Benchmark

In [1]:

import pandas as pd
import numpy as np
import seaborn as  sns

import pandas as pd import numpy as np import seaborn as sns

Gene body¶

In [33]:

gb_bench_martrix = pd.read_csv('./feature_metrics_gene_body_v3.csv')

gb_bench_martrix = pd.read_csv('./feature_metrics_gene_body_v3.csv')

In [34]:

gb_bench_martrix.head()

gb_bench_martrix.head()

Out[34]:

	Unnamed: 0	ARI_Louvain	AMI_Louvain	Homogeneity_Louvain	ARI_leiden	AMI_leiden	Homogeneity_leiden	method	feature_number
0	0	NaN	NaN	NaN	0.228188	0.554318	0.600083	feature_select_random_	3000
1	1	NaN	NaN	NaN	0.229584	0.510317	0.558553	feature_select_var_	3000
2	2	NaN	NaN	NaN	0.439136	0.783196	0.834210	feature_select_sr_	3000
3	3	NaN	NaN	NaN	0.205842	0.386525	0.431316	feature_select_dispersion_	3000
4	4	NaN	NaN	NaN	0.424175	0.764100	0.817355	feature_select_residual_	3000

In [5]:

# df = df.rename(columns={'variable':'metric','value':'metric value'})
plot = sns.catplot(x='k',y='value',hue='gene_selection_via',data=df,kind='point',col='variable',)
for ax in plot.axes.flatten():
    for item in ax.get_xticklabels():
        item.set_rotation(45)
#plot.savefig('./figures/gene_body_feature_metric_v3_300_1000.pdf', dpi=300, format=None)

# df = df.rename(columns={'variable':'metric','value':'metric value'}) plot = sns.catplot(x='k',y='value',hue='gene_selection_via',data=df,kind='point',col='variable',) for ax in plot.axes.flatten(): for item in ax.get_xticklabels(): item.set_rotation(45) #plot.savefig('./figures/gene_body_feature_metric_v3_300_1000.pdf', dpi=300, format=None)

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[5], line 2
      1 # df = df.rename(columns={'variable':'metric','value':'metric value'})
----> 2 plot = sns.catplot(x='k',y='value',hue='gene_selection_via',data=df,kind='point',col='variable',)
      3 for ax in plot.axes.flatten():
      4     for item in ax.get_xticklabels():

NameError: name 'df' is not defined

In [35]:

df_metrics = gb_bench_martrix

df_metrics = gb_bench_martrix

In [36]:

def get_plot_df(res):
    gene_selection_via = []
    k = []
    dim_reduction_method = []
    ari = []
    ami = []
    homogeneity = []
    # 使用 DataFrame 的列名直接访问数据并添加到各自的列表
    for idx, row in res.iterrows():
        gene_selection_via.append(row['method'])
        k.append(row['feature_number'])
        ari.append(row['ARI_leiden'])
        ami.append(row['AMI_leiden'])
        homogeneity.append(row['Homogeneity_leiden'])

    data_dict = dict(gene_selection_via=gene_selection_via,
                     k=k,
                     ari = ari,
                     ami = ami,
                     homogeneity = homogeneity
                        )
    data = pd.DataFrame(data_dict)
    return data.melt(id_vars=['gene_selection_via','k'])

# 重置索引，将索引列转为普通列
df = df_metrics

df = get_plot_df(df_metrics)
df = df.reset_index(drop=True)

def get_plot_df(res): gene_selection_via = [] k = [] dim_reduction_method = [] ari = [] ami = [] homogeneity = [] # 使用 DataFrame 的列名直接访问数据并添加到各自的列表 for idx, row in res.iterrows(): gene_selection_via.append(row['method']) k.append(row['feature_number']) ari.append(row['ARI_leiden']) ami.append(row['AMI_leiden']) homogeneity.append(row['Homogeneity_leiden']) data_dict = dict(gene_selection_via=gene_selection_via, k=k, ari = ari, ami = ami, homogeneity = homogeneity ) data = pd.DataFrame(data_dict) return data.melt(id_vars=['gene_selection_via','k']) # 重置索引，将索引列转为普通列 df = df_metrics df = get_plot_df(df_metrics) df = df.reset_index(drop=True)

In [37]:

# 删除字符末尾的下划线
df['gene_selection_via'] = df['gene_selection_via'].str.rstrip('_')

# 将 'scm' 替换为 'scMethtools'
df['gene_selection_via'] = df['gene_selection_via'].str.replace('scm', 'scMethtools')

# 删除字符末尾的下划线 df['gene_selection_via'] = df['gene_selection_via'].str.rstrip('_') # 将 'scm' 替换为 'scMethtools' df['gene_selection_via'] = df['gene_selection_via'].str.replace('scm', 'scMethtools')

In [38]:

df

df

Out[38]:

	gene_selection_via	k	variable	value
0	feature_select_random	3000	ari	0.228188
1	feature_select_var	3000	ari	0.229584
2	feature_select_sr	3000	ari	0.439136
3	feature_select_dispersion	3000	ari	0.205842
4	feature_select_residual	3000	ari	0.424175
...	...	...	...	...
115	scMethtools_feature_select_random	28537	homogeneity	0.847398
116	scMethtools_feature_select_var	28537	homogeneity	0.847398
117	scMethtools_feature_select_sr	28537	homogeneity	0.847398
118	scMethtools_feature_select_dispersion	28537	homogeneity	0.847398
119	scMethtools_feature_select_residual	28537	homogeneity	0.847398

120 rows × 4 columns

In [40]:

method = ['feature_select_random','feature_select_var','feature_select_dispersion','scMethtools_feature_select_sr']

method = ['feature_select_random','feature_select_var','feature_select_dispersion','scMethtools_feature_select_sr']

In [85]:

filtered_df = df[df['gene_selection_via'].isin(method)]

filtered_df = df[df['gene_selection_via'].isin(method)]

In [86]:

filtered_df

filtered_df

Out[86]:

	gene_selection_via	k	variable	value
0	feature_select_random	3000	ari	0.228188
1	feature_select_var	3000	ari	0.229584
3	feature_select_dispersion	3000	ari	0.205842
5	feature_select_random	5000	ari	0.320749
6	feature_select_var	5000	ari	0.266127
8	feature_select_dispersion	5000	ari	0.228710
10	feature_select_random	10000	ari	0.387008
11	feature_select_var	10000	ari	0.352698
13	feature_select_dispersion	10000	ari	0.284312
15	feature_select_random	28537	ari	0.422763
16	feature_select_var	28537	ari	0.422763
18	feature_select_dispersion	28537	ari	0.422763
22	scMethtools_feature_select_sr	3000	ari	0.446674
27	scMethtools_feature_select_sr	5000	ari	0.438144
32	scMethtools_feature_select_sr	10000	ari	0.514701
37	scMethtools_feature_select_sr	28537	ari	0.493098
40	feature_select_random	3000	ami	0.554318
41	feature_select_var	3000	ami	0.510317
43	feature_select_dispersion	3000	ami	0.386525
45	feature_select_random	5000	ami	0.659198
46	feature_select_var	5000	ami	0.618409
48	feature_select_dispersion	5000	ami	0.516154
50	feature_select_random	10000	ami	0.720927
51	feature_select_var	10000	ami	0.716170
53	feature_select_dispersion	10000	ami	0.646381
55	feature_select_random	28537	ami	0.772037
56	feature_select_var	28537	ami	0.772037
58	feature_select_dispersion	28537	ami	0.772037
62	scMethtools_feature_select_sr	3000	ami	0.789079
67	scMethtools_feature_select_sr	5000	ami	0.788785
72	scMethtools_feature_select_sr	10000	ami	0.801440
77	scMethtools_feature_select_sr	28537	ami	0.801598
80	feature_select_random	3000	homogeneity	0.600083
81	feature_select_var	3000	homogeneity	0.558553
83	feature_select_dispersion	3000	homogeneity	0.431316
85	feature_select_random	5000	homogeneity	0.701129
86	feature_select_var	5000	homogeneity	0.666750
88	feature_select_dispersion	5000	homogeneity	0.554015
90	feature_select_random	10000	homogeneity	0.777818
91	feature_select_var	10000	homogeneity	0.772963
93	feature_select_dispersion	10000	homogeneity	0.702188
95	feature_select_random	28537	homogeneity	0.824734
96	feature_select_var	28537	homogeneity	0.824734
98	feature_select_dispersion	28537	homogeneity	0.824734
102	scMethtools_feature_select_sr	3000	homogeneity	0.836942
107	scMethtools_feature_select_sr	5000	homogeneity	0.837973
112	scMethtools_feature_select_sr	10000	homogeneity	0.844679
117	scMethtools_feature_select_sr	28537	homogeneity	0.847398

In [48]:

import matplotlib.pyplot as plt

import matplotlib.pyplot as plt

In [110]:

import scanpy  as sc

import scanpy as sc

In [114]:

sc.set_figure_params()
# 设置 Seaborn 样式为白色背景
sns.set_style('white')

sc.set_figure_params() # 设置 Seaborn 样式为白色背景 sns.set_style('white')

In [117]:

plot = sns.catplot(x='k',y='value',hue='gene_selection_via',data=filtered_df,kind='point',col='variable')

plot.legend.remove()  
plot.add_legend(title='Gene Selection Method', title_fontsize='14', fontsize='14', ncol=4, loc='lower center')
# 设置各个子图标题的字体大小
for ax in plot.axes.flat:
    ax.set_title(ax.get_title(), fontsize=16)

# 调整图形布局以确保图例不会被遮挡
plt.subplots_adjust(bottom=0.25)

plot.savefig('./figures/gene_body_feature_metric_v5_300_all.pdf', dpi=300, format=None)

plot = sns.catplot(x='k',y='value',hue='gene_selection_via',data=filtered_df,kind='point',col='variable') plot.legend.remove() plot.add_legend(title='Gene Selection Method', title_fontsize='14', fontsize='14', ncol=4, loc='lower center') # 设置各个子图标题的字体大小 for ax in plot.axes.flat: ax.set_title(ax.get_title(), fontsize=16) # 调整图形布局以确保图例不会被遮挡 plt.subplots_adjust(bottom=0.25) plot.savefig('./figures/gene_body_feature_metric_v5_300_all.pdf', dpi=300, format=None)

下采样结果¶

In [118]:

downsample_bench_martrix = pd.read_csv('./downsample_feature_metrics_v1.csv')

downsample_bench_martrix = pd.read_csv('./downsample_feature_metrics_v1.csv')

In [120]:

downsample_bench_martrix.head()

downsample_bench_martrix.head()

Out[120]:

	Unnamed: 0	Sample	Feature	ARI	AMI	Homogeneity
0	0	10	1000	0.376579	0.660696	0.752957
1	1	10	3000	0.346674	0.659011	0.754903
2	2	10	5000	0.355119	0.676024	0.788836
3	3	10	10000	0.401669	0.695064	0.805697
4	4	10	28537	0.394273	0.702486	0.808378

In [121]:

def get_plot_df(res):
    sample_num = []
    k = []
    dim_reduction_method = []
    ari = []
    ami = []
    homogeneity = []
    # 使用 DataFrame 的列名直接访问数据并添加到各自的列表
    for idx, row in res.iterrows():
        sample_num.append(row['Sample'])
        k.append(row['Feature'])
        ari.append(row['ARI'])
        ami.append(row['AMI'])
        homogeneity.append(row['Homogeneity'])

    data_dict = dict(sample_num=sample_num,
                     k=k,
                     ari = ari,
                     ami = ami,
                     homogeneity = homogeneity
                        )
    data = pd.DataFrame(data_dict)
    return data.melt(id_vars=['sample_num','k'])

# 重置索引，将索引列转为普通列
df = downsample_bench_martrix

df = get_plot_df(downsample_bench_martrix)
df = df.reset_index(drop=True)

def get_plot_df(res): sample_num = [] k = [] dim_reduction_method = [] ari = [] ami = [] homogeneity = [] # 使用 DataFrame 的列名直接访问数据并添加到各自的列表 for idx, row in res.iterrows(): sample_num.append(row['Sample']) k.append(row['Feature']) ari.append(row['ARI']) ami.append(row['AMI']) homogeneity.append(row['Homogeneity']) data_dict = dict(sample_num=sample_num, k=k, ari = ari, ami = ami, homogeneity = homogeneity ) data = pd.DataFrame(data_dict) return data.melt(id_vars=['sample_num','k']) # 重置索引，将索引列转为普通列 df = downsample_bench_martrix df = get_plot_df(downsample_bench_martrix) df = df.reset_index(drop=True)

In [126]:

# 将 'k' 列转换为整数类型
df['k'] = df['k'].astype(int)

# 将 'k' 列转换为整数类型 df['k'] = df['k'].astype(int)

In [134]:

# 确保列 'k' 是浮点数类型
df['sample_num'] = df['sample_num'].astype(float)

# 将列 'k' 转换为百分数
df['k_percentage'] = df['sample_num'] * 100

# 确保列 'k' 是浮点数类型 df['sample_num'] = df['sample_num'].astype(float) # 将列 'k' 转换为百分数 df['k_percentage'] = df['sample_num'] * 100

In [139]:

# 创建映射字典
mapping = {
    10.0: '10% (n=231)',
    30.0: '30% (n=693)',
    50.0: '50% (n=1156)',
    70.0: '70% (n=1619)',
    90.0: '90% (n=2081)'
}
# 根据 sample 列的值匹配 detail 列的内容
df['detail'] = df['sample_num'].map(mapping)

# 创建映射字典 mapping = { 10.0: '10% (n=231)', 30.0: '30% (n=693)', 50.0: '50% (n=1156)', 70.0: '70% (n=1619)', 90.0: '90% (n=2081)' } # 根据 sample 列的值匹配 detail 列的内容 df['detail'] = df['sample_num'].map(mapping)

In [140]:

df

df

Out[140]:

	sample_num	k	variable	value	k_percentage	detail
0	10.0	1000	ari	0.376579	1000.0	10% (n=231)
1	10.0	3000	ari	0.346674	1000.0	10% (n=231)
2	10.0	5000	ari	0.355119	1000.0	10% (n=231)
3	10.0	10000	ari	0.401669	1000.0	10% (n=231)
4	10.0	28537	ari	0.394273	1000.0	10% (n=231)
...	...	...	...	...	...	...
70	90.0	1000	homogeneity	0.831687	9000.0	90% (n=2081)
71	90.0	3000	homogeneity	0.834220	9000.0	90% (n=2081)
72	90.0	5000	homogeneity	0.840011	9000.0	90% (n=2081)
73	90.0	10000	homogeneity	0.839620	9000.0	90% (n=2081)
74	90.0	28537	homogeneity	0.846224	9000.0	90% (n=2081)

75 rows × 6 columns

In [141]:

plot = sns.catplot(x='k',y='value',hue='detail',data=df,kind='point',col='variable')

plot.legend.remove()  
plot.add_legend(title='Sample fraction', title_fontsize='14', fontsize='14', ncol=5, loc='lower center')
# 设置各个子图标题的字体大小
for ax in plot.axes.flat:
    ax.set_title(ax.get_title(), fontsize=16)

# 调整图形布局以确保图例不会被遮挡
plt.subplots_adjust(bottom=0.25)

plot.savefig('./figures/downsample_feature_metric_all.pdf', dpi=300, format=None)

plot = sns.catplot(x='k',y='value',hue='detail',data=df,kind='point',col='variable') plot.legend.remove() plot.add_legend(title='Sample fraction', title_fontsize='14', fontsize='14', ncol=5, loc='lower center') # 设置各个子图标题的字体大小 for ax in plot.axes.flat: ax.set_title(ax.get_title(), fontsize=16) # 调整图形布局以确保图例不会被遮挡 plt.subplots_adjust(bottom=0.25) plot.savefig('./figures/downsample_feature_metric_all.pdf', dpi=300, format=None)

In [142]:

df.to_csv('downsample_feature_metrics_v2.csv')

df.to_csv('downsample_feature_metrics_v2.csv')

In [ ]: