more than 73% is unclassified.

[jojyjohn28]

I am running Kaiju classification for my Biofilm based meta-genome reads and I used Nr-euk database. However Kaiju yielded less than 25 % classification at phylum level. Is there any ways to increase the percentage of classification?

[pmenzel]

That's indeed quite low. Do you run it with default parameters? You can of course always reduce thresholds for match score (-s) or E-value (-E), which will also increase false positive classifications. Are your reads short or bad quality? I suggest to blastx some of the reads by hand to the nr database using the NCBI blast web server and see what hits you get and how the alignments look like. If you get good hits, but alignments have a lot of mismatches, then kaiju is probably not able to find an alignment and a more sensitive method might be the better choice.

[jojyjohn28]

I ran it with default parameters. I have high quality reads (average phred score is 36).
I did a blast X as you suggested for few representative reads, it showed good alignment but the percentage similarity is low, almost 56 to 60%.
I will try reducing the thresholds for match score and will update you soon. let me know if you have any other advice for me.