2

Objective: Create a dendrogram with branches colored by a factor variable with the resultant plot containing a legend to translate the branch colors to each factor variable value.

I have some data with factor variables followed by the numeric data that I am creating a dendrogram with:

> cleaned_mayo[1:5,1:20]
          patient                Source         Tissue RIN Diagnosis Gender  AgeAtDeath ApoE   FLOWCELL PMI N_unmapped N_multimapping N_noFeature N_ambiguous ENSG00000223972
1924_TCX 1924_TCX MayoBrainBank_Dickson TemporalCortex 5.6   Control      F 90_or_above   33 AC5R6PACXX   2    2773880        9656114     8225967     2876479               1
1926_TCX 1926_TCX MayoBrainBank_Dickson TemporalCortex 7.8   Control      F          88   33 AC44HKACXX   2    2279283       12410116     9503353     3600252               2
1935_TCX 1935_TCX MayoBrainBank_Dickson TemporalCortex 8.6   Control      F          88   33 AC5T2GACXX   3    3120169        8650081     9640468     4603751               0
1925_TCX 1925_TCX MayoBrainBank_Dickson TemporalCortex 6.6   Control      F          89   33 BC6178ACXX   4    2046886       10627577     7533671     3361385               1
1963_TCX 1963_TCX MayoBrainBank_Dickson TemporalCortex 9.7   Control      M 90_or_above   33 AC5T1WACXX   4    1810116        9611375     5343437     2983079               2
         ENSG00000227232 ENSG00000278267 ENSG00000243485 ENSG00000274890 ENSG00000237613
1924_TCX              80               7               1               0               0
1926_TCX             113              22               9               0               0
1935_TCX             181              21               2               0               0
1925_TCX              75               9               5               0               0
1963_TCX              73              14               1               0               0

The data dimensions are: 161 x 60,739. With this data I have achieved a dendrogram with colored branches but no legend, and a dendrogram with colored labels (not branches) with a legend. I would like to combine the two.

Create dendrogram with colored branches but no legend:

# Create the dendrogram for visualization
dend_expr<- cleaned_mayo[,15:60739] %>% # Isolate expression data
                  scale %>% # Normalize
                  dist  %>% # Compute distance measure
                  hclust %>% # Cluster hierarchically
                  as.dendrogram()

# Arrange labels in order with tree
tree_labels<- cleaned_mayo[order.dendrogram(dend_expr),]

# Color branches by diagnosis
dend_expr<- assign_values_to_leaves_edgePar(dend_expr, value= tree_labels$Diagnosis, edgePar= "col") %>%
            as.ggdend()

# Plot dendrogram
ggplot(dend_expr, horiz= T, theme= NULL, labels= F) +
  ggtitle("Mayo Cohort: Hierarchical Clustering of Patients Colored by Diagnosis")

Create dendrogram with colored labels (not branches) and a legend:

# Create the dendrogram for visualization
dend_expr<- cleaned_mayo[,15:60739] %>% # Isolate expression data
                  scale %>% # Normalize
                  dist  %>% # Compute distance measure
                  hclust %>% # Cluster hierarchically
                  as.dendrogram()

tree_labels<- dendro_data(dend_expr, type = "rectangle")
tree_labels$labels<- merge(x= tree_labels$labels, y= cleaned_mayo, by.x= "label", by.y= "patient")

ggplot() +
  geom_segment(data=segment(tree_labels), aes(x=x, y=y, xend=xend, yend=yend)) +
  geom_text(data = label(tree_labels), aes(x=x, y=y, label=label, colour = Diagnosis, hjust=0), size=3) +
  #geom_point(data = label(tree_labels), aes(x=x, y=y), size=2, shape = 21) +
  coord_flip() +
  scale_y_reverse(expand=c(0.2, 0)) +
  scale_colour_brewer(palette = "Dark2") + 
  theme_dendro() +
  ggtitle("Mayo Cohort: Hierarchical Clustering of Patients Colored by Diagnosis")

Examples of the respective outputs: Colored branches; Colored labels with legend.

Any help is appreciated. Thanks!

4

1 回答 1

1

以下是有关如何实现所需着色的示例：

library(tidyverse)
library(ggdendro)
library(dendextend)

一些数据：

matrix(rnorm(1000), ncol = 10) %>%
  scale %>% 
  dist  %>% 
  hclust %>% 
  as.dendrogram() -> dend_expr

tree_labels<- dendro_data(dend_expr, type = "rectangle")
tree_labels$labels <- cbind(tree_labels$labels, Diagnosis = as.factor(sample(1:2, 100, replace = T)))

阴谋：

ggplot() +
  geom_segment(data = segment(tree_labels), aes(x=x, y=y, xend=xend, yend=yend))+
  geom_segment(data = tree_labels$segments %>%
                 filter(yend == 0) %>%
                 left_join(tree_labels$labels, by = "x"), aes(x=x, y=y.x, xend=xend, yend=yend, color = Diagnosis)) +
  geom_text(data = label(tree_labels), aes(x=x, y=y, label=label, colour = Diagnosis, hjust=0), size=3) +
  coord_flip() +
  scale_y_reverse(expand=c(0.2, 0)) +
  scale_colour_brewer(palette = "Dark2") + 
  theme_dendro() +
  ggtitle("Mayo Cohort: Hierarchical Clustering of Patients Colored by Diagnosis")

在此处输入图像描述

关键在我做的第二个 geom_segment 调用中：

tree_labels$segments %>%
     filter(yend == 0) %>%
     left_join(tree_labels$labels, by = "x")

过滤所有叶子并通过yend == 0左连接tree_labels$labelsx

于 2017-12-04T19:52:26.343 回答