Mapping multiple dataframe based on the matching columns

Question

I have 25 data frames which I need to merge and find recurrently occurring rows from all 25 data frames, For example, my data frame looks like following,

df1
chr start   end     name
1   12334   12334   AAA
1   2342    2342    SAP
2   3456    3456    SOS
3   4537    4537    ABR
df2
chr start   end     name
1   12334   12334   DSF
1   3421    3421    KSF
2   7689    7689    LUF
df3 
chr start   end     name
1   12334   12334   DSF
1   3421    3421    KSF
2   4537    4537    LUF
3   8976    8976    BAR
4   6789    6789    AIN

And In the end, I am aiming to have an output data frame like following,

chr start   end     name    Sample
1   12334   12334   AAA df1
1   12334   12334   AAA df2
1   12334   12334   AAA df3

I can get there with the following solution, By dictionary which adds all these three data frames into one bigger data frame dfs

dfs = {'df1': df1, 'df2': df2}

Then further,

common_tups = set.intersection(*[set(df[['chr', 'start', 'end']].drop_duplicates().apply(tuple, axis=1).values) for df in dfs.values()])
pd.concat([df[df[['chr', 'start', 'end']].apply(tuple, axis=1).isin(common_tups)].assign(Sample=name) for (name, df) in dfs.items()])

This gives out the resulting data frame with matching rows from all three data frames, but I have 25 data frames which I am calling as list from the directory as following,

path         = 'Fltered_vcfs/' 
files        = os.listdir(path)
results      = [os.path.join(path,i) for i in files if i.startswith('vcf_filtered')]

And so how can I show the list 'results' in the dictionary and proceed further to get the desired output. Any help or suggestions are greatly appreciated.

Thank you

Community · Accepted Answer · 2017-05-23 12:19:29Z

1

Using the glob module, you can use

import os
from glob import glob

path = 'Fltered_vcfs' 
f_names = glob(os.path.join(path, 'vcf_filtered*.*'))

Then, your dictionary can be created with dictionary comprehension using

import pandas as pd

 {os.path.splitext(os.path.split(f_name)[1])[0]: pd.read_csv(f_name,sep='\t') for f_name in f_names}

edited May 23, 2017 at 12:19

CommunityBot

11 silver badge

answered Sep 26, 2016 at 11:44

Ami Tavory

76.7k13 gold badges152 silver badges196 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

ARJ Over a year ago

There is a problem, my actual datafarmes has more columns, say like 18 for each of them.

Ami Tavory Over a year ago

@user1017373 Not sure what the problem is. If you'd like to load only a specific subset of the columns, you can use usecols=['start', 'end', 'name']. Is that your problem?

ARJ Over a year ago

No, loading whole columns and then merging based on matching rows from those three columns : [['chr', 'start', 'end']., the question is same, but the input slightly differs, and with the above solution its throwing index error message even after stripping the whitespace for all data frames as, KeyError: "['chr' 'start' 'end'] not in index"

Ami Tavory Over a year ago

@user1017373 But then it looks like they're just not in the index, no? So it doesn't look like your input matches the the specifications of your algorithm.

ARJ Over a year ago

Actually the name of the column is different from what I specify in the demo algorithm and that I have changed when I am using the same for real data frames. Like the real dfs have columns which I needed are as , chrom, start and end

|

Collectives™ on Stack Overflow

Mapping multiple dataframe based on the matching columns

1 Answer 1

7 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related