Creates a plot showing the time series expression of genes within clusters, overlain by the w/in cluster mean

mean_time_series_per_cluster.R

library(dendextend)
library(tidyr)
library(dplyr)
library(magrittr)
library(ggplot2)

# load data in "wide" format (genes in columns)
spellman <- read.csv("spellman-reformated.csv")

# restructure in "long" format
spellman.long <- gather(spellman, gene, expression, -expt, -time)

# calculate corr based on alpha factor expt measures
alpha.cor <- 
  spellman %>%
  filter(expt == "alpha") %>%
  select(-expt, -time) %>%
  cor(use="pairwise.complete.obs")

# cluster the alpha factor data using hclust
alpha.clust <-
  as.dist(1 - alpha.cor) %>%
  hclust(method="complete")

# cut tree, to yield 8 clusters
clusters <- dendextend::cutree(alpha.clust, k=8, 
                               order_clusters_as_data = TRUE)

# create a data frame holding cluster membership for each gene
clusters.df <- data.frame(gene = names(clusters), 
                          cluster = as.factor(clusters))

# do a left_join, combining the information in spellman.long
# with clusters.df (matched on gene).  This effectively adds the
# cluster information as a new column to spellman.long data frame
# keeping the appropriate matches by gene name
alpha.long <- 
  spellman.long %>%
  filter(expt == "alpha") %>%
  left_join(clusters.df, by = c("gene"))

# calculate the mean at each time point within each cluster
cluster.means <-
  alpha.long %>%
  group_by(cluster, time) %>%
  summarize(mean.exp = mean(expression, na.rm = TRUE))

# draw a figure showing time varying gene expression
# in each cluster, overlain with the each clusters 
# mean time series
alpha.long %>%
  ggplot(aes(time, expression, group=gene)) +
  geom_line(alpha=0.25) + 
  geom_line(aes(time, mean.exp, group=NULL,color=cluster),
            data = cluster.means,
            size=1.1) +
  facet_wrap(~cluster, ncol=4)