Thanks. Yes, the ability to identify repetitive regions would be bad. So is it best to just arbitrarily remove reads pre-assembly? From: mira_talk-bounce@xxxxxxxxxxxxx [mailto:mira_talk-bounce@xxxxxxxxxxxxx] On Behalf Of Robert Bruccoleri Sent: Thursday, April 21, 2011 3:08 PM To: mira_talk@xxxxxxxxxxxxx Subject: [mira_talk] Re: reducing number of Illumina reads However, there is a problem with taking this approach -- you could significantly change the statistics for the determination of repetitive regions and cause misassemblies as a result. Mira will generate coverage equivalent reads which can help in this situation. --Bob ALLO (Alfredo Lopez De Leon) wrote: You can try to collapse your reads with the FASTX toolkit this will leave you with a set made of unique reads. This method will preserve the sequence coverage and remove the redundancy. http://hannonlab.cshl.edu/fastx_toolkit/ AlLo From: mira_talk-bounce@xxxxxxxxxxxxx [mailto:mira_talk-bounce@xxxxxxxxxxxxx] On Behalf Of Goldman, Thomas Sent: Thursday, April 21, 2011 2:27 PM To: mira_talk@xxxxxxxxxxxxx Subject: [mira_talk] reducing number of Illumina reads Hello all, I have a 24GB RHEL5 machine on which I was able to do a de novo assembly of 454 paired-end and fragment reads (~1.5 million reads). I also have about 6 million 36bp Illumina reads. I would like to: Map the Illumina reads to the 454 backbone Include the Illumina reads with the 454 reads for a de novo assembly But I believe I don't have enough memory to handle all the Illumina reads. I think my VM could handle maybe 20% of the Illumina reads. What is the best way to reduce the Illumina reads used for the mapping and/or the de novo assemblies? Would it be to just randomly pick 20% of the reads out of the fastq file? Is there a tool out there I could use for this? Thanks, Tom