【原】跟着iMeta学做图｜NMDS分析展示群落beta多样性

宏基因组 2022-11-21 发布于北京

展开全文

原始教程链接：https://github.com/iMetaScience/iMetaPlot/tree/main/221108NMDS

写在前面

非度量多维尺度分析(Non-metric multidimensional scaling, NMDS)，是基于相异矩阵或距离矩阵进行排序分析的间接梯度分析方法，在微生物组研究中可以用来展示群落beta多样性。本期我们挑选2022年2月24日刊登在iMeta上的Linking soil fungi to bacterial community assembly in arid ecosystems - iMeta：西农韦革宏团队焦硕等-土壤真菌驱动细菌群落的构建，选择文章的Figure 6C进行复现，基于vegan包，讲解和探讨和NMDS分析和可视化的方法，先上原图：

接下来，我们将通过详尽的代码逐步拆解原图，最终实现对原图的复现。代码编写及注释：农心生信工作室。

R包检测和安装

安装核心R包vegan以及ggplot2，并载入所有R包。

if (!require("vegan"))  install.packages('vegan')if (!require("ggplot2"))  install.packages('ggplot2') # 加载包library(vegan)library(ggplot2)

生成测试数据

由于缺少原始数据，因此本例使用vegan包自带的dune数据集进行测试。dune包含了20个样品，每个样品有30个物种丰度，每一行是一个样品，每一列是一个物种。

# 载入dune数据集data(dune)#载入dune包含分组信息等的元数据（即metadata），分组信息为Management列data(dune.env)

NMDS分析

获取数据后，即可利用vegan包进行NMDS分析。

#计算bray_curtis距离distance <- vegdist(dune, method = 'bray')#NMDS排序分析，k = 2预设两个排序轴nmds <- metaMDS(distance, k = 2)#> Run 0 stress 0.1192678 #> Run 1 stress 0.1192678 #> ... Procrustes: rmse 4.495733e-05  max resid 0.0001375161 #> ... Similar to previous best#> Run 2 stress 0.1183186 #> ... New best solution#> ... Procrustes: rmse 0.02026799  max resid 0.06495211 #> Run 3 stress 0.1183186 #> ... New best solution#> ... Procrustes: rmse 1.832694e-05  max resid 5.57604e-05 #> ... Similar to previous best#> Run 4 stress 0.1809577 #> Run 5 stress 0.1192678 #> Run 6 stress 0.1183186 #> ... New best solution#> ... Procrustes: rmse 5.582524e-06  max resid 1.803473e-05 #> ... Similar to previous best#> Run 7 stress 0.1192678 #> Run 8 stress 0.1192678 #> Run 9 stress 0.1192678 #> Run 10 stress 0.1192678 #> Run 11 stress 0.1192679 #> Run 12 stress 0.1808911 #> Run 13 stress 0.1192678 #> Run 14 stress 0.1183186 #> ... Procrustes: rmse 5.943311e-06  max resid 1.823899e-05 #> ... Similar to previous best#> Run 15 stress 0.1886532 #> Run 16 stress 0.1192678 #> Run 17 stress 0.1183186 #> ... Procrustes: rmse 3.001088e-06  max resid 9.607646e-06 #> ... Similar to previous best#> Run 18 stress 0.1192679 #> Run 19 stress 0.1808911 #> Run 20 stress 0.1183186 #> ... Procrustes: rmse 2.027412e-05  max resid 6.520856e-05 #> ... Similar to previous best#> *** Best solution repeated 4 times#查看结果#summary(nmds)

获取可视化所需数据。

#获得应力值（stress）stress <- nmds$stress#将绘图数据转化为数据框df <- as.data.frame(nmds$points)#与分组数据合并df <- cbind(df, dune.env)

NMDS可视化

根据分组绘制一个最基础的散点图。

p <- ggplot(df, aes(MDS1, MDS2))+  geom_point(aes(color = Management), size = 5)

我们注意到，原图中，每个分组被连接成不规则的多边形并用不同颜色表示，我们可以通过ggplot2中geom_polygon()来绘制。geom_polygon()会按照数据中出现的顺序连接观测值，内部可填充颜色。

p <- ggplot(df, aes(MDS1, MDS2))+  geom_point(aes(color = Management), size = 5)+  geom_polygon(aes(x = MDS1, y = MDS2, fill = Management, group = Management, color = Management),               alpha = 0.3, linetype = "longdash", linewidth = 1.5) #通过按顺序连接观测值绘制多边形

由于geom_polygon()会按照数据中出现的顺序连接观测值，因此如果我们按照df自身顺序来绘制多边形，多边形会非常奇怪，没法代表不同分组。因此，我们需要预先处理df的顺序，按合理的顺序连接观测值。

df <- df[order(df$Management), ]#先按分组排序df$Order <- c(2, 1, 3, 1, 2, 3, 4, 5, 3, 5, 1, 6, 2, 4, 1, 2, 6, 3, 5, 4)#添加一列Order，给每个分组内观测点的手动排序df <- df[order(df$Management, df$Order), ]#按分组和Order排序p <- ggplot(df, aes(MDS1, MDS2))+  geom_point(aes(color = Management), size = 5)+  geom_polygon(aes(x = MDS1, y = MDS2, fill = Management, group = Management, color = Management),               alpha = 0.3, linetype = "longdash", linewidth = 1.5)

分别进行Anosim分析（Analysis of similarities）和PERMANOVA（即adonis）检验分析。

#设置随机种子set.seed(123)#基于bray-curtis距离进行PERMANOVA分析adonis <-  adonis2(dune ~ Management, data = dune.env, permutations = 999, method = "bray")#基于bray-curtis距离进行anosim分析anosim = anosim(dune, dune.env$Management, permutations = 999, distance = "bray")

美化图片，并用AI微调。

# 应力值stress，Adonis R2与显著性，Anosim R与显著性stress_text <- paste("Stress  =", round(stress, 4))adonis_text <- paste(paste("Adonis  =", round(adonis$R2, 2)), "**")[1]anosim_text <- paste(paste("Anosim  =", round(anosim$statistic, 2)), "**")
p <- ggplot(df, aes(MDS1, MDS2))+  geom_point(aes(color = Management), size = 5)+  geom_polygon(aes(x = MDS1, y = MDS2, fill = Management, group = Management, color = Management), alpha = 0.3, linetype = "longdash", linewidth = 1.5)+  theme(plot.margin = unit(rep(1, 4), 'lines'),         panel.border = element_rect(fill = NA, color = "black", size = 0.5, linetype = "solid"),         panel.grid = element_blank(),         panel.background = element_rect(fill = 'white'))+  guides(color = "none", fill = "none")+  ggtitle(paste(paste(stress_text, adonis_text), anosim_text))

完整代码

if (!require("vegan"))  install.packages('vegan')if (!require("ggplot2"))  install.packages('ggplot2') # 加载包library(vegan)library(ggplot2) # 载入dune数据集data(dune)#载入dune包含分组信息等的元数据（即metadata），分组信息为Management列data(dune.env)#计算bray_curtis距离distance <- vegdist(dune, method = 'bray')#NMDS排序分析，k = 2预设两个排序轴nmds <- metaMDS(distance, k = 2)#> Run 0 stress 0.1192678 #> Run 1 stress 0.1192678 #> ... Procrustes: rmse 1.505128e-05  max resid 4.673581e-05 #> ... Similar to previous best#> Run 2 stress 0.1192678 #> ... Procrustes: rmse 3.715749e-06  max resid 1.009651e-05 #> ... Similar to previous best#> Run 3 stress 0.1889642 #> Run 4 stress 0.1192679 #> ... Procrustes: rmse 0.0001542849  max resid 0.0004702712 #> ... Similar to previous best#> Run 5 stress 0.1886532 #> Run 6 stress 0.2341212 #> Run 7 stress 0.1192678 #> ... Procrustes: rmse 1.328909e-05  max resid 4.273575e-05 #> ... Similar to previous best#> Run 8 stress 0.1886532 #> Run 9 stress 0.1192678 #> ... Procrustes: rmse 1.903819e-05  max resid 5.828243e-05 #> ... Similar to previous best#> Run 10 stress 0.1192678 #> ... Procrustes: rmse 6.358457e-06  max resid 1.687026e-05 #> ... Similar to previous best#> Run 11 stress 0.119268 #> ... Procrustes: rmse 5.501506e-05  max resid 0.0001605112 #> ... Similar to previous best#> Run 12 stress 0.1192678 #> ... New best solution#> ... Procrustes: rmse 5.074111e-06  max resid 1.393603e-05 #> ... Similar to previous best#> Run 13 stress 0.1192678 #> ... Procrustes: rmse 3.160318e-05  max resid 9.85043e-05 #> ... Similar to previous best#> Run 14 stress 0.1886532 #> Run 15 stress 0.2003486 #> Run 16 stress 0.2035424 #> Run 17 stress 0.1192678 #> ... Procrustes: rmse 2.440829e-05  max resid 7.079487e-05 #> ... Similar to previous best#> Run 18 stress 0.1183186 #> ... New best solution#> ... Procrustes: rmse 0.02027171  max resid 0.06497302 #> Run 19 stress 0.1183186 #> ... New best solution#> ... Procrustes: rmse 3.78469e-06  max resid 9.699447e-06 #> ... Similar to previous best#> Run 20 stress 0.1192678 #> *** Best solution repeated 1 times#查看结果#summary(nmds)#获得应力值（stress）stress <- nmds$stress#将绘图数据转化为数据框df <- as.data.frame(nmds$points)#与分组数据合并df <- cbind(df, dune.env)df <- df[order(df$Management), ]#先按分组排序df$Order <- c(2, 1, 3, 1, 2, 3, 4, 5, 3, 5, 1, 6, 2, 4, 1, 2, 6, 3, 5, 4)#添加一列Order，给每个分组内观测点的手动排序df <- df[order(df$Management, df$Order), ]#按分组和Order排序 #设置随机种子set.seed(123)#基于bray-curtis距离进行PERMANOVA分析adonis <- adonis2(dune ~ Management, data = dune.env, permutations = 999, method = "bray")#基于bray-curtis距离进行anosim分析anosim = anosim(dune, dune.env$Management, permutations = 999, distance = "bray") # 应力值stress，Adonis R2与显著性，Anosim R与显著性stress_text <- paste("Stress  =", round(stress, 4))adonis_text <- paste(paste("Adonis  =", round(adonis$R2, 2)), "**")[1]anosim_text <- paste(paste("Anosim  =", round(anosim$statistic, 2)), "**") p <- ggplot(df, aes(MDS1, MDS2))+  geom_point(aes(color = Management), size = 5)+  geom_polygon(aes(x = MDS1, y = MDS2, fill = Management, group = Management, color = Management),               alpha = 0.3, linetype = "longdash", linewidth = 1.5)+  theme(plot.margin = unit(rep(1, 4), 'lines'),         panel.border = element_rect(fill = NA, color = "black", size = 0.5, linetype = "solid"),         panel.grid = element_blank(),         panel.background = element_rect(fill = 'white'))+  guides(color = "none", fill = "none")+  ggtitle(paste(paste(stress_text, adonis_text), anosim_text))
ggsave("Figure6C.pdf", p, height = 5.69, width = 7.42)