【TCGA数据挖掘】R/mclust做Model-based clustering分析整合临床分类

在常见的临床生信文章中，需要进行Model-based clustering来将临床数据与分子数据结合，以下就是使用mclust实现的过程

加载所需要的包

library(mclust)
library(Biobase)
library(ConsensusClusterPlus)
library(BS831)
library(cba)
library(pheatmap)

对数据进行聚类

MC <- Mclust(t(Biobase::exprs(cancerSet)),G=1:4)
summary(MC)

## ---------------------------------------------------- 
## Gaussian finite mixture model fitted by EM algorithm 
## ---------------------------------------------------- 
## 
## Mclust VEI (diagonal, equal shape) model with 3 components: 
## 
##  log-likelihood  n    df       BIC       ICL
##       -200111.4 45 16004 -461144.6 -461144.6
## 
## Clustering table:
##  1  2  3 
## 18 20  7

由以上可以看到分成了三类，那么这三类在临床样本的分布是怎样的呢？

## show cluster assignments (and disease state)
head(data.frame(MC$classification,pData(cancerSet)$Characteristics.DiseaseState))

##          MC.classification pData.cancerSet..Characteristics.DiseaseState
## GSM85480                 1                           sporadic basal-like
## GSM85490                 1                           sporadic basal-like
## GSM85500                 2                                non-basal-like
## GSM85484                 1                           sporadic basal-like
## GSM85478                 1                           sporadic basal-like
## GSM85516                 3                                        normal

以上我们只是对样本查看实际上还是要对整个的分布，比如1，2，3聚类在临床表型的差别进行统计。

## contingency table
print(ftable(MC$classification,pData(cancerSet)$Characteristics.DiseaseState))

##    non-basal-like normal sporadic basal-like
##                                             
## 1               0      0                  18
## 2              20      0                   0
## 3               0      7                   0

得到这些数据就可以用pheatmap,或者complexheatmap进行绘制图片了，大致如下：