KO-length normalization

[benoitdepins]

Hi all!

First, thanks for FunProfiler. Such a great tool.

I would like to aggregate multiple KOs (from the same pathway) into a single pathway-level value per metagenome (e.g., mean), then compare these pathway values across metagenomes. Intuitively, it seems I should normalize each KO’s abundance by its “length” first (longer KOs might have more chances to get matches than shorter ones).

My questions:

• Are FunProfiler’s KO “abundances” length-normalized in any way (e.g., by KO sequence length or KO k-mer cardinality)?

• If not, would you recommend normalizing by KO length before aggregating KOs within a pathway? (and if so, maybe normalizing another output/metric of FunProfiler instead of your calculated relative abundance? Like the intersect_bp value from prefetch_out?)

• Or do you see any other caveats or limitations to this idea when using FunProfiler?

Thank you very much

[mahmudhera]

Hi, thanks for your interests in the pipeline.

Yes, the abundances are length-normalized. We achieve this using k-mer cardinality. In other words, if a KO contains fewer k-mers, then a smaller number of matching k-mers in the query is sufficient for it to be considered equally abundant compared to a KO with more k-mers.

Let me know if this answers your question.