velvet_ begginer

[mmmm]

I have started using velvet for denovo assembly-

$velveth output/ 31 -fastq -shortPaired -separate read1.fastq read2.fastq

$velevtg output/ -cov_cutoff 0 -exp_cov 0 -unused_reads yes -very_clean yes

(using these parametes, I got >6000 nodes at n50 of 325

I think I should adjust (-cov_cutoff? and -exp_cov?) in order to get good assembly?- what are the right paprametrs to try please?- this a bacterial genome

[mastal]

In the first instance you should set -exp_cov auto, and velvet will calculate
the exp cov. You could also set -cov_cutoff auto, but by default this is set to half the -exp_cov value.

You need to read the velvet manual, it gives a formula for calculating the coverage in kmers, which is what velvet uses, and that will give you an idea of how much coverage you have from your data.

In order for velvet to assemble your reads as paired reads, you also need to set the -ins_length parameter. If you don't have an estimate for the insert length from the sequence provider, there is a script in the contrib directory of velvet that will calculate a value for the insert length.

You also need to do various preliminary runs changing the kmer length, to see what kmer length gives you the best assembly. Again, see the velvet manual for more details.

[mmmm]

are these velvet output good for a bacterial genome (5MB):
633 nodes, N50 38471, max 96792, total 4872056, using 760787/765644

I have set "-exp_cov and cov_cutoff" to auto

in the stats.txt, nodes coverage range greatly from 20x to 90x- so I am not sure, which to use?

what is the best N50, I should get?

[mastal]

How long are your reads?

What kmer size did you use?

One of the first things to do with velvet is to vary the kmer size, and see what size gives the best N50.

You may need to recompile velvet so that it will use kmer lengths longer than 31.

[mmmm]

thanks mastal-
How long are your reads?
reads are 250PE (but they are combined from 2 different runs)- as the first run has very big library size and gave low cluster density
What kmer size did you use?
31 (I will use 31,45,2 to check the best kmer)

now do you think I should try -ins_length* (with -exp_coverage auto and -cov_cutoff auto)?- thank you for the advice and help

[mmmm]

regarding Kmer length- velvet doe snot accept any values above 31- is this correct?

have tried kmer 21 to 31 and the best N50 was for kmer 27 (n50 39430 but number of nodes increased to 756)- it was 633 nodes for kmer31 (n50 38471)

I think I should use kmer 27?- I got this assembly using -cov_cutoff auto and -exp_cov auto

is it possible to improve this assembly?? (should insert size length be included?

[mastal]

It would probably be helpful to have a look at the examples given in the velvet manual:

https://www.ebi.ac.uk/~zerbino/velvet/Manual.pdf

With reads of 250 you could theoretically use any kmer size up to 249, or whatever is the length of your longest reads (presumably you have done some adapter/quality trimming prior to assembly).

The default kmer length velvet uses is 31. To find out what parameters your version of velvet is compiled with just type

$ velveth

and it should give you the MAXKMERLENGTH it was compiled with at the top of the help page.

to change the maxkmer length, navigate to the velvet directory and type

$ make 'MAXKMERLENGTH=249'

[mmmm]

thank you very much- have got the highest N50- but with 500 nodes and the node length is >1kb only in 40 nodes-
when mapped the contigs to the raw reads- got 99% mapping? (so only 1% did not map to the original reads)