Clump gene entries for metagenomic data/ humann2

Question

[xinyuanzhang22]

Hi, I was doing some associations using gene abundance data from humann2 output, and got top results with hundreds of gene entries. By mapping them back to uniprot, I get the information on which species it come from and proteins they encode. Is there a way to clump all gene entries that encode the same proteins together? --for better annotation of the result.

A example of 4 gene entries from different organisms all encode for HTH cro/C1-type domain-containing protein. Thanks!

• R5DNF1 Parabacteroides johnsonii CAG:246
• R6K7I4 Eubacterium sp. CAG:252
• H1CLY0 Lachnospiraceae bacterium 7_1_58FAA
• D4C9T9 Clostridium sp. M62/1

click here for the screenshot from uniprot

Answer 1

[zorbax]

If you have the 'Protein names' and 'Organism' columns in one table you can use Pandas.

df.groupby('Protein names')['Organism'].agg(lambda col: ','.join(col)).reset_index()

For this table:

class   order
bird    Falconiformes
bird    Psittaciformes
mammal  Carnivora
mammal  Primates
mammal  Carnivora

You will get something like this:

class   order
bird    Falconiformes,Psittaciformes
mammal  Carnivora,Primates,Carnivora

[xinyuanzhang22]

Hi Thanks! Just that the protein names are usually not exactly the same, so it's a bit hard to do it for all proteins...