minPercentIdentity and lcaCoveragePercent

User Questions
1

I’m using MEGAN6 CE in the command line. I was wondering if I change the minPercentIdentity parameter, do I also need to change the lcaCoveragePercent parameter?

For example, I’m working on some DNA metabarcoding data for a diet study in ecology. The common parameter used for percent identity in other studies (not using MEGAN) seems to be 94% minimum percent identity to a reference sequence through BLAST. So I assume I should try to set my minPercentIdentity to 94. Do I then need to change the lcaCoveragePercent parameter too? I’ve seen that 80 is recommended for long reads, but my data isn’t long reads (all sequences <150 bp), and in another forum question I saw the developer say that 51% is ok too.

What impact would changing the lcaCoveragePercent parameter have on any classifications? And what would people recommend changing it to, if I should change it? Ideally I’d get down to Genus for most of my classifications, but I’d be happy with Family.

Thanks so much for your help, this is a great forum.

2

Hi @TPM,

The lcaCoveragePercent parameter doesn’t matter if you are performing just the naïve LCA, which I believe is the default for short reads. It only becomes relevant if you intend to use weighted LCA (short reads) or interval union LCA (typically designed for long reads).

Anupam

3

Hi Anupam,

Thanks for your reply. I was using weighted LCA. So what should I put in for this parameter if I’m using weighted?

Thanks

4

Hi @TPM,

Sorry for the delayed response. If I remember correctly, 80% is the recommended threshold for Weighted LCA. The 51% threshold was changed for Long Reads Interval Union LCA.

However, you can try tuning the threshold for Weighted LCA as needed. I assume the default is 80%.

Best,
Anupam