Recently I came across a post where user wants to average over triplicates for each time point. I wish there were easy solutions in R. For eg. group and sum by rows or group and sum by columns without common column name. For eg triplicates will have names R1, R2 and R3. There is no way to group them esp when one has multiple triplicates. Here are three solutions for such a scenario.

Create a data frame with two triplicates, for four genes

df1 = data.frame(
genes = paste("gene", seq(1, 4), sep = "_"),
X_T0_R1 = seq(1, 4),
X_T0_R2 = seq(1, 4),
X_T0_R3 = seq(1, 4),
X_T1_R1 = seq(3, 6),
X_T1_R2 = seq(3, 6),
X_T1_R3 = seq(3, 6)
)

df1
genes X_T0_R1 X_T0_R2 X_T0_R3 X_T1_R1 X_T1_R2 X_T1_R3
gene_1    1    1    1    3    3   
gene_2    2    2    2    4    4   
gene_3    3    3    3    5    5   
gene_4    4    4    4    6    6 

User wants to average (mean) triplicates for each gene for each time point (in this case: X_T0 and X_T1). Triplicates are represented by R1,R2 and R3.

There are 3 methods in R to do this:

Method 1: Using tidyverse

> suppressPackageStartupMessages(library(tidyr))
> suppressPackageStartupMessages(library(stringr))
> gdf1 = gather(df1, "group", "Expression", -genes)
> gdf1$tgroup = apply(str_split_fixed(gdf1$group, "_", 3)[, c(1, 2)], 1, paste, collapse ="_")
> suppressPackageStartupMessages(library(dplyr))
> final_df=gdf1 %>% group_by(genes, tgroup) %>% summarize(expression_mean = mean(Expression)) %>% spread(., tgroup, expression_mean)
> final_df

Method 2: Using apply and arrays: Note that headers are not stored

Method 3: Using a loop, apply