计算 pd.DataFrame 中条目值的差异比较多个索引

Question

我想从包含多个碱基（A、C、G、T、U¹ 和 U 的数据的 pd.DataFrame 中插入每两个能量值差异的差异²) 用于每个环境中的每个基地，依此类推。我想到了类似fhe的东西。我想过遍历整个数据框，但它没有用，也绝对不是 pd.DataFrames 的用途。
例如，我希望在其“中性”和“正”值之间以 pbs 为单位负责基础 A 的差异，从而扩展每个基础、环境和更多标准的原则，这些标准可能会在未来添加。

虽然我听说过pandas.MultiIndex，但我没有太多经验，但这对我来说似乎是一个可能的解决方案，不是吗？

import seaborn as sns
import pandas as pd
import os

with open (os.path.join (os.environ ['HOME'], 'data.csv'), 'r') as f :
    df = pd.read_csv (f, index_col = 0, header = 0, thousands = None, decimal = '.')

df = df.loc [df ['base'].isin (['A', 'C', 'G', 'U'])]

# one way I tried
pos = df.loc [df ['charge'] == 'pos']
neg = df.loc [df ['charge'] == 'neg']
neu = df.loc [df ['charge'] == 'neu']

df.loc [df ['charge'] == 'neg', 'difference'] = neg ['energy'] - neu ['energy']
df.loc [df ['charge'] == 'pos', 'difference'] = pos ['energy'] - neu ['energy']

# another way I tried
for posneg in ['pos', 'neg'] :
    df.loc [df ['charge'] == posneg, 'difference'] = lit.query ("(charge == 'neu') - (charge == '@posneg')")

~/data.csv：

,environment,base,charge,energy,type
0,pbs,A,neg,0.34835,1
1,pbs,C,neg,0.40194,2
2,pbs,G,neg,0.34959,1
3,pbs,T,neg,0.40738,2
4,pbs,U1,neg,0.34904,2
5,pbs,U2,neg,0.40016,2
6,pbs,A,neu,0.40151,3
7,pbs,C,neu,0.34494,3
8,pbs,G,neu,0.40193,3
9,pbs,T,neu,0.34458,3
10,pbs,U1,neu,0.34646,3
11,pbs,U2,neu,0.40871,3
12,pbs,A,pos,0.34047,2
13,pbs,C,pos,0.40157,2
14,pbs,G,pos,0.34232,2
15,pbs,T,pos,0.40854,2
16,pbs,U1,pos,0.34611,2
17,pbs,U2,pos,0.34414,2
18,polymeric,A,neg,0.28333,2
19,polymeric,C,neg,0.46908,3
20,polymeric,G,neg,0.33224,3
21,polymeric,T,neg,0.35825,1
22,polymeric,U1,neg,0.33033,3
23,polymeric,U2,neg,0.39167,3
24,polymeric,A,neu,0.36964,2
25,polymeric,C,neu,0.33979,2
26,polymeric,G,neu,0.41815,3
27,polymeric,T,neu,0.30786,2
28,polymeric,U1,neu,0.40727,1
29,polymeric,U2,neu,0.36719,3
30,polymeric,A,pos,0.38173,1
31,polymeric,C,pos,0.35060,3
32,polymeric,G,pos,0.37617,1
33,polymeric,T,pos,0.44172,2
34,polymeric,U1,pos,0.31267,3
35,polymeric,U2,pos,0.34478,2

Answer 1

代码中只反映了第一个减法，因为不清楚有多少计算。我认为遵循这种格式很容易。

工艺流程：

1. 数据框变换
2. 消除多个索引。
3. 更改列名
4. 计算 A，计算 B

# Extract the condition
df = df.loc[df['base'].isin (['A','C','G','U'])]
# long to wide
df = df.set_index(['environment','base','charge']).unstack()
# Eliminate multi-indexing
df.columns = df.columns.droplevel()
# Eliminate indexing
df = df.rename_axis(columns=None).reset_index()
# rename columns
df.columns = ['environment','base','neg','neu','pos','neg_type','neu_type','pos_type']
df['neg-neu'] = df['neg']-df['neu']
df['pos-neu'] = df['pos']-df['neu']
df
    environment base    neg neu pos neg_type    neu_type    pos_type    neg-neu pos-neu
0   pbs A   0.34835 0.40151 0.34047 1   3   2   -0.05316    -0.06104
1   pbs C   0.40194 0.34494 0.40157 2   3   2   0.05700 0.05663
2   pbs G   0.34959 0.40193 0.34232 1   3   2   -0.05234    -0.05961
3   polymeric   A   0.28333 0.36964 0.38173 2   2   1   -0.08631    0.01209
4   polymeric   C   0.46908 0.33979 0.35060 3   2   3   0.12929 0.01081
5   polymeric   G   0.33224 0.41815 0.37617 3   3   1   -0.08591    -0.04198