Hi,
We have a high quality, relatively small metagenomic dataset (from Miseq), with 4976646 reads that was processed by diamond (–sensitive option). After meganization it turned out that there are hundreds of non displayed Arenavirus (NP_694845) reads/hits which definitely present in
- original daa / blasttab file
- diamond NR database (latest)
- Megan prot_acc2tax-May2017.abin database
- meganized daa file (opened by text editor)
LCA parameters was set to extra sensitivity:
min score: 40
max expected: 0.01
min perc. identity: 0
top percent: 10
min. support percent: 1.0E-5
min support: 1
min complexity: 0
LCA algorithm: naive
read assigned mode: readcount
min perced read to cover: 0
With this parameters we able to detect as few as 1-3 reads in another samples.
Thanks for any idea: Blaize