美文网首页pcoa图
R：肠型Enterotypes

R：肠型Enterotypes

作者: 胡童远 | 来源:发表于2020-08-04 21:03 被阅读0次

导读

2011年，肠型(Enterotypes)的概念首次在《自然》杂志上由Arumugam等【1】提出，该研究发现可以将人类肠道微生物组分成稳定的3种类型，因为这3种类型不受年龄、性别、体重以及地域限制，表现出较高的稳定性，与血型具有很高的相似性，所以将其定义为肠型。后来的有的研究称发现2种肠型，也有称发现4种，这引起了国际上对于肠型概念的广泛讨论与争议。最新的《自然∙微生物学》杂志报道了学界对于肠型的最新认识，调解了之前关于肠型的一些争论【2】。

方法

目前流行的肠型计算方法有2种，一种是基于样品间的Jensen-Shannon距离，利用围绕中心点划分算法(PAM)进行聚类，最佳分类数目通过Calinski-Harabasz (CH)指数确定【3】；另一种则是直接基于物种丰度，利用狄利克雷多项混合模型(DMM)进行肠型分类【4】。

参考：人体肠道宏基因组生物信息分析方法 [J].微生物学报, 2018
Nature 2011 肠型分析教程：https://enterotype.embl.de/enterotypes.html#
Reference-Based肠型分析：http://enterotypes.org/
三大肠型ET_B、ET_P、ET_F分别代表Bacteroides（拟杆菌属）, Prevotella（普氏菌属）和Firmicutes（厚壁菌门） 肠型。
Github：https://github.com/tapj/biotyper

一、举例

来自：Quantitative microbiome profiling disentangles inflammation- and bile duct obstruction-associated microbiota alterations across PSC/IBD diagnoses.nature microbiology. 2019
方法：DirichletMultinomial.
描述：Enterotyping (or community typing) based on the DMM approach was performed in R as described previously. Enterotyping was performed on a combined genus-level abundance matrix including PSC/IBD and mHC samples compiled with 1,054 samples originating from the FGFP.

来自：Impact of early events and lifestyle on the gut microbiota and metabolic phenotypes in young school-age children. micorbiome. 2019
方法：DMM, PAM-JSD, PAM-BC
描述：Genus-level enterotypes analysis was performed according to the Dirichlet multinomial mixtures (DMM) and partitioning around medoid (PAM)-based clustering protocols using Jensen-Shannon divergence (PAM-JSD) and Bray-Curtis (PAM-BC)

二、函数：JSD计算距离、PAM聚类

思路：
1 丰度值为0的加上0.000001（1e-6），避免分子分母出现0
2 新建matrix，样品数作行数，样品数作列数
3 双重for循环计算样品间JSD到matrix
4 样品名为行列名，as.dist去重复，attr设置dist属性
5 确定cluster数，用PAM对JSD样品matirx进行聚类

library(ade4)
library(cluster)
library(clusterSim)

## 1. 菌属-样品相对丰度表
data=read.table("input.txt", header=T, row.names=1, dec=".", sep="\t")
# dec：标志小数点符号，有些国家的小数点是逗号
data=data[-1,]

## 写函数
## JSD计算样品距离，PAM聚类样品，CH指数估计聚类数，比较Silhouette系数评估聚类质量。
dist.JSD <- function(inMatrix, pseudocount=0.000001, ...)
{
  ## 函数：JSD计算样品距离
  KLD <- function(x,y) sum(x *log(x/y))
  JSD <- function(x,y) sqrt(0.5 * KLD(x, (x+y)/2) + 0.5 * KLD(y, (x+y)/2))

  matrixColSize <- length(colnames(inMatrix))
  matrixRowSize <- length(rownames(inMatrix))
  colnames <- colnames(inMatrix)
  resultsMatrix <- matrix(0, matrixColSize, matrixColSize)
  
  inMatrix = apply(inMatrix, 1:2, function(x) ifelse (x==0, pseudocount, x))
  
  for(i in 1:matrixColSize)
  {
    for(j in 1:matrixColSize)
    { 
      resultsMatrix[i, j] = JSD(as.vector(inMatrix[, i]), as.vector(inMatrix[, j]))
    }
  }
      
  colnames -> colnames(resultsMatrix) -> rownames(resultsMatrix)
  as.dist(resultsMatrix) -> resultsMatrix
  # 两两重复矩阵去重，as.dist
  attr(resultsMatrix, "method") <- "dist"
  return(resultsMatrix)
}

pam.clustering = function(x, k)
{ 
  ## 函数：PAM聚类样品
  # x is a distance matrix and k the number of clusters
  require(cluster)
  cluster = as.vector(pam(as.dist(x), k, diss=TRUE)$clustering)
  return(cluster)
}

三、计算CH确定最佳Cluster

思路：
1 CH指数作为评估cluster数好坏的指标
2 尝试不同的cluster数，挨个进行PAM聚类，用index.G1计算CH指数
3 CH指数绘图观察最佳cluster数，which(, arr.ind=T)通过获取最大CH数的位置得到该数

## 2. 选择CH指数最大的K值作为最佳聚类数
data.dist = dist.JSD(data)
                   
# data.cluster=pam.clustering(data.dist, k=3) # k=3 为例 
# nclusters = index.G1(t(data), data.cluster, d = data.dist, centrotypes = "medoids") # 查看CH指数

nclusters = NULL

for(k in 1:20)
{ 
  if(k==1)
  {
    nclusters[k] = NA 
  }
  else
  {
    data.cluster_temp = pam.clustering(data.dist, k)
    nclusters[k] = index.G1(t(data), data.cluster_temp,  d = data.dist, centrotypes = "medoids")
  }
}

plot(nclusters, type="h", xlab="k clusters", ylab="CH index", main="Optimal number of clusters") # 查看K与CH值得关系

如图，CH最大时的cluster数为3，最佳。

nclusters[1] = 0
k_best = which(nclusters == max(nclusters), arr.ind = TRUE)

三、对JSD矩阵进行正式PAM聚类，silhouette评估聚类质量

## 3. PAM根据JSD距离对样品聚类(分成K个组)
data.cluster=pam.clustering(data.dist, k = k_best)

## 4. silhouette评估聚类质量，-1=<S(i)=<1，越接近1越好
mean(silhouette(data.cluster, data.dist)[, k_best])

四、可视化

between-class analysis (BCA) 和principal coordinates analysis (PCoA)是两种常用的肠型可视化方法。PCOA更常用生态学研究，推荐使用PCOA对JSD matrix进行可视化。

1 基于BCA

## plot 1
obs.pca=dudi.pca(data.frame(t(data)), scannf=F, nf=10)
obs.bet=bca(obs.pca, fac=as.factor(data.cluster), scannf=F, nf=k-1) 

s.class(obs.bet$ls, fac=as.factor(data.cluster), grid=F,sub="Between-class analysis", col=c(1,2,3))

2 基于PCoA

#plot 2
obs.pcoa = dudi.pco(data.dist, scannf = F, nf = 3)

s.class(obs.pcoa$li, fac = as.factor(data.cluster), grid = F, sub = "Principal coordiante analysis", col=c(1,2,3))

五、一个案例

分析思路：
1 利用菌属样品表对样品进行JSD-PAM聚类
2 PCOA得到样品坐标，envfit多元回归分析细菌对聚类的贡献度
3 挑选细菌作可视化

1 输入数据
菌属丰度表


2 删除存在重复的菌属 
# 删除存在重复的菌属
genus = data.frame(table(input[,1]))
delete = as.character(genus[genus$Freq != 1,]$Var1)
input2 = input[!(input$Taxonomy%in%delete),]
rownames(input2) = input2$Taxonomy
input2 = input2[, -1]

3 排序选择，这里没选

# 选用丰度前100的菌属
input2$sum = apply(input2, 1, sum)
input2 = input2[order(input2$sum, decreasing=T),]
input2 = input2[, -ncol(input2)]

4 获取样品举例矩阵

inMatrix = input2
#dist.JSD <- function(inMatrix, pseudocount=0.000001, ...)
#{
  ## 函数：JSD计算样品距离
  pseudocount=0.000001
  KLD <- function(x,y) sum(x *log(x/y))
  JSD <- function(x,y) sqrt(0.5 * KLD(x, (x+y)/2) + 0.5 * KLD(y, (x+y)/2))

  matrixColSize <- length(colnames(inMatrix))
  matrixRowSize <- length(rownames(inMatrix))
  colnames <- colnames(inMatrix)
  resultsMatrix <- matrix(0, matrixColSize, matrixColSize)
  
  inMatrix = apply(inMatrix, 1:2, function(x) ifelse (x==0, pseudocount, x))
  
  for(i in 1:matrixColSize)
  {
    for(j in 1:matrixColSize)
    { 
      resultsMatrix[i, j] = JSD(as.vector(inMatrix[, i]), as.vector(inMatrix[, j]))
    }
  }
      
  colnames -> colnames(resultsMatrix) -> rownames(resultsMatrix)
  as.dist(resultsMatrix) -> resultsMatrix
  # 两两重复矩阵去重，as.dist
  attr(resultsMatrix, "method") <- "dist"
  # class, comment, dim, dimnames, names, row.names and tsp
  #return(resultsMatrix)
#}
data.dist = resultsMatrix

5 聚类，获取最佳聚类数

pam.clustering = function(x, k)
{ 
  ## 函数：PAM聚类样品
  # x is a distance matrix and k the number of clusters
  require(cluster)
  cluster = as.vector(pam(as.dist(x), k, diss=TRUE)$clustering)
  return(cluster)
}
                   
#data.dist = dist.JSD(inMatrix)
                  
# data.cluster = pam.clustering(data.dist, k=3) # k=3 为例 
# nclusters = index.G1(t(input2), data.cluster, d = data.dist, centrotypes = "medoids")
# 查看CH指数
nclusters = NULL

for(k in 1:10)
{ 
  if(k==1)
  {
    nclusters[k] = NA
  }
  else
  {
    data.cluster_temp = pam.clustering(data.dist, k)
    nclusters[k] = index.G1(t(inMatrix), data.cluster_temp,  d = data.dist, centrotypes = "medoids")
  }
}

plot(nclusters, type="h", xlab="k clusters", ylab="CH index", main="Optimal number of clusters")  
                   
## 3. PAM根据JSD距离对样品聚类(分成K个组)
nclusters[1] = 0
k_best = which(nclusters == max(nclusters), arr.ind = TRUE)
data.cluster = pam.clustering(data.dist, k = k_best)

## 4. silhouette评估聚类质量，-1=<S(i)=<1，越接近1越好
silhouette(data.cluster, data.dist)[, k_best]      
mean(silhouette(data.cluster, data.dist)[, k_best])

6 PCOA获取坐标，envfit计算细菌相关度，

obs.pcoa = dudi.pco(data.dist, scannf=F, nf=2)

data = inMatrix
env = data.frame(t(data))

ord = obs.pcoa$li
fit = envfit(ord, env, perm=999)
p = data.frame(fit$vectors$pvals)
p$label = rownames(p)
p = p[order(p$fit.vectors.pvals, decreasing=F),]
#head(fit$vectors)
#fit$arrows
#fit$r
#fit$pvals
#sign_site = which(p[,1]>0.001, arr.ind=T)[1] - 1

7 根据p值选择细菌（方案一）

sign_site = which(p[,1]>0.001, arr.ind=T)[1] - 1
env = data.frame(t(inMatrix[rownames(inMatrix)%in%p$label[1:sign_site],]))
fit = envfit(ord, env, perm=999)

tmp = data.frame(fit$vectors$arrows)
color=c()
for(i in 1:nrow(tmp))
{
    if(tmp[i,1] < 0 ) color=c(color, "red")
    else if(tmp[i,1] > 0) color=c(color, "blue")
}

genus = data.frame(head(fit$vectors)$arrows)
#lab = rbind(ord, genus)
sample = ord
for(i in 1:nrow(ord))
{
    if(ord[i,2] < 0) sample[i,2] = ord[i,2] - 0.01
        else if(ord[i,2] > 0) sample[i,2] = ord[i,2] + 0.01 
            else if(ord[i,2] == 0) sample[i,2] = ord[i,2] + 0.01
}

8 可视化

pdf("test1.pdf", width=25, height=21)

par(family = "serif")
#s.class(ord, fac=as.factor(data.cluster), grid=F,sub="Principal coordiante analysis", csub=3, pch=21, col=c("red","blue"), cpoint=1, clabel=2, cstar=1)
s.class(ord, fac=as.factor(data.cluster), grid=F, sub="Principal coordiante analysis", csub=3, pch=21, cpoint=2, clabel=2, cstar=1, col=c("Firebrick3", "DeepSkyBlue2"))
#plot(fit, font=3, col=color, cex=3)
text(x = sample$A1, y = sample$A2, labels = rownames(sample), cex=2)
text(x = genus[,1]/4, y = genus[,2]/4, labels = rownames(genus), cex=5, col="black", font=4)
#text(x = lab$A1, y = lab$A2, labels = rownames(lab), cex=2)
dev.off()

9 筛选p值前六细菌（方案二）

# 二、选择靶标细菌（前六）
env = data.frame(t(inMatrix[rownames(inMatrix)%in%p$label[c(1:6)],]))
fit = envfit(ord, env, perm=999)

# 箭头配色
# fit$vectors$arrows
tmp = data.frame(fit$vectors$arrows)
color=c()
for(i in 1:nrow(tmp))
{
    if(tmp[i,1] < 0 ) color=c(color, "red")
    else if(tmp[i,1] > 0) color=c(color, "blue")
}

# 画图保存
pdf("test-arrow.pdf", width=25, height=21)

par(family = "serif")
#s.class(ord, fac=as.factor(data.cluster), grid=F,sub="Principal coordiante analysis", csub=3, pch=21, col=c("red","blue"), cpoint=1, clabel=2, cstar=1)
s.class(ord, fac=as.factor(data.cluster), grid=F, sub="Principal coordiante analysis", csub=3, pch=21, cpoint=2, clabel=2, cstar=1, col=c("Firebrick3", "DeepSkyBlue2"))
plot(fit, font=3, col=color, cex=3)

dev.off()

10 选择靶标细菌（方案三）

# 二、选择靶标细菌
env = data.frame(t(inMatrix[rownames(inMatrix)%in%p$label[c(1,2,5)],]))
fit = envfit(ord, env, perm=999)

# 给每个样本点添加label，防止覆盖 + 0.01
genus = data.frame(head(fit$vectors)$arrows)
lab = rbind(ord, genus)
sample = ord
for(i in 1:nrow(ord))
{
    if(ord[i,2] < 0) sample[i,2] = ord[i,2] - 0.01
        else if(ord[i,2] > 0) sample[i,2] = ord[i,2] + 0.01 
            else if(ord[i,2] == 0) sample[i,2] = ord[i,2] + 0.01
}
# https://cran.r-project.org/web/packages/ade4/ade4.pdf
pdf("test1.pdf", width=25, height=21)

par(family = "serif")
#s.class(ord, fac=as.factor(data.cluster), grid=F,sub="Principal coordiante analysis", csub=3, pch=21, col=c("red","blue"), cpoint=1, clabel=2, cstar=1)
s.class(ord, fac=as.factor(data.cluster), grid=F, sub="Principal coordiante analysis", csub=3, pch=21, cpoint=2, clabel=2, cstar=1, col=c("Firebrick3", "DeepSkyBlue2"))
#plot(fit, font=3, col=color, cex=3)
text(x = sample$A1, y = sample$A2, labels = rownames(sample), cex=2)
text(x = genus[,1]/4, y = genus[,2]/4, labels = rownames(genus), cex=5, col="black", font=4)
#text(x = lab$A1, y = lab$A2, labels = rownames(lab), cex=2)
dev.off()
图片.png

参考：
【1】Enterotypes of the human gut microbiome. Nature, 2011
【2】人体肠道宏基因组生物信息分析方法 [J].微生物学报, 2018
【3】Linking long-term dietary patterns with gut microbial enterotypes. Science, 2011
【4】Dirichlet multinomial mixtures: generative models for microbial metagenomics. PLoS One, 2012
【5】Statin drugs might boost healthy gut microbes. Nature News. 2020

如何鉴定肠型
排序图（beta多样性）添加各种辅助线-无非是让不明显的规律看得到
ggplot2添加散点图文字标记
https://cran.r-project.org/web/packages/ade4/ade4.pdf

2020.10.26更新

相关文章

网友评论

    本文标题：R：肠型Enterotypes

    本文链接：https://www.haomeiwen.com/subject/qzzhrktx.html

    延伸阅读

    深度阅读

    • 您也可以注册成为美文阅读网的作者，发表您的原创作品、分享您的心情！

    栏目导航

    热点阅读

    pcoa图

    关于我们|服务条款|联系我们|R：肠型Enterotypes|投稿指南|网站地图|RSS订阅|排版工具|手机版

    提供经典美文摘抄,优美散文欣赏,现代诗歌精选,短篇小说,心情随笔,表白情书范文,故事会在线阅读欣赏

    Copyright © 2014-2023 Haomeiwen.com All Rights Reserved. 好美文阅读网 版权所有

    备案信息：桂公网安备 45052102000051号 · 桂ICP备13007215号-3

    本站所收录作品、热点评论等信息部分来源互联网，目的只是为了系统归纳学习和传递资讯

    所有作品版权归原创作者所有，与本站立场无关，如不慎侵犯了你的权益，请联系我们告知，我们将做删除处理！