Miao Sun | Miao Sun | 孙苗

How to built a completely resolved species-level phylogeny using TCAT

Sun, 06 Sep 2020 00:00:00 +0000

Phylogenetic tree is a crucial reference system for biological research. It carries the key information for species diversification through geological time. A phylogeny tree with all members of a lineage presented is necessary for questions in macroecology and macroevolution. Hence, there are many comparative biological methods (e.g., diversification) integreated “sampling fractions” or other approches to correct sparsely sampled phylogenies of higher taxonomic groups. For example, “sampling fractions” in RPANDA (Morlon et al. 2016) and “global sampling fraction” and “clade-specific sampling fraction” in BAMM (Rabosky, 2014; Sun et al., 2020). However, this approach could lead to low statistical power to infer rate variation, or spurious patterns of diversification, which may further impact the evolutionary implication (Chang et al., 2019).
A new stochastic polytomy resolution method — Taxonomic Addition for Complete Trees (TACT) is able to avoid the biases arised above, and also can greatly improve the power of estimation in diversification analyses (Chang et al., 2019). It uses birth–death-sampling model and taxonomic information to place unsampled taxa onto the ultrametric target backbone phylogeny. It produces a pseduposterior distribution of phylogenies, integrating diversification rates with phylogenetic uncertainty and speciation times, while this approach also can be labor and computation time intense as well as require topology constraint. Besides TACT, other stochastic polytomy resolvers, including PASTIS (Thomas et al., 2013) and CorSiM (Cusimano et al., 2012), but TACT performes better (See Chang et al., 2019).

The caveats which researchers should pay attention is that complete phylogenies generared by stochastic polytomy resolvers should NOT be used for estimating the trait evolution. Because the way of how complete tree is generated, the true rate of trait evolution can be overestimated, and the phylogenetic signal is lost.

The Author Jonathan Chang already has a detailed tutorial of how to install and run TACT. Here I only focus on how to prepare a taxonomy list and built up pipeline for my own project.

General steps

First thing first, you need two files to run TACT (assuming that you have successfully installed TACT and its dependencies):

An ultrametric backbone phylogeny (e.g., Vitales_dated.tre)
A taxonomy of target clade

this can be taxonomy tree (see Chang et al., 2019)
can also be a taxonomic list in csv format with all taxonomic ranks listed (e.g., genus, family, order; here as Vitales.csv)

For the case in my post, I use Vitales as an example. The Vitales taxonomic list is generated from WCSP website, and the higher taxonomic ranks are queried from APWeb. The backbone tree was built using RAxML with molecular data mined from GenBank via PyPHLAWD, and then dated using treePL.

1. Prepare a taxonomic list

A csv format
All taxonomic ranks referenced from APWeb
Columns layout as order,family,genus,genus.species

# devtools::install_github("wcornwell/taxonlookup")
rm(list=ls())
library("taxonlookup")
library("ape")
library("dplyr", warn.conflicts = FALSE, quietly = TRUE)
#read in the complete taxonomic list
Vitales <- read.csv("Vitales_rawlist.csv", header = T, stringsAsFactors = F)
# > head(Vitales)
# [1] "Acareosperma spireanum"
# [2] "Ampelocissus abyssinica"
# [3] "Ampelocissus acapulcensis"
# [4] "Ampelocissus acetosa"
# [5] "Ampelocissus aculeata"
# [6] "Ampelocissus africana var. migeodii"
# lookup APGWeb based on species names
Vitales.apg <- lookup_table(unique(Vitales), missing_action = "NA", by_species=TRUE)
#just manual check if there are some genera are failed to have query
if(sum(is.na(Vitales.apg$family)) >0){
tt <- Vitales.apg[is.na(Vitales.apg$family),]
print(tt)
}
# rearrange columns as taxonomic ranks
APG_table <- Vitales.apg %>% mutate(genus.species=row.names(Vitales.apg)) %>% select(order, family, genus, genus.species)
#
#> head(APG_table)
#order family genus genus.species
#1 Vitales Vitaceae Acareosperma Acareosperma spireanum
#2 Vitales Vitaceae Ampelocissus Ampelocissus abyssinica
#3 Vitales Vitaceae Ampelocissus Ampelocissus acapulcensis
#4 Vitales Vitaceae Ampelocissus Ampelocissus acetosa
#5 Vitales Vitaceae Ampelocissus Ampelocissus aculeata
#6 Vitales Vitaceae Ampelocissus Ampelocissus africana var. migeodii
#replace space as underscore
APG_table$genus.species <- gsub(" ", "_",APG_table$genus.species)
#save the apg lookup table, and prepare for next step TACT
write.csv(APG_table, "Vitales.csv", row.names=FALSE, quote=FALSE)

2. Run TACT

#Build a taxonomic tree
tact_build_taxonomic_tree Vitales.csv --output Vitales.taxonomy.tre
# this step adove can be skipped, if you already have a taxonomy tree
#Run the TACT stochastic polytomy resolver algorithm in conjunction with the backbone phylogeny
tact_add_taxa --backbone Vitales_dated.tre --taxonomy Vitales.taxonomy.tre --output Vitales.tacted --verbose
#check the TACT results
tact_check_results Vitales.tacted.newick.tre --backbone Vitales_dated.tre --taxonomy Vitales.taxonomy.tre > Vitales_checkresults.csv

3. Visualize and compare trees

library("ape")
tree1 <- ladderize(read.tree("./Vitales_dated.tre"))
plot.phylo(tree1, cex=0.1, type = "fan", main="Backbone tree with NCBI data")

This is the original backbone tree before TACT treatment

Vitales backbone

Now we show which tips on the complete species-level tree are added by TACT:

tree2 <- ladderize(read.tree("./Vitales.tacted.newick.tre"))
tip.cl <- ifelse(tree2$tip.label %in% tree1$tip.label, "black", "red")
plot.phylo(tree2, tip.color=tip.cl, cex=0.1, type = "fan", main="complete tree with NCBI + Taxonomic data")
mtext("Red tips are species added by TACT", col = "red", side=1, line=-5, outer=TRUE)

Vitales Tacted Tree1

Last, We highlight which tree edges on the complete species-level tree are added by TACT:

library("phytools")
x <- setNames(ifelse(tree2$tip.label %in% tree1$tip.label, "0", "1"), tree2$tip.label)
a <- names(x)[x=="0"]
b <- names(x)[x=="1"]
tt<-paintBranches(tree2,edge=sapply(b,match,tree2$tip.label),
state="b",anc.state="a")
cols<-setNames(c("black","red"),c("a","b"))
plot(tt, colors=cols, fsize=0.1, type="fan", lwd=0.5)
mtext("Red branches are imputated by TACT", col = "red", side=3, line=-1.5, outer=TRUE)

Vitales Tacted Tree2

That’s it.

Reference

Thomas G.H. et al. PASTIS: an R package to facilitate phylogenetic assembly with soft taxonomic inferences. Methods Ecol. Evol. 4:1011–1017 (2013).
Chang et al. Estimating Diversification Rates on Incompletely Sampled Phylogenies: Theoretical Concerns and Practical Solutions. Syst. Biol. 69(3):602–611, 2020
Cusimano N.et al. A new method for handling missing species in diversification analysis applicable to randomly or nonrandomly sampled phylogenies. Syst. Biol. 61:785–792 (2012).
Morlon, H. et al. RPANDA: an R package for macroevolutionary analyses on phylogenetic trees. Methods Ecol. Evol. 7, 589–597 (2016).
Rabosky, D. L. Automatic detection of key innovations, rate shifts, and diversity-dependence on phylogenetic trees. PLoS ONE 9, e89543 (2014).
Revell, J. L. Phytools: an R package for phylogenetic comparative biology (and other things). Methods Ecol. Evol. 3, 217–223 (2012).
Sun, M. et al. Estimating rates and patterns of diversification with incomplete sampling: A case study in the rosids. Am. J. Bot. 107, 1–15 (2020).

AJB: Estimating rates and patterns of diversification with incomplete sampling: a case study in the rosids

Tue, 09 Jun 2020 00:00:00 +0000

Demonstrating gene tree conflict with Phyparts Piecharts

Sun, 27 Oct 2019 00:00:00 +0000

We have marched into genome era already, especially using biparental inherited nuclear genome data to examine the framework of tree of life established by plastid data. However, not all gene trees tell one same story, even the most homogeneous genes from plastid organelle. Not to mention nuclear genes may have more than one copy, and other potential biological processes (e.g., hybridization, incomplete lineage sorting (ILS), or horizontal gene transfer, etc); also see Sun et al. (2015). Hence, Smith et al. (2015) did a great job in summarizing current situation, and setting examples of visualization of concordance information in animals and plants genomic phylogenies. They also developed an open source java software to do this job — phyparts.

In this post I will use Sassafras 322 gene trees as an example to show you how to use ASTRAL-III to estimate a species tree, and phyparts to summarize the conflict and concordance information of those individual homologous gene regions, and finally using PhyParts PieCharts to visualize the Phyparts Output. These data was generated by Target Enrichment method using Universal Probe Set for Targeted Sequencing of 353 Nuclear Genes under Dimension project. My other relevant workingflow and scripts are available here.

Note: Matt Johnson has a great tutorial of how to run his script using Jupyter Notebook. The interpretation of PieCharts is well explained as well.

My instructions here focus on the overall procedures from gene trees all the way down to the Pie chart. Hopefully, I’m able to integrate all the steps together into one pipeline.

General steps

Here we need to assess how a number of gene trees how they agree with ours pecies tree, and displaying this discordance and agreeent information with satisfying visualization.

1. Building phylogeny, so that we have all gene trees

My data was generated from 353 target enrichment method
I used 322 gene trees from Sassafras samples as example
Note:
- All trees were rooted by either phyx or The Newick Utilities, or DendroPy
- Though ASTRAL is able to take any gene trees rooted or unrooted, however, for downstream Phyparts performance, rooted trees are preferred. Such so all gene trees and species trees are towards to the same root direction (See Matt Johnson’s post).
- As quote in ASTRAL tutorial:
  “Importantly, we will reroot the tree at the correct node, which is always necessary, since the rooting of the ASTRAL trees is arbitrary and meaningless."
Therefore, rooting all the trees (species tree and gene trees) is preferred for Phyparts Piecharts. Sometimes, when you have hundreds of gene trees, however, not all outgroups presented in your gene tree; in these case, I recommend either use pxrr function from phyx, or MAD (root the tree by Minimal Ancestor Deviation); the latter works well with those trees without outgroups at all; for details see my other post here

2. Species tree estimation

Species tree was estimated by ASTRAL-III, and see tutorials here
Note: Collapse gene tree nodes with BS support less than certain value (saying 10%; see cmd below), will help to improve accuracy; sometime increasing the threshold for collapse may yield better results.
There are many ways to do this: phyx or The Newick Utilities
- using The Newick Utilities as example from ASTRAL-III tutorial e.g., nw_ed 1KP-genetrees.tre 'i & b<=10' o > 1KP-genetrees-BS10.tre
ASTRAL CMD:
java -jar astral.5.6.3.jar -i collapse_genetrees.tre -o output_species_tree.tre 2> running.log

3. Statistic information about conflict, concordance, or even gene duplications

You need to run phyparts with instruction on the repo webpage.
Phyparts “conflict” option CMD:
java -jar target/phyparts-0.0.1-SNAPSHOT-jar-with-dependencies.jar -a 1 -v -d gene_trees -m ASTRAL_species_tree -o output_name

Note: run cmd below, you’ll able to see all the Phyparts options (see snapshot below):
java -jar target/phyparts-0.0.1-SNAPSHOT-jar-with-dependencies.jar

4. Mapping these information on species tree

You need PhyParts PieCharts python script and tutorial from Matt Johnson

Note: Please read the tutorial, Python >2.7 and ETE3 with the graphical options need to be installed before run the python script

how to run script:
python3 phypartspiecharts.py species_tree output_name gene_number

Note: output_name must be the same as you names at the phyparts (# 3) step

5. Visualize the gene tree support and conflic statics with Piecharts

the output format of PhyParts PieCharts is in svg, then you need to Inkscape or AI to convert it as pdf. You also can refine the overall layout of that plot.
Note: make sure Python3, ETE3, and X server installed.

As my plot below:

Last updated: Mon Dec 7 2020

MPE: Accelerated diversification correlated with functional traits shapes extant diversity of the early divergent angiosperm family Annonaceae

Sat, 26 Oct 2019 00:00:00 +0000

Research Talk at IBCAS Youth Forum

Wed, 25 Sep 2019 00:00:00 +0000

I had the great honor and was invited by Dr. Zhiduan Chen to visit his lab. We had great conversations and we also discussed to collaborate a few great project in comming futher.

Later on I was invited by Dr. Limin Lu, who hosted a Youth Forum organized by the State Key Laboratory of Systematic and Evolutionary Botany (LSEB) at IBCAS.

I briedly introduced my research works conducted in these 5-year post-doc life in Soltis Lab. I used rosids as an example to illustrate how to build and use large-scale phylogentic trees. Key points are:

Data mining public database (e.g, GenBank and GBIF), and data cleaning for downstream analyses. E.g, recent manuscript in bioRxiv and paper in AJB.
Knowing the potential issues when applying large-scale data for macro-ecology and -evolution. E.g., papers in AJB, Ecology, and another recent manuscript in bioRxiv.
Large-scale trees can be used for [Diversification]. E.g., recent manuscript in bioRxiv and an accepted paper collaborated with Dr. Xue.
Large-scale trees can be used for Phylogentic Diversity (PD)

Relationships matter!!

Please let me know if you have any questions and want to know more details.

Overall, it turned out full house!! More than expected!! ✌️ ✌️

Especially I have met Prof. Jianfeng Mao from Beijing Forestry University. I had great cinversation with him and his students.

Research Lightning Talks at Florida Museum of Natural History

Fri, 01 Feb 2019 00:00:00 +0000

All my research is phylogeny based. Phylogeny trees are illustrating relationships among organisms. Relationships matter! In my understanding, phylogeny serves as a beautiful platform, which links all the biology data together with an evolutionary background. I’m insterested in phylogeny of angiosperms, particularly rosids.

I briedly introduced three main research interests:

It turned out that people like my talk! ✌️ ✌️

You can download my talk slides here 👈

Field work activities from 2006-2012 in China

Sun, 13 Jan 2019 00:00:00 +0000

I have visited a lot places in south part of China. These places are genreally hostspots of biodiversity of China. Especially I love Xizang! I hope I have chance to introduce more details of this facinating landscape, spelendid vegation, and each plant I came across during my trip.

🍀 🌸 🌿 🌵 🍄 🌴 🌳 🌼

Here some bullets of my trips I have make since 2006 - 2012:

2012
Trip to southern Yunnan along Myanmar border with Pam S. Soltis and Doug E. Soltis, collecting materials for Tree of life — China Project. Trip to Hainan tropical rain forest with Doug E. Soltis, collecting materials of Tree for life — China Project.
2006 ~ 2011
Joined in the expedition team of national herbarium (PE) for collection many times, successively visited a series of biodiversity hotspots, such as Jilin, Hubei, Chongqing, Guangxi, Sichuan, Xizang, etc.

I use R code and Baidumap API to show my trips:

Warning: This script may not work well in R markdown enviroment, but I tried by copying out the code as a single script in R studio. It works.

rm(list=ls())
# you need to install these packages if have not done so.
#install.packages("devtools")
# devtools::install_github("lchiffon/REmap")
# devtools::install_github("badbye/baidumap"")
install.packages("png")
library("png")
library("REmap")
library("baidumap")
# I realized chinese string have more success match with their coordinates
city<- c("北京","吉林省", "广东","湖北","重庆","广西","西藏","四川", "云南", "海南")
begin<-rep("beijing",length(city)) #(起点)
mapdata<-data.frame(begin,city)
getBaiduMap2(city, width = 400, height = 400, zoom = 10, scale = 2,
color = "color", messaging = TRUE)
map_pic <- remap(mapdata = mapdata, #数据框
title = "Footsteps covered most hotspots of China", #标题
subtitle = "Field work activities from 2006-2012", #副标题
theme = get_theme(theme = "Dark")) #主题
plot(map_pic)

Field trip trail

Plotting Christmas Tree

Mon, 17 Dec 2018 00:00:00 +0000

I saw a figure of all potential Christmas tree 🎄 at 🐦 Polypompholyx’s Tweet.

It looks strange for my first impression. So I did a quick check from Open Tree. This looks much better!

See code below:

rm(list=ls())
# install.packages("rotl")
# install.packages("ape")
#loading libraries
library("rotl") #for querying open tree database
library("ape") #for tree manipulation
#First I preapred a table for those potential **Christmas Trees** with first column is common names,a nd second column is the scientific names
# then we read in the table as a query list to the Open Tree
query <- read.csv("XsmasTree.csv", header=TRUE, sep=",", stringsAsFactors=FALSE, quote="")
# check what our table look like this
head(query)
# query names and get ottid
species <- tnrs_match_names(names = query$Scientific_name_rept.)
# check results
head(species)
# fix fuzzy match on row 16, then updated it
new_species<- update(species, row_number=16, new_ott_id = 248313)
# checking object
head(new_species)
species <- new_species
species.ottid <- ott_id(species)
# extracting the phylogeny summarized in the Open tree
tree <- tol_induced_subtree(ott_ids=species$ott_id, label_format = "name")
# make it looks good
tree <- ladderize(tree)
plot.phylo(tree, cex=0.8)
# rename the tree tip labels
query[[2]][16] <- "Picea pungens" #remove one extra space
query[[2]] <- gsub(pattern="\\s", "_", query[[2]])
new_label <- paste0(tree$tip.label,"(", query[[1]][match(tree$tip.label, query[[2]])], ")")
new_tree <- tree
new_tree$tip.label <- new_label
#png("The Tree of Christams Tree.png")
#plot(new_tree, cex=0.7)
#dev.off()
#write.tree(new_tree, "renamed_Xsmas_tree.tre")
#plot a prettier tree
plot.phylo(new_tree, edge.color = "dark green", edge.width = 2, tip.color = "red",
label.offset = 0.5, main="The Tree of Christmas Tree", col.main="red")
mtext("---source from Open Tree", col="dark green", 3)

Tree of Xsmas Trees

❄️ 🎅 🌵 🔔 ☃️ 🌲

Nature---ISME: Phylogenetic imprint of woody plants on the soil mycobiome in natural mountain forests of eastern China

Mon, 01 Jan 2018 00:00:00 +0000

JSE: Phylogeny of the Rosidae: A dense taxon sampling analysis

Fri, 01 Jan 2016 00:00:00 +0000

Exploring deep phylogenetic incongruence of the COM clade in Rosidae: Phylogenomics approach [PhD dissertation]

Wed, 28 May 2014 00:00:00 +0000

Genome: Identification of nuclear low-copy genes and their phylogenetic utility in rosids

Wed, 01 Jan 2014 00:00:00 +0000

JSE: A revision of Elaeagnus L. (Elaeagnaceae) in mainland China

Sat, 01 May 2010 00:00:00 +0000

Primary Taxonomic Study of Elaeagnus ( Elaeagnaceae) in Mainland China

Mon, 01 Jun 2009 00:00:00 +0000

Lectotypification of five scientific names in Rosaceae [In Chinese]

Tue, 01 Jan 2008 00:00:00 +0000