堆栈后删除多个值答案

【问题标题】：Remove multiple values after stack堆栈后删除多个值
【发布时间】：2019-05-11 12:50:48
【问题描述】：

我有这样的方阵。

            ACSM3     ACSX12    ADXM28  ...   UGT2B15      VCAN        XK
ACSM3    1.000000  0.929347  0.999914  ...  0.986433  0.999947 -0.999680
ACSX12    0.929347  1.000000  0.924428  ...  0.977350  0.925496 -0.919704
ADXM28   0.999914  0.924428  1.000000  ...  0.984196  0.999996 -0.999925
ADAM28   0.999976  0.926774  0.999981  ...  0.985275  0.999994 -0.999831
ADH1B   -0.999509 -0.917317 -0.999834  ... -0.980802 -0.999778  0.999982
ADTRP   -0.999039 -0.912273 -0.999528  ... -0.978290 -0.999438  0.999828
AEBP1    0.983312  0.846668  0.985611  ...  0.940104  0.985133 -0.987601
AKR1B10 -0.999658 -0.919371 -0.999915  ... -0.981800 -0.999874  1.000000
UBL3     0.997347  0.900002  0.998215  ...  0.971864  0.998043 -0.998870
UGT2B15  0.986433  0.977350  0.984196  ...  1.000000  0.984690 -0.981961
VCAN     0.999947  0.925496  0.999996  ...  0.984690  1.000000 -0.999887
XK      -0.999680 -0.919704 -0.999925  ... -0.981961 -0.999887  1.000000

使用堆栈函数后，我将数据变成我想要的形状，但正如您所见，由于相互比较，所有数据都有多个值。

dfHealty = df_healtyWithGenes.stack().reset_index()
dfHealty.columns = ['gene1', 'gene2', 'score']
dfHealty = dfHealty[dfHealty.gene1 != dfHealty.gene2]

我可以按分数过滤，但这不是个好主意，数据可能会损坏。

如何按基因列过滤？

gene1   gene2   score
EPB41L4B PGC    0.496713249 
PGC EPB41L4B    0.496713249 
CHGA    MT1G    0.496751983
MT1G    CHGA    0.496751983
AEBP1   FCER1G  0.497061368 
FCER1G  AEBP1   0.497061368 
ADTRP   CAPN9   0.497122603
CAPN9   ADTRP   0.497122603
FAM189A2 GLUL   0.49721763
GLUL FAM189A2 0.49721763
CA9 DUOX1   0.497233294
DUOX1   CA9 0.497233294
EDNRA   MSLN    0.497267565
MSLN    EDNRA   0.497267565
HRASLS2 LIPF    0.497581499
LIPF    HRASLS2 0.497581499
EPB41L4B    NEDD4L  0.497613643
NEDD4L  EPB41L4B    0.497613643

我需要像这样转换数据。

gene1   gene2   score
EPB41L4B PGC    0.496713249 
CHGA    MT1G    0.496751983
AEBP1   FCER1G  0.497061368 
ADTRP   CAPN9   0.497122603
FAM189A2 GLUL   0.49721763
CA9 DUOX1   0.497233294
EDNRA   MSLN    0.497267565

【问题讨论】：

标签： python pandas dataframe matrix

【解决方案1】：

使用给定的数据，您可以像这样删除数据中的重复对

import pandas as pd

cols = ['gene1','gene2','score']
data = [['EPB41L4B', 'PGC',0.496713249], 
        ['PGC','EPB41L4B',0.496713249], 
        ['CHGA','MT1G',0.496751983],
        ['MT1G','CHGA',0.496751983],
        ['AEBP1','FCER1G',0.497061368 ],
        ['FCER1G','AEBP1',0.497061368], 
        ['ADTRP','CAPN9',0.497122603],
        ['CAPN9','ADTRP',0.497122603],
        ['FAM189A2','GLUL',0.49721763],
        ['GLUL','FAM189A2',0.49721763],
        ['CA9','DUOX1',0.497233294],
        ['DUOX1','CA9',0.497233294],
        ['EDNRA','MSLN',0.497267565],
        ['MSLN','EDNRA',0.497267565],
        ['HRASLS2','LIPF',0.497581499],
        ['LIPF','HRASLS2',0.497581499],
        ['EPB41L4B','NEDD4L',0.497613643],
        ['NEDD4L','EPB41L4B',0.497613643]]

df = pd.DataFrame(data,columns=cols)
df = df[df['gene1'] < df['gene2']]
print(df)

产生这样的输出

       gene1   gene2     score
0   EPB41L4B     PGC  0.496713
2       CHGA    MT1G  0.496752
4      AEBP1  FCER1G  0.497061
6      ADTRP   CAPN9  0.497123
8   FAM189A2    GLUL  0.497218
10       CA9   DUOX1  0.497233
12     EDNRA    MSLN  0.497268
14   HRASLS2    LIPF  0.497581
16  EPB41L4B  NEDD4L  0.497614

【讨论】：

【解决方案2】：

在我看来，你可以每隔一排。

print(df.iloc[::2].reset_index(drop=True))

输出：

      gene1   gene2        score
0  EPB41L4B     PGC  0.496713249
1      CHGA    MT1G  0.496751983
2     AEBP1  FCER1G  0.497061368
3     ADTRP   CAPN9  0.497122603
4  FAM189A2    GLUL   0.49721763
5       CA9   DUOX1  0.497233294
6     EDNRA    MSLN  0.497267565
7   HRASLS2    LIPF  0.497581499
8  EPB41L4B  NEDD4L  0.497613643

您还可以使用frozenset 过滤掉重复项：

without_dupes = {frozenset([first, second]): score for first, second, score in df.values}.items()

result = pd.DataFrame([(*k, v) for k, v in without_dupes], columns=['gene1', 'gene2', 'score'])
print(result)

输出：

     gene1     gene2        score
0       PGC  EPB41L4B  0.496713249
1      CHGA      MT1G  0.496751983
2    FCER1G     AEBP1  0.497061368
3     ADTRP     CAPN9  0.497122603
4  FAM189A2      GLUL   0.49721763
5     DUOX1       CA9  0.497233294
6     EDNRA      MSLN  0.497267565
7   HRASLS2      LIPF  0.497581499
8    NEDD4L  EPB41L4B  0.497613643

【讨论】：

【解决方案3】：

我认为你可以这样做。这是您尝试做的一个最小示例：

import pandas as pd
import numpy as np

# sample data frame
df = pd.DataFrame({'col1': ['a','b'], 'col2':['b','a'], 'col3':[1,1]})

   col1 col2  col3
0    a    b     1
1    b    a     1

# take first two columns from where to remove duplicates
df2 = df.iloc[:,:2]

# sort the columns based on their corresponding values and create a new df 
df3 = pd.DataFrame(np.sort(df2.values), axis=1), df2.index, df2.columns)

# finally drop duplicates
result = pd.concat([df3, df['col3']], axis=1).drop_duplicates(subset=['col1','col2'])

  col1 col2  col3
0    a    b     1

【讨论】：

@tyasird 看起来我错误地删除了一些代码。现已修复。