FastQC kmer content output

Question

[Maria]

I have never seen FastQC reports where more than 20 Kmers are flagged or reported as failed. The manual says that "Any Kmers with positionally biased enrichment are reported." Has anyone seen more than 20 reported Kmers and if not, are there any suggestions as why this is happening?

Thanks!

Answer 1

[Istvan Albert]

It is not the number of k-mers that matter but their counts. The normalization method (and the corresponding plot) is misleading and frankly, just about useless.

Do also note that a longer k-mer will usually be reported as multiple shorter ones, each shifted a base from its original starting position. This is a limitation on how long the detected kmers can be. Hence a long k-mer will be reported many times over with different sub-kmers.