在上篇文章中，我们初步探索了PHATE的使用方法，发现它在揭示一些连续分化过程中不同细胞状态之间的微小局部差异具有很好的效果，同时也能保留细胞全局的整体结构。在本节教程中，我们将复现演示近期发表在Science Immunology期刊上的一篇文章的结果，进一步学习PHATE的相关使用方法。
image.png
文章结果图：
image.png
image.png
本文复现图：
image.png
image.png
文章数据下载
文章处理后的基因表达矩阵文件存放在GEO数据库中，检索号为 GSE182509，这里只提供了pkl格式的表达矩阵，需要用python的pickle包进行读取，我已将其转换为TSV文件存放在我的百度云盘中，有需要的可以下载使用。
链接：https://pan.baidu.com/s/1IoSIYoEfTzZarLWXWvgvzg
提取码：gkd9
image.png
加载示例数据
# 导入所需的python包
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import phate
import scprep
import magic
# 加载示例数据
chronic = scprep.io.load_tsv("GSM5530565_Chronic_LCMV_processed.matrix.txt")
acute = scprep.io.load_tsv("GSM5530564_Acute_LCMV_processed.matrix.txt")
tumor8 = scprep.io.load_tsv("GSM5530560_Lung_week8_processed.matrix.txt")
dln8 = scprep.io.load_tsv("GSM5530561_LN_week8_processed.matrix.txt")
tumor17 = scprep.io.load_tsv("GSM5530562_Lung_week17_processed.matrix.txt")
dln17 = scprep.io.load_tsv("GSM5530563_LN_week17_processed.matrix.txt")
# 查看示例数据
chronic.head()
# Mrpl15 (ENSMUSG00000033845) ... CAAA01147332.1 (ENSMUSG00000095742)
#0 ...
#AAACCTGAGAGTAAGG-1 0.0 ... 0.000000
#AAACCTGGTGTGAAAT-1 0.0 ... 0.000000
#AAACGGGAGGCTCATT-1 0.0 ... 0.000000
#AAACGGGGTAGCTTGT-1 0.0 ... 0.000000
#AAACGGGGTCGAATCT-1 0.0 ... 0.552073
#[5 rows x 11595 columns]
chronic.shape
#(1185, 11595)
acute.shape
#(10768, 12960)
tumor8.shape
#(806, 11749)
dln8.shape
#(1742, 12116)
tumor17.shape
#(731, 11150)
dln17.shape
#(876, 11595)
数据质控预处理和数据归一化
由于下载的数据是预先已经进行质控过滤和归一化处理的，这里将不再进行处理。详细用法见上期 使用PHATE进行单细胞高维数据的可视化
使用PHATE进行低维嵌入降维可视化
### analysis for chronic sample ###
#Embedding Data Using PHATE
# Instantiating the PHATE estimator
phate_operator = phate.PHATE(n_jobs=-2)
Y_phate = phate_operator.fit_transform(chronic)
Y_phate
#array([[ -5.34563999, 6.11670333],
# [ 29.8771147 , 8.33029219],
# [ 9.44218856, -25.94568946],
# ...,
# [-10.86402299, -1.15554 ],
# [-14.65615727, -1.03794057],
# [-16.53150258, 4.50993521]])
细胞聚类分群
# cell clustering
clusters = phate.cluster.kmeans(phate_operator, n_clusters=8)
clusters
#array([0, 2, 1, ..., 4, 4, 4], dtype=int32)
# 聚类结果可视化
scprep.plot.scatter2d(Y_phate, c=clusters, figsize=(8,7), cmap="Spectral",
ticks=False, label_prefix="PHATE", title= "Chronic LCMV")
plt.savefig("plot_phate_chronic_2d_by_cluster.png")
image.png
使用MAGIC进行数据填充
#Creating the MAGIC operator
magic_op = magic.MAGIC()
# Running MAGIC for all genes
chronic_magic = magic_op.fit_transform(chronic, genes='all_genes')
chronic_magic.head()
# Mrpl15 (ENSMUSG00000033845) ... CAAA01147332.1 (ENSMUSG00000095742)
#0 ...
#AAACCTGAGAGTAAGG-1 0.150924 ... 0.076723
#AAACCTGGTGTGAAAT-1 0.153466 ... 0.067584
#AAACGGGAGGCTCATT-1 0.146601 ... 0.064838
# rename columns names
chronic_magic.columns = [i.split(" ")[0] for i in chronic_magic.columns.tolist()]
# marker基因可视化
markers = ["Sell", "Ccr7", "Tcf7", "Slamf6", "Xcl1", "Il7r", "Eomes", "Tbx21", "Gzmb", "Prf1", "Pdcd1", "Havcr2", "Cd101", "Cx3cr1", "Cxcr6"]
for marker in markers:
# 2d plot
scprep.plot.scatter2d(Y_phate, c=chronic_magic[marker], figsize=(8,7), cmap="Reds",
ticks=False, label_prefix="PHATE", title=marker + " magic expression")
plt.savefig("plot_chronic_magic_marker_2d_" + marker + ".png")
image.png
根据这些marker基因的表达情况，我们将不同的cluster进行细胞类型的注释，得到以下的细胞注释结果。
Tnaive：Sell, Ccr7
Tsl: Tcf7, Slamf6, Xcl1
Ttrans: Cx3cr1, Cxcr6
Tex: Pdcd1, Havcr2, Cd101
image.png
多样本合并分析
### analysis for combined five sample ###
# Merge all datasets and create a vector representing the time point of each sample
alldata = [chronic,tumor8,tumor17,dln8,dln17]
EBT_counts, sample_labels = scprep.utils.combine_batches(
alldata,
["Chronic","Early Tumor","Late Tumor","Early LN","Late LN"],
append_to_cell_names=True
)
del alldata # removes objects from memory
EBT_counts.head()
# Mrpl15 (ENSMUSG00000033845) ... CAAA01147332.1 (ENSMUSG00000095742)
#AAACCTGAGAGTAAGG-1_Chronic 0.0 ... 0.000000
#AAACCTGGTGTGAAAT-1_Chronic 0.0 ... 0.000000
#AAACGGGAGGCTCATT-1_Chronic 0.0 ... 0.000000
#AAACGGGGTAGCTTGT-1_Chronic 0.0 ... 0.000000
#AAACGGGGTCGAATCT-1_Chronic 0.0 ... 0.552073
#[5 rows x 10246 columns]
EBT_counts.shape
#(5340, 10246)
sample_labels
#AAACCTGAGAGTAAGG-1_Chronic Chronic
#AAACCTGGTGTGAAAT-1_Chronic Chronic
#AAACGGGAGGCTCATT-1_Chronic Chronic
#AAACGGGGTAGCTTGT-1_Chronic Chronic
#AAACGGGGTCGAATCT-1_Chronic Chronic
# ...
#Name: sample_labels, Length: 5340, dtype: object
PHATE降维可视化
#Embedding Data Using PHATE
# Instantiating the PHATE estimator
phate_operator = phate.PHATE(n_jobs=-2)
Y_phate = phate_operator.fit_transform(EBT_counts)
Y_phate
#array([[ 79.05651647, 15.42592929],
# [ 20.72815444, 25.65566379],
# [ 90.57712893, -4.77917562],
# ...,
# [-52.39011592, 39.20142516],
# [-28.51731009, -8.66499775],
# [-46.00734805, -17.37265621]])
scprep.plot.scatter2d(Y_phate, c=sample_labels, figsize=(10,8), cmap="Spectral",
ticks=False, label_prefix="PHATE")
plt.savefig("plot_phate_2d_by_sample.png")
test4.png
# 3D visualization
phate_operator.set_params(n_components=3)
Y_phate_3d = phate_operator.transform()
Y_phate_3d
#array([[ 77.85894712, 13.21732854, -11.29708591],
# [ 17.23971363, 19.47621167, -27.54015408],
# [ 87.69788489, 4.0231593 , 13.16082805],
# ...,
# [-46.46850791, 42.68065699, -2.93836416],
# [-20.53357918, -4.9821892 , -22.70265013],
# [-36.18351133, -9.56494547, -29.67206442]])
scprep.plot.scatter3d(Y_phate_3d, c=sample_labels, figsize=(8,6), cmap="Spectral",
ticks=False, label_prefix="PHATE")
plt.savefig("plot_phate_3d_by_sample.png")
test5.png
细胞聚类分群
# cell clustering
clusters = phate.cluster.kmeans(phate_operator, n_clusters=12)
clusters
#array([8, 6, 1, ..., 2, 4, 4], dtype=int32)
# save meta data
meta = pd.merge(pd.DataFrame(sample_labels),pd.DataFrame(clusters,index=sample_labels.index,columns=["cluster"]),left_index=True,right_index=True)
meta.head()
# sample_labels cluster
#AAACCTGAGAGTAAGG-1_Chronic Chronic 8
#AAACCTGGTGTGAAAT-1_Chronic Chronic 6
#AAACGGGAGGCTCATT-1_Chronic Chronic 1
#AAACGGGGTAGCTTGT-1_Chronic Chronic 8
#AAACGGGGTCGAATCT-1_Chronic Chronic 8
meta.to_csv("metadata.csv")
scprep.plot.scatter3d(Y_phate_3d, c=clusters, figsize=(8,6), cmap="Spectral",
ticks=False, label_prefix="PHATE")
plt.savefig("plot_phate_3d_by_cluster.png")
test6.png
scprep.plot.scatter2d(Y_phate, c=clusters, figsize=(8,6), cmap="Spectral",
ticks=False, label_prefix="PHATE")
plt.savefig("plot_phate_2d_by_cluster.png")
test7.png
# to save as a gif:
scprep.plot.rotate_scatter3d(Y_phate_3d, c=sample_labels,
figsize=(8,6), cmap="Spectral",
ticks=False, label_prefix="PHATE", filename="phate.gif")
phate.gif
使用MAGIC进行数据填充
# rename columns names
EBT_counts.columns = [i.split(" ")[0] for i in EBT_counts.columns.tolist()]
# MAGIC imputation
#Creating the MAGIC operator
magic_op = magic.MAGIC()
#Running MAGIC with gene selection
#bmmsc_magic = magic_op.fit_transform(bmmsc_data, genes=["Mpo", "Klf1", "Ifitm1"])
#bmmsc_magic.head()
#Visualizing gene-gene relationships
#fig, (ax1, ax2) = plt.subplots(1,2, figsize=(16, 6))
#scprep.plot.scatter(x=bmmsc_data['Mpo'], y=bmmsc_data['Klf1'], c=bmmsc_data['Ifitm1'], ax=ax1,
# xlabel='Mpo', ylabel='Klf1', legend_title="Ifitm1", title='Before MAGIC')
#scprep.plot.scatter(x=bmmsc_magic['Mpo'], y=bmmsc_magic['Klf1'], c=bmmsc_magic['Ifitm1'], ax=ax2,
# xlabel='Mpo', ylabel='Klf1', legend_title="Ifitm1", title='After MAGIC')
#plt.tight_layout()
#plt.show()
# Running MAGIC for all genes
EBT_counts_magic = magic_op.fit_transform(EBT_counts, genes='all_genes')
EBT_counts_magic.head()
markers = ["Sell", "Ccr7", "Tcf7", "Slamf6", "Xcl1", "Il7r", "Eomes", "Tbx21", "Gzmb", "Prf1", "Pdcd1", "Havcr2", "Cd101", "Cx3cr1", "Cxcr6"]
for marker in markers:
# 2d plot
scprep.plot.scatter2d(Y_phate, c=EBT_counts_magic[marker], figsize=(8,6), cmap="Reds",
ticks=False, label_prefix="PHATE", title=marker + " magic expression")
plt.savefig("plot_magic_marker_2d_" + marker + ".png")
# 3d plot
scprep.plot.scatter3d(Y_phate_3d, c=EBT_counts_magic[marker], figsize=(8,6), cmap="Reds",
ticks=False, label_prefix="PHATE", title=marker + " magic expression")
plt.savefig("plot_magic_marker_3d_" + marker + ".png")
image.png
基因差异表达分析
# Differential analysis
# By samples
de_samples = scprep.stats.differential_expression_by_cluster(EBT_counts_magic,clusters=sample_labels,direction="up")
de_samples
de_samples["Chronic"]
for key,value in de_samples.items():
print(value.head(n=10))
# save DE data
de_samples["Chronic"].to_csv("DEs_Chronic.csv")
de_samples["Early Tumor"].to_csv("DEs_Early_Tumor.csv")
de_samples["Early LN"].to_csv("DEs_Early_LN.csv")
# By clusters
de_clusters = scprep.stats.differential_expression_by_cluster(EBT_counts_magic,clusters=clusters,direction="up")
de_clusters
de_clusters[0]
for key,value in de_clusters.items():
print(value.head(n=10))
