关于 etetoolkit
etetoolkit 是用于系统发育和层级树分析的工具包,面向生物信息学研究者。它结合树操作、进化事件分析、NCBI 分类信息和可视化,也支持聚类分析及树比较。
主要功能
- 01读取、编辑和转换多种树格式
- 02检测复制与物种形成事件
- 03识别直系同源与旁系同源基因
- 04整合 NCBI 分类信息
- 05生成树图并比较树拓扑
- 06分析聚类并绘制热图
适用场景
- 01整理和修剪系统发育树
- 02从基因树识别直系同源基因
- 03为树添加分类谱系信息
- 04比较多棵树或展示聚类结果
安装
复制下面的命令,在终端中运行。
npx skills add https:/ / github.com/ jackspace/ claudeskillz --skill etetoolkit安装后在 Claude Code 中输入/etetoolkit确认已加载。
附带文件SKILL.json
详细介绍
ETE Toolkit 技能
概述
ETE(Environment for Tree Exploration,树探索环境)是一款用于系统发育和层次树分析的工具包。它支持操作树、分析进化事件、可视化结果,并与生物数据库集成,适用于系统基因组学研究和聚类分析。
核心功能
1. 树操作与分析
加载、操作和分析层次树结构,支持:
- 树的输入/输出:读取和写入 Newick、NHX、PhyloXML 和 NeXML 格式
- 树遍历:使用前序、后序或层序策略遍历树
- 拓扑修改:剪枝、设定根节点、折叠节点、解析多分叉
- 距离计算:计算节点之间的分支长度和拓扑距离
- 树比较:计算 Robinson-Foulds 距离并识别拓扑差异
常见模式:
from ete3 import Tree
# Load tree from file
tree = Tree("tree.nw", format=1)
# Basic statistics
print(f"Leaves: {len(tree)}")
print(f"Total nodes: {len(list(tree.traverse()))}")
# Prune to taxa of interest
taxa_to_keep = ["species1", "species2", "species3"]
tree.prune(taxa_to_keep, preserve_branch_length=True)
# Midpoint root
midpoint = tree.get_midpoint_outgroup()
tree.set_outgroup(midpoint)
# Save modified tree
tree.write(outfile="rooted_tree.nw")
使用 scripts/tree_operations.py 从命令行操作树:
# Display tree statistics
python scripts/tree_operations.py stats tree.nw
# Convert format
python scripts/tree_operations.py convert tree.nw output.nw --in-format 0 --out-format 1
# Reroot tree
python scripts/tree_operations.py reroot tree.nw rooted.nw --midpoint
# Prune to specific taxa
python scripts/tree_operations.py prune tree.nw pruned.nw --keep-taxa "sp1,sp2,sp3"
# Show ASCII visualization
python scripts/tree_operations.py ascii tree.nw
2. 系统发育分析
通过检测进化事件分析基因树:
- 序列比对集成:将树与多序列比对(FASTA、Phylip)关联
- 物种命名:自动从基因名称提取物种,或自定义提取方式
- 进化事件:使用物种重叠或树重建检测复制和物种形成事件
- 直系同源关系检测:根据进化事件识别直系同源基因和旁系同源基因
- 基因家族分析:按复制事件拆分树,折叠谱系特异性扩增
基因树分析工作流:
from ete3 import PhyloTree
# Load gene tree with alignment
tree = PhyloTree("gene_tree.nw", alignment="alignment.fasta")
# Set species naming function
def get_species(gene_name):
return gene_name.split("_")[0]
tree.set_species_naming_function(get_species)
# Detect evolutionary events
events = tree.get_descendant_evol_events()
# Analyze events
for node in tree.traverse():
if hasattr(node, "evoltype"):
if node.evoltype == "D":
print(f"Duplication at {node.name}")
elif node.evoltype == "S":
print(f"Speciation at {node.name}")
# Extract ortholog groups
ortho_groups = tree.get_speciation_trees()
for i, ortho_tree in enumerate(ortho_groups):
ortho_tree.write(outfile=f"ortholog_group_{i}.nw")
查找直系同源基因和旁系同源基因:
# Find orthologs to query gene
query = tree & "species1_gene1"
orthologs = []
paralogs = []
for event in events:
if query in event.in_seqs:
if event.etype == "S":
orthologs.extend([s for s in event.out_seqs if s != query])
elif event.etype == "D":
paralogs.extend([s for s in event.out_seqs if s != query])
3. NCBI 分类数据库集成
集成 NCBI 分类数据库中的分类信息:
- 数据库访问:自动下载 NCBI 分类数据库并缓存到本地(约 300MB)
- Taxid/名称转换:在分类 ID 和学名之间进行转换
- 谱系检索:获取完整的进化谱系
- 分类树:构建连接指定分类单元的物种树
- 树注释:自动为树添加分类信息注释
构建基于分类信息的树:
from ete3 import NCBITaxa
ncbi = NCBITaxa()
# Build tree from species names
species = ["Homo sapiens", "Pan troglodytes", "Mus musculus"]
name2taxid = ncbi.get_name_translator(species)
taxids = [name2taxid[sp][0] for sp in species]
# Get minimal tree connecting taxa
tree = ncbi.get_topology(taxids)
# Annotate nodes with taxonomy info
for node in tree.traverse():
if hasattr(node, "sci_name"):
print(f"{node.sci_name} - Rank: {node.rank} - TaxID: {node.taxid}")
为现有树添加注释:
# Get taxonomy info for tree leaves
for leaf in tree:
species = extract_species_from_name(leaf.name)
taxid = ncbi.get_name_translator([species])[species][0]
# Get lineage
lineage = ncbi.get_lineage(taxid)
ranks = ncbi.get_rank(lineage)
names = ncbi.get_taxid_translator(lineage)
# Add to node
leaf.add_feature("taxid", taxid)
leaf.add_feature("lineage", [names[t] for t in lineage])
4. 树可视化
创建适用于出版物的高质量树可视化图:
- 输出格式:用于出版物的 PNG(栅格)、PDF 和 SVG(矢量)格式
- 布局模式:矩形和圆形树布局
- 交互式 GUI:通过缩放、平移和搜索交互式探索树
- 自定义样式:使用 NodeStyle 设置节点外观(颜色、形状、大小)
- 图形元素:向节点添加图形元素(文本、图像、图表、热图)
- 布局函数:根据节点属性动态设置样式
基本可视化工作流:
from ete3 import Tree, TreeStyle, NodeStyle
tree = Tree("tree.nw")
# Configure tree style
ts = TreeStyle()
ts.show_leaf_name = True
ts.show_branch_support = True
ts.scale = 50 # pixels per branch length unit
# Style nodes
for node in tree.traverse():
nstyle = NodeStyle()
if node.is_leaf():
nstyle["fgcolor"] = "blue"
nstyle["size"] = 8
else:
# Color by support
if node.support > 0.9:
nstyle["fgcolor"] = "darkgreen"
else:
nstyle["fgcolor"] = "red"
nstyle["size"] = 5
node.set_style(nstyle)
# Render to file
tree.render("tree.pdf", tree_style=ts)
tree.render("tree.png", w=800, h=600, units="px", dpi=300)
使用 scripts/quick_visualize.py 快速生成可视化图:
# Basic visualization
python scripts/quick_visualize.py tree.nw output.pdf
# Circular layout with custom styling
python scripts/quick_visualize.py tree.nw output.pdf --mode c --color-by-support
# High-resolution PNG
python scripts/quick_visualize.py tree.nw output.png --width 1200 --height 800 --units px --dpi 300
# Custom title and styling
python scripts/quick_visualize.py tree.nw output.pdf --title "Species Phylogeny" --show-support
使用图形元素进行高级可视化:
from ete3 import Tree, TreeStyle, TextFace, CircleFace
tree = Tree("tree.nw")
# Add features to nodes
for leaf in tree:
leaf.add_feature("habitat", "marine" if "fish" in leaf.name else "land")
# Layout function
def layout(node):
if node.is_leaf():
# Add colored circle
color = "blue" if node.habitat == "marine" else "green"
circle = CircleFace(radius=5, color=color)
node.add_face(circle, column=0, position="aligned")
# Add label
label = TextFace(node.name, fsize=10)
node.add_face(label, column=1, position="aligned")
ts = TreeStyle()
ts.layout_fn = layout
ts.show_leaf_name = False
tree.render("annotated_tree.pdf", tree_style=ts)
5. 聚类分析
结合数据分析层次聚类结果:
- ClusterTree:专用于聚类树状图的类
- 数据矩阵关联:将树叶与数值特征关联
- 聚类指标:轮廓系数、Dunn 指数、簇间/簇内距离
- 验证:使用不同距离指标测试聚类质量
- 热图可视化:在树旁显示数据矩阵
聚类工作流:
from ete3 import ClusterTree
# Load tree with data matrix
matrix = """#Names\tSample1\tSample2\tSample3
Gene1\t1.5\t2.3\t0.8
Gene2\t0.9\t1.1\t1.8
Gene3\t2.1\t2.5\t0.5"""
tree = ClusterTree("((Gene1,Gene2),Gene3);", text_array=matrix)
# Evaluate cluster quality
for node in tree.traverse():
if not node.is_leaf():
silhouette = node.get_silhouette()
dunn = node.get_dunn()
print(f"Cluster: {node.name}")
print(f" Silhouette: {silhouette:.3f}")
print(f" Dunn index: {dunn:.3f}")
# Visualize with heatmap
tree.show("heatmap")
6. 树比较
量化树之间的拓扑差异:
- Robinson-Foulds 距离:树比较的标准指标
- 标准化 RF:与尺度无关的距离(0.0 至 1.0)
- 划分分析:识别独有和共有的二分划分
- 共识树:分析多棵树的支持情况
- 批量比较:两两比较多棵树
比较两棵树:
from ete3 import Tree
tree1 = Tree("tree1.nw")
tree2 = Tree("tree2.nw")
# Calculate RF distance
rf, max_rf, common_leaves, parts_t1, parts_t2 = tree1.robinson_foulds(tree2)
print(f"RF distance: {rf}/{max_rf}")
print(f"Normalized RF: {rf/max_rf:.3f}")
print(f"Common leaves: {len(common_leaves)}")
# Find unique partitions
unique_t1 = parts_t1 - parts_t2
unique_t2 = parts_t2 - parts_t1
print(f"Unique to tree1: {len(unique_t1)}")
print(f"Unique to tree2: {len(unique_t2)}")
比较多棵树:
import numpy as np
trees = [Tree(f"tree{i}.nw") for i in range(4)]
# Create distance matrix
n = len(trees)
dist_matrix = np.zeros((n, n))
for i in range(n):
for j in range(i+1, n):
rf, max_rf, _, _, _ = trees[i].robinson_foulds(trees[j])
norm_rf = rf / max_rf if max_rf > 0 else 0
dist_matrix[i, j] = norm_rf
dist_matrix[j, i] = norm_rf
安装和设置
安装 ETE 工具包:
# Basic installation
pip install ete3
# With external dependencies for rendering (optional but recommended)
# On macOS:
brew install qt@5
# On Ubuntu/Debian:
sudo apt-get install python3-pyqt5 python3-pyqt5.qtsvg
# For full features including GUI
pip install ete3[gui]
首次设置 NCBI Taxonomy:
首次实例化 NCBITaxa 时,它会自动将 NCBI taxonomy 数据库(约 300MB)下载到 ~/.etetoolkit/taxa.sqlite。此操作只会执行一次:
from ete3 import NCBITaxa
ncbi = NCBITaxa() # Downloads database on first run
更新 taxonomy 数据库:
ncbi.update_taxonomy_database() # Download latest NCBI data
常见用例
用例 1:系统基因组学工作流
从基因树到直系同源基因识别的完整工作流:
from ete3 import PhyloTree, NCBITaxa
# 1. Load gene tree with alignment
tree = PhyloTree("gene_tree.nw", alignment="alignment.fasta")
# 2. Configure species naming
tree.set_species_naming_function(lambda x: x.split("_")[0])
# 3. Detect evolutionary events
tree.get_descendant_evol_events()
# 4. Annotate with taxonomy
ncbi = NCBITaxa()
for leaf in tree:
if leaf.species in species_to_taxid:
taxid = species_to_taxid[leaf.species]
lineage = ncbi.get_lineage(taxid)
leaf.add_feature("lineage", lineage)
# 5. Extract ortholog groups
ortho_groups = tree.get_speciation_trees()
# 6. Save and visualize
for i, ortho in enumerate(ortho_groups):
ortho.write(outfile=f"ortho_{i}.nw")
用例 2:树预处理和格式化
批量处理树以供分析:
# Convert format
python scripts/tree_operations.py convert input.nw output.nw --in-format 0 --out-format 1
# Root at midpoint
python scripts/tree_operations.py reroot input.nw rooted.nw --midpoint
# Prune to focal taxa
python scripts/tree_operations.py prune rooted.nw pruned.nw --keep-taxa taxa_list.txt
# Get statistics
python scripts/tree_operations.py stats pruned.nw
用例 3:出版级图表
创建经过样式设计的可视化图表:
from ete3 import Tree, TreeStyle, NodeStyle, TextFace
tree = Tree("tree.nw")
# Define clade colors
clade_colors = {
"Mammals": "red",
"Birds": "blue",
"Fish": "green"
}
def layout(node):
# Highlight clades
if node.is_leaf():
for clade, color in clade_colors.items():
if clade in node.name:
nstyle = NodeStyle()
nstyle["fgcolor"] = color
nstyle["size"] = 8
node.set_style(nstyle)
else:
# Add support values
if node.support > 0.95:
support = TextFace(f"{node.support:.2f}", fsize=8)
node.add_face(support, column=0, position="branch-top")
ts = TreeStyle()
ts.layout_fn = layout
ts.show_scale = True
# Render for publication
tree.render("figure.pdf", w=200, units="mm", tree_style=ts)
tree.render("figure.svg", tree_style=ts) # Editable vector
用例 4:自动化树分析
系统地处理多棵树:
from ete3 import Tree
import os
input_dir = "trees"
output_dir = "processed"
for filename in os.listdir(input_dir):
if filename.endswith(".nw"):
tree = Tree(os.path.join(input_dir, filename))
# Standardize: midpoint root, resolve polytomies
midpoint = tree.get_midpoint_outgroup()
tree.set_outgroup(midpoint)
tree.resolve_polytomy(recursive=True)
# Filter low support branches
for node in tree.traverse():
if hasattr(node, 'support') and node.support < 0.5:
if not node.is_leaf() and not node.is_root():
node.delete()
# Save processed tree
output_file = os.path.join(output_dir, f"processed_{filename}")
tree.write(outfile=output_file)
参考文档
如需完整的 API 文档、代码示例和详细指南,请参阅 references/ 目录中的以下资源:
api_reference.md:所有 ETE 类和方法(Tree、PhyloTree、ClusterTree、NCBITaxa)的完整 API 文档,包括参数、返回类型和代码示例workflows.md:按任务分类的常见工作流模式(树操作、系统发育分析、树比较、taxonomy 集成、聚类分析)visualization.md:全面的可视化指南,涵盖 TreeStyle、NodeStyle、Faces、布局函数和高级可视化技术
需要详细信息时,请加载这些参考资料:
# To use API reference
# Read references/api_reference.md for complete method signatures and parameters
# To implement workflows
# Read references/workflows.md for step-by-step workflow examples
# To create visualizations
# Read references/visualization.md for styling and rendering options
故障排除
导入错误:
# If "ModuleNotFoundError: No module named 'ete3'"
pip install ete3
# For GUI and rendering issues
pip install ete3[gui]
渲染问题:
如果 tree.render() 或 tree.show() 因 Qt 相关错误而失败,请安装系统依赖项:
# macOS
brew install qt@5
# Ubuntu/Debian
sudo apt-get install python3-pyqt5 python3-pyqt5.qtsvg
NCBI Taxonomy 数据库:
如果数据库下载失败或损坏:
from ete3 import NCBITaxa
ncbi = NCBITaxa()
ncbi.update_taxonomy_database() # Redownload database
大树的内存问题:
对于非常大的树(>10,000 个叶节点),请使用迭代器,而不要使用列表推导式:
# Memory-efficient iteration
for leaf in tree.iter_leaves():
process(leaf)
# Instead of
for leaf in tree.get_leaves(): # Loads all into memory
process(leaf)
Newick 格式参考
ETE 支持多种 Newick 格式规范(0-100):
- 格式 0:灵活,包含分支长度(默认)
- 格式 1:包含内部节点名称
- 格式 2:包含 bootstrap/支持值
- 格式 5:内部节点名称 + 分支长度
- 格式 8:包含所有特性(名称、距离、支持值)
- 格式 9:仅叶节点名称
- 格式 100:仅拓扑结构
读写时指定格式:
tree = Tree("tree.nw", format=1)
tree.write(outfile="output.nw", format=5)
NHX(New Hampshire eXtended)格式可保留自定义特性:
tree.write(outfile="tree.nhx", features=["habitat", "temperature", "depth"])
最佳实践
- 保留分支长度:在系统发育分析中进行剪枝时,使用
preserve_branch_length=True - 缓存内容:对于大型树上节点内容的重复访问,使用
get_cached_content() - 使用迭代器:使用
iter_*方法,以节省内存地处理大型树 - 选择合适的遍历方式:自底向上分析使用后序遍历,自顶向下分析使用前序遍历
- 验证单系性:始终检查返回的分支类型(单系、并系或多系)
- 用于出版的矢量格式:出版图表使用 PDF 或 SVG(可缩放、可编辑)
- 交互式测试:先使用
tree.show()测试可视化效果,再渲染到文件 - 系统发育分析使用 PhyloTree:对基因树和进化分析使用 PhyloTree 类
- 选择复制方法:追求速度时使用 "newick",追求完整保真时使用 "cpickle",处理复杂对象时使用 "deepcopy"
- 缓存 NCBI 查询:存储 NCBI taxonomy 查询结果,以避免重复访问数据库