Differences between kos.parsed.counts.tsv files and my own script for summing counts per KEGG KO

[Dries-B]

I'm finding differences between kos.parsed.counts.tsv (in results/annotation/$SAMPLE/) and the results of my own script for summing counts annotated per KEGG KO. My script revolves around the following (pandas) lines:

kegg_ko_sr = merged_df['KEGG_ko'].str.replace('ko:','').str.split(',')
kegg_ko_df = merged_df.assign(KEGG_ko=kegg_ko_sr).explode('KEGG_ko')

group_kegg_ko = (kegg_ko_df.groupby('KEGG_ko', as_index = False)
                            [['counts','length']]
                              .agg('sum')
                    )

Do you know what could be causing the difference? Is there any additional functionality in the code creating kos.parsed.counts.tsv which could be the cause?

[Dries-B]

Many differences seem to be caused by duplicate KOs. For example, the result of my script only has 24 counts for both K11191 and K11192, while yours only includes the latter.

On the other hand, K11189 (with 302 counts) is only included in 'my' result. Also, the number of Unclassified reads differs between the result files.