图片来自Google

“进化树的构建怎么操作？”

“那肯定是用MEGA啊！”

图片来自MEGA官网

可是真的好麻烦啊，要先比对再建树，然后再进行各种美化，习惯了R就用R解决吧。

简单Google之后发现R有现成的包可以完成分析，包括了从序列读取、进化树构建、进化树美化等 。相关的R包主要是：ape、phangorn、seqinr、ggtree。

本文中的测试数据来自一个教程，公众号后PLANTOMIX台回复“进化树”获取下载链接。

---

关于进化树

常用的方法包括基于距离矩阵的UPGMA、ME（Minimum Evolution，最小进化法）和 NJ（Neighbor-Joining，邻接法）与非距离矩阵的MP（Maximum parsimony，最大简约法）、ML（Maximum likelihood，最大似然法）以及贝叶斯（Bayesian）推断等方法。现在UPGMA很少使用，在大多数文章里面基本是NJ或者ML。就速度来说，NJ是较快的，ML是很耗时的。

本文就简单描述如何在R中利用DNA序列构建进化树，本文的方法适用于DNA序列和氨基酸序列，蛋白序列的方法还在探索中。

 rm(list = ls())
 ​
 setwd('../../20200727【R】进化树构建与美化/')
 ​
 if (!requireNamespace(c('ape','phangorn','seqinr'))) {
  install.packages(c('ape','phangorn','seqinr'))
 }
 ​
 library(ape)
 library(phangorn)
 library(seqinr)
 library(ggtree)
 library(ggplot2)
 library(patchwork)
 ​
 test_dna = read.dna('data/test.fasta', format = 'fasta')
 test_phyDat = phyDat(test_dna, type = 'DNA', levels = NULL)
 ​
 # 模型评估
 mt = modelTest(test_phyDat)
 print(mt)
 ​
 # 计算距离
 dna_dist = dist.ml(test_phyDat,model = 'JC69')
 ​
 # NJ树
 tree_nj = NJ(dna_dist)
 write.tree()
 p_nj = ggtree(tree_nj, layout="radial") +
  geom_tiplab(size = 3, color = 'red')+
  labs(title = 'Neighbor Joining Tree')

 p_nj
 # UPGMA树
 tree_upgma = upgma(dna_dist)
 p_UPGMA = ggtree(tree_upgma) +
  geom_tiplab(size = 3, color = 'green')+
  labs(title = 'UPGMA Tree')
 p_UPGMA
 ​
 # 最大简约树
 parsimony(tree_upgma, data = test_phyDat)
 parsimony(tree_nj, data = test_phyDat)
 test_optim = optim.parsimony(tree_nj, data = test_phyDat)
 tree_peatchet = pratchet(test_phyDat)
 p_Maximum_Parsimony = ggtree(tree_peatchet,layout="radial") +
  geom_tiplab(size = 3, color = 'blue')+
  labs(title = 'Maximum Parsimony Tree')
 p_Maximum_Parsimony
 ​
 # 最大似然法
 tree_fit = pml(tree_nj, data = test_phyDat)
 print(tree_fit)
 fitJC = optim.pml(tree_fit, model = 'JC', rearrangement = 'stochastic')
 logLik(fitJC)
 tree_bs = bootstrap.pml(fitJC, bs = 100, optNni = T,
  multicore = F, 
  control = pml.control(trace = 0))
 tree_ml_bootstrap = plotBS(midpoint(fitJC$tree), tree_bs, p = 50, type = 'p')
 ​
 p_Maximum_Likelihood = ggtree(tree_ml_bootstrap) +
  geom_tiplab(size = 3, color = 'purple')+
  labs(title = 'Maximum Likelihood Tree')
 p_Maximum_Likelihood
 ​
p_all = p_nj + p_UPGMA + p_Maximum_Parsimony + p_Maximum_Likelihood +
        plot_layout(ncol = 2)
ggsave(p_all, filename = 'figures/all.pdf', width = 8, height = 8)