VELVET de novo plant genome, strategy question

Question

[dsoronellas]

Dear community,

I'm planning to use Velvet to assemble a plant genome, but because I'm new to de-novo genome assembly field I have a couple of questions which I looked to resolve with no fortune at all.

The situation is that I have four different libraries:

1. Nextera Mate-Pair 7Kb (HiSeq 2000) [2x101]
2. Truseq Paired-end (HiSeq 2000) [2x101]
3. Paired-End (MiSeq)-1 [2x300] (adapter has been trimmed, length is variable between 35-300)
4. Paired-End (MiSeq)-1 [2x300] (adapter has been trimmed, length is variable between 35-300)

I was planning to perform an assembly of libs 1 and 2 within the same Velvet command (using shortPaired and shortPaired2, inspired by this blog), and then in parallel run another Velvet command of a pooled file of libs 3+4. Finally combining them-all using any other program/tool.

As a hash value I will perform runnings between 21-99 bp.

• What do you think about this strategy? Do you recommend something different?
• What do you recommend as a next step to pool down all the assemblies?

Thanks for your help!!

Answer 1

[Philipp Bayer]

I'm not sure why you don't assemble all libraries at once - velvet can have more than 2 tracks, the maximum number is controlled by the CATEGORIES variable during Velvet compilation, so you can have shortPaired, shortPaired2, shortPaired3, shortPaired4 etc. Order these by insert size (as it says in your linked blog post) and you should be fine, provided you did quality trimming before.

By the way, you can also use VelvetOptimiser - that one will try the range of k-mers for you and report the "best" assembly (by default the assembly with highest N50).

[dsoronellas]

Sorry for being late in answering the post!

I will give it a try to this definetly!

[Philipp Bayer]

No worries!

One word of warning about VelvetOptimiser: If you run in with several threads, the threads don't share memory. So if you have an assembly that will take 1GB in memory, and you run VelvetOptimiser with 4 threads, you will need about 4GB of memory. Try running the memory estimation function first to see how many threads you need, example from manual:

VelvetOptimiser.pl -s 27 -e 31 -f '-short -fastq s_1_sequence.txt' -g 4.5 -t 8

to estimate how much memory you'll need for a 4.5MB genome and 8 threads