Chapter 3 Data Exploration and Normalization
The first step for differential abundance analysis is to explore the data and select the best normalization method, which will be achieved using some custom functions.
3.1 Exploring multiple normalization methods
Data can be normalized in multiple ways, normalization methods presented here include: - Global sum normalization - Median normalization - Mean Normalization - VSN Normalization - Cyclic LOESS Normalization
The custom function normalize_by_all() will apply all the normalizations and will log-transform the data
library(ggpubr)
source('data/functions_da.R')
normalize_by_all(metabolite_abundance)
Figure 3.1: Normalization methods
From the plot the best normalization method seem to be either VSN or cyclic LOESS.