MEGAN - how many BLAST hits to keep for each?

[camilaz]

Hi

I'm wondering if I could get some advice for experts on this forum. I am working with the Illumina metagenomic dataset and using MEGAN to determine the taxonomic diversity.

In summary I have done this:
- Assembled the reads into contigs (82.5% of reads aligned in contigs >500 bp)
- Predicted CDS on contigs >500 bp
- Run BLASTP on CDSs against nr NCBI, only 1 blast hits for each gene was kept
- Uploaded BLASTP results to MEGAN for analysis

My understanding is that MEGAN needs more than one blast hit for each query for the LCA algorithm to be properly working and since I have only 1 hit for each query, the sequences are not assigned to taxon using LCA. My question is: is the method I used for taxonomic classification correct? I am at the stage of writing manuscript and worry that I did the analysis wrong.

Thank you in advance for all answers, advices and criticism!

Camila

[salamay]

Are you manually setting the number of hits returned to 1 or do all your sequences that have hits actually only return 1 hit?
In any case, I believe going into options->change LCA paramters, and setting Min Support to 1 would assign the reads to the corresponding nodes.

[camilaz]

I am using -max_target_seqs 1 in the BLAST command line. But my question was not - how to set 1 hit per sequence in MEGAN - but - is one hit per sequence good enough for sequence classification? Perhaps I should have set -max_target_seqs option to 10 or 100?

And I apologise for the mistake in title should be "MEGAN - how many BLAST hits to keep for each query?"

[salamay]

Well I don't think it is optimal to provide only one hit when using a lowest common ancestor algorithm which MEGAN uses so I would probably not use -max_target_seqs 1 and just let it get as many hits as possible and let the LCA algorithm deal with it.