放弃Venn-Upset-花瓣图，拥抱二分网络

生物信息学习的正确姿势

NGS系列文章包括NGS基础、在线绘图、转录组分析 （Nature重磅综述|关于RNA-seq你想知道的全在这）、ChIP-seq分析 （ChIP-seq基本分析流程）、单细胞测序分析 (重磅综述：三万字长文读懂单细胞RNA测序分析的最佳实践教程)、DNA甲基化分析、重测序分析、GEO数据挖掘（典型医学设计实验GEO数据分析 (step-by-step)）、批次效应处理等内容。

写在前面

让点随机排布在一个区域，保证点之间不重叠,并且将点的图层放到最上层，保证节点最清晰，然后边可以进行透明化，更加突出节点的位置。这里我新构建了布局函数 PolyRdmNotdCirG 来做这个随机排布。调用的是packcircles包的算法。使用和其他相似函数一样，这里我们重点介绍一下使用这种算法构造的二分网络布局。

微生物网络

ggClusterNet 安装

ggClusterNet包依赖的R包均在cran或者biocductor中，所以未能成功安装，需要检查依赖是否都顺利安装。如果网路问题，无法下载R包，可以在github中手动下载安装：具体安装方法参考：玩转R包

#---ggClusterNet devtools::install_github("taowenmicro/ggClusterNet") #--如果无法安装请检查网络或者换个时间

导入R包和输入文件

#--导入所需R包#------- library(ggplot2) library(ggrepel) library(ggClusterNet) library(phyloseq) library(dplyr)# 数据内置 #-----导入数据#------- data(ps)#--可选 #-----导入数据#------- ps = readRDS("../ori_data/ps_liu.rds")

这里我们提取一部分OTU，节省出图时间。

# ps data(ps)ps_sub = filter_taxa(ps, function(x) sum(x ) > 20 , TRUE) ps_sub = filter_taxa(ps_sub, function(x) sum(x ) < 30 , TRUE) ps_sub

div_network函数 用于计算共有和特有关系

这个函数是之前我写的专门用于从OTU表格整理成Gephi的输入文件，所以大家直接用这个函数即可转到gephi进行操作。这次为了配合二分网络，我设置了参数flour = TRUE，代表是否仅仅提取共有部分和特有部分。

# ?div_network result = div_network(ps_sub,num = 6)edge = result[[1]] head(edge)# levels(edge$target) # node = result[[2]] # head(node) # # tail(node) data = result[[3]] dim(data)#----计算节点坐标 # flour参数，设置是否仅仅展示共有和特有的二分网络

div_culculate函数 核心算法，用于计算二分网络的节点和边的表格

参数解释：

• distance = 1.1：

  中心一团点到样本点距离

• distance2 = 1.5：

  中心点模块到独有OTU点之间距离

• distance3 = 1.3：

  样本点和独有OTU之间的距离

• order = FALSE ：

  节点是否需要随机扰动效果

result <- div_culculate(table = result[[3]],distance = 1.1,distance2 = 1.5,distance3 = 1.3,order = FALSE)edge = result[[1]] head(edge)plotdata = result[[2]] head(plotdata) #--这部分数据是样本点数据 groupdata <- result[[3]]

对OTU进行注释，方便添加到图形上

为了让节点更加丰富，这里我对节点文件添加了注释信息。

# table(plotdata$elements) node = plotdata[plotdata$elements == unique(plotdata$elements), ]otu_table = as.data.frame(t(vegan_otu(ps_sub))) tax_table = as.data.frame(vegan_tax(ps_sub)) res = merge(node,tax_table,by = "row.names",all = F) dim(res) head(res) row.names(res) = res$Row.names res$Row.names = NULL plotcord = resxx = data.frame(mean =rowMeans(otu_table)) head(xx) plotcord = merge(plotcord,xx,by = "row.names",all = FALSE) head(plotcord) # plotcord$Phylum row.names(plotcord) = plotcord$Row.names plotcord$Row.names = NULL head(plotcord)p = ggplot() + geom_segment(aes(x = X1, y = Y1, xend = X2, yend = Y2),data = edge, size = 0.3,color = "yellow") +geom_point(aes(X1, X2,fill = Phylum,size =mean ),pch = 21, data = plotcord) +geom_point(aes(X1, X2),pch = 21, data = groupdata,size = 5,fill = "blue",color = "black") +geom_text_repel(aes(X1, X2,label = elements ), data = groupdata) +theme_void()pggsave("4.png",p,width = 12,height = 8)

map = as.data.frame(sample_data(ps_sub))map$Group2 <- rep(c("A1","A2","A3","A4","A5","A6"),3)sample_data(ps_sub) <- map# ?div_network result = div_network(ps_sub,num = 3,group = "Group2",flour = TRUE)edge = result[[1]] head(edge)# levels(edge$target) # node = result[[2]] # head(node) # # tail(node)data = result[[3]] dim(data)#----计算节点坐标 # flour参数，设置是否仅仅展示共有和特有的二分网络result <- div_culculate(table = result[[3]],distance = 1.1,distance2 = 1.5,distance3 = 1.3,order = FALSE)edge = result[[1]] head(edge)plotdata = result[[2]] head(plotdata)groupdata <- result[[3]]# table(plotdata$elements) node = plotdata[plotdata$elements == unique(plotdata$elements), ]otu_table = as.data.frame(t(vegan_otu(ps_sub))) tax_table = as.data.frame(vegan_tax(ps_sub)) res = merge(node,tax_table,by = "row.names",all = F) dim(res) head(res) row.names(res) = res$Row.names res$Row.names = NULL plotcord = resxx = data.frame(mean =rowMeans(otu_table)) head(xx) plotcord = merge(plotcord,xx,by = "row.names",all = FALSE) head(plotcord) # plotcord$Phylum row.names(plotcord) = plotcord$Row.names plotcord$Row.names = NULL head(plotcord)p = ggplot() + geom_segment(aes(x = X1, y = Y1, xend = X2, yend = Y2),data = edge, size = 0.3,color = "yellow") +geom_point(aes(X1, X2,fill = Phylum,size =mean ),pch = 21, data = plotcord) +geom_point(aes(X1, X2),pch = 21, data = groupdata,size = 5,fill = "blue",color = "black") +geom_text_repel(aes(X1, X2,label = elements ), data = groupdata) +theme_void() p ggsave("4.png",p,width = 12,height = 8)

map = as.data.frame(sample_data(ps_sub))map = map[1:12,]# map$Group2 <- rep(c("A1","A2","A3","A4","A5","A6"),2) sample_data(ps_sub) <- mapresult = div_network(ps_sub,num = 3,group = "Group",flour = TRUE)edge = result[[1]] head(edge)# levels(edge$target) # node = result[[2]] # head(node) # # tail(node)data = result[[3]] dim(data)result <- div_culculate(table = result[[3]],distance = 1.1,distance2 = 1.5,distance3 = 1.3,order = FALSE)edge = result[[1]] head(edge)plotdata = result[[2]] head(plotdata)groupdata <- result[[3]]# table(plotdata$elements) node = plotdata[plotdata$elements == unique(plotdata$elements), ]otu_table = as.data.frame(t(vegan_otu(ps_sub))) tax_table = as.data.frame(vegan_tax(ps_sub)) res = merge(node,tax_table,by = "row.names",all = F) dim(res) head(res) row.names(res) = res$Row.names res$Row.names = NULL plotcord = resxx = data.frame(mean =rowMeans(otu_table)) head(xx) plotcord = merge(plotcord,xx,by = "row.names",all = FALSE) head(plotcord) # plotcord$Phylum row.names(plotcord) = plotcord$Row.names plotcord$Row.names = NULL head(plotcord)p = ggplot() + geom_segment(aes(x = X1, y = Y1, xend = X2, yend = Y2),data = edge, size = 0.3,color = "yellow") +geom_point(aes(X1, X2,fill = Phylum,size =mean ),pch = 21, data = plotcord) +geom_point(aes(X1, X2),pch = 21, data = groupdata,size = 5,fill = "blue",color = "black") +geom_text_repel(aes(X1, X2,label = elements ), data = groupdata) +theme_void()p# ggsave("4.png",p,width = 12,height = 22)

在线绘制 Venn 图和二分网络图，点击阅读原文或扫描二维码访问吧！

往期精品(点击图片直达文字对应教程)

后台回复“生信宝典福利第一波”或点击阅读原文获取教程合集

总结

以上是生活随笔为你收集整理的放弃Venn-Upset-花瓣图，拥抱二分网络的全部内容，希望文章能够帮你解决所遇到的问题。

如果觉得生活随笔网站内容还不错，欢迎将生活随笔推荐给好友。