Hey Katya!

I would recommend using the "bbmap" tool included in the "BBTools" suite for that task.

For an introduction to the "bbmap" tool: 

http://seqanswers.com/forums/showthread.php?t=41057

For a description of how to use bbmap to remove human contaminant DNA (also includes an already processed reference genome available for download):

http://seqanswers.com/forums/showthread.php?t=42552

You could try to map everything on a human genome and then take what is left over. 

Some people BLAST the reads against the human genome as this algorithm is faster than mapping directly.

Thank you for your answers. i was blasting against human genome miseq data and it was taking long time. Now with hiseq, i hoped there was a faster solution as number of reads is much higher. 

thanks, Artur. i downloaded the human database from ncbi and blasted my fragments against it on our server. which software would you recommend for mapping against the human genome?

I would try 2 ways (either bwa/bowtie) map to a single index of human genome and get rid of those reds, or map to joined index of human + whatever the bacterial was and remove the MAPq=0 reads (note however that here you should look for the reads, where the supplementary mapping was to human, as MAPQ=0 is generated when equally likely mapping (that means more than 1 mapping position) is reported for the read)

(I would advise against blasting. BWA/Bowtie where developed for that manner to perform much better than blast)

Hey Katya!

I would recommend using the "bbmap" tool included in the "BBTools" suite for that task.

For an introduction to the "bbmap" tool: 

http://seqanswers.com/forums/showthread.php?t=41057

For a description of how to use bbmap to remove human contaminant DNA (also includes an already processed reference genome available for download):

http://seqanswers.com/forums/showthread.php?t=42552

 In my experience, mapping with bowtie2 is much faster than using BLAST. I downloaded and use the hg19 human reference db from http://support.illumina.com/sequencing/sequencing_software/igenome.html. I still don't have a particularly fast way to split up the original fastq files based on the mapping results, but the standard mapping output can be parsed to generate a fasta file with the original definition lines and sequences of the reads that do NOT map to the genome (they will have a "*" rather than a reference match). 

If you simply want to get mapped or unmapped you can go with samtools view -f4/-F4 arguments, and to get by mapping quality (not the read mean phred score) just run samtools with -q 20 for example. Those are most basic flags, if you were to use newer versions of BWA it has flags dealing with supplementary/primary/secondary aligments (in the essence they allow you to quickly get unique/non-unique mappings.To put it simply, anything with MAPQ =0 is non-unique)

Hey Katya,

I am also wondering that which is the efficient way of removing human dna reads from metagenome. Which tool did you finally use to remove the human DNA data from the metagenome?

Hei Sajan,

i followed Cyril's recommendations and used BBMap tool, taking an additional filtering step after read clustering; i then blasted all of them against ncbi human genome database and removed those that matched (it was not many, but there were still some). 

Thank you Ekaterina. 

There are two ways which I consider proper for that. The first is by mapping every read against the ref genome, by using blast However, before that, you should merge your fastq files into a single one and the use seqtk to convert from fastq to fasta, easily, and then perform blast, and sort by hits that are 95% identical to the reference (I use awk commands to perform that), and then create a list from those positive hits, and extract them from the read files. The second is by using BWA/BBMAP/BOWTIE2 tools, and after that, using BamToFastq to generate just one fastq file with your mapped read files.