stats':

    filter, lag

The following objects are masked from 'package:base':

    intersect, setdiff, setequal, union

expr %>%
  group_by(geno) %>%
  summarise(
    n = n(),
    median_expression = median(exp, na.rm = TRUE)
  )

# A tibble: 3 × 3
  geno      n median_expression
  <chr> <int>             <dbl>
1 A/A     108              31.2
2 A/G     233              25.1
3 G/G     121              20.1

library(ggplot2)

Let’s make a boxplot

ggplot(expr) + aes(geno, exp, fill=geno) +
  geom_boxplot(notch=TRUE) +
  theme_bw()

Given the relative expression values provided, I can infer that A/A has the highest median expression and G/G has the lowest median expression value. This hints at the idea that there is a clear stepwise decrease in gene expression as the G allele increases in dosage. Based on this plot, the SNP appears to influence expression levels ORMDL3.