Relationship between kmer size and read length and total no.of reads. ?

[vaibhavvsk]

To start the assembly process we have to set ther kmer size.I'm not clear about the relationship between kmer size,read length and total no.of reads so that i can decide proper kmer size.

[jgibbons1]

Hi vaibhavvsk,
What is your read size? Is this paired end or single end? What kind of data is this (Illumina, 454, pacbio etc.)?

You may find this blog post helpful:

http://www.homolog.us/blogs/2012/10/10/multi-kmer-de-bruijn-graphs/

[krobison]

A lot of this must be worked out empirically, though you can use others' experience as a guide. kmer is always going to be shorter than your read length, and probably quite a bit shorter. For a lot of Illumina data and many assemblers, small k (21 or 31, for example) work well as a first cut, and then it is worth trying larger values

Also note that most assemblers require odd k values (so that no kmer is palindromic), so if you are scanning a range don't bother with evens.

Also note that some assemblers (e.g. Ray) have a default maximum on the kmer size which isn't very large; you need to set a parameter during the build to increase this. SOAPdenovo comes in various flavors compiled for different maxima; the memory requirements for all assemblies are greater if you use one of the binaries that can run a higher k.