[ufwsse]

What do you mean by N50 of scaffold? The N50 after scaffolding runs?
It looks quite strange cause this http://seqanswers.com/forums/attachm...0&d=1331038646 clearly shows that your base quality is up to 24 with length of approx. 60nt. Thus trimming all reads to a length of 60nt all reads should pass your pre-processing filter. So my guess is that something in your filter works wrong.

Yes, The N50 after scaffolding.
the picture is filtered reads' property, handled with my own filter.
the original reads' property is as below.
Or perhaps I didn't get you meaning clearly.
Should I try to trim them to 60bp?
BTW: the http://seqanswers.com/forums/attachm...9&d=1331037505 is shown with a red cross, is anything wrong?

[kopi-o]

I'm not sure that you would expect better results, with such a small file (667M). How many reads is that equivalent to, and what average coverage (roughly) of the bacterial genome does that correspond to?

[sphil]

Yes, The N50 after scaffolding.
the picture is filtered reads' property, handled with my own filter.
the original reads' property is as below.
Or perhaps I didn't get you meaning clearly.
Should I try to trim them to 60bp?
BTW: the http://seqanswers.com/forums/attachm...9&d=1331037505 is shown with a red cross, is anything wrong?

ok if these are from the filtered one i also would suggest that your coverage is too low....

[ufwsse]

I'm not sure that you would expect better results, with such a small file (667M). How many reads is that equivalent to, and what average coverage (roughly) of the bacterial genome does that correspond to?

the length of reference is 2119959, there are 2 * 3418625 reads whose length is 76. so the average coverage is 2 * (3418625 * 76 / 2119959) = 2 * 122. Is it right?

[ufwsse]

problem solved with another assemble tool...
Thank all you guys' help!!!

[kopi-o]

Good to hear!

It would be nice to know which assembler that eventually solved the problem, and what kind of N50 etc you got ... sorry for being nosey, I am just curious!

[ufwsse]

Good to hear!

It would be nice to know which assembler that eventually solved the problem, and what kind of N50 etc you got ... sorry for being nosey, I am just curious!

errr.....shame on myself
I use the velvet and get a better output. It is a pitty that it can't set a longer kmer than 31.
Acctually, soapdenovo can solve this problem with a longer kmer, but I can't run it on my PC since it costs a lot of memory. When the k is small, the result is so bad that I think there must be something wrong with my pre-processing procedure.
I am sorry that I waste your precious time.

[kopi-o]

Thanks for your reply! You have not wasted my time at all, I think it's interesting to exchange experiences about different software packages, parameters, etc.