Denovo RNA-Seq assembly using Velvetoptimiser

[nareshvasani]

Hello Fellows,

I used velvetoptimiser to find optimal value of Kmer:
1] I run it from 45 to 71,
##VelvetOptimiser.pl -s 45 -e 71 -f '-short -fastq Barcode10_trp1_unmatched_forward_1mismatch_removed-trim300.fastq' -t 8 -k 'n50' -c Lbp

It ran fine but shows warning like:

Aug 6 17:45:32 Setting cov_cutoff to 1.483.
**********
Warning: Assembly score for assembly_id 14 is 0
You may want to consider choosing a different optimisation variable or function.
Current optimisation functions are n50 for k value and Lbp for cov_cutoff
**********
Aug 6 17:51:35 Setting cov_cutoff to 0.917.
**********
Warning: Assembly score for assembly_id 14 is 0
You may want to consider choosing a different optimisation variable or function.
Current optimisation functions are n50 for k value and Lbp for cov_cutoff
**********
Looking for best cutoff score between 0.000 and 2.400
Max cutoff lies between 0.000 & 1.483
fc = 0 fd = 0 abs diff = 0
Aug 6 17:52:17 Setting cov_cutoff to 0.567.
**********
Warning: Assembly score for assembly_id 14 is 0
You may want to consider choosing a different optimisation variable or function.
Current optimisation functions are n50 for k value and Lbp for cov_cutoff
**********
Max cutoff lies between 0.000 & 0.917
fc = 0 fd = 0 abs diff = 0
Aug 6 17:52:59 Setting cov_cutoff to 0.350.
**********
Warning: Assembly score for assembly_id 14 is 0
You may want to consider choosing a different optimisation variable or function.
Current optimisation functions are n50 for k value and Lbp for cov_cutoff
**********
Optimum value of cutoff is 0.92
Took 3 iterations
Aug 6 17:53:41


Final optimised assembly details:
********************************************************
Assembly id: 14
Velveth timestamp: Aug 6 2013 17:30:03
Velvetg timestamp: Aug 6 2013 17:53:39
Velveth version: 1.2.03
Velvetg version: 1.2.03
Readfile(s): -short -fastq Barcode10_trp1_unmatched_forward_1mismatch_removed-trim300.fastq
Velveth parameter string: auto_data_71 71 -short -fastq Barcode10_trp1_unmatched_forward_1mismatch_removed-trim300.fastq
Velvetg parameter string: auto_data_71 -clean yes -exp_cov 3 -cov_cutoff 0.3502176
Assembly directory: /media/DATAPART3/RNA_seq_Analysis_7_26_13/Barcode_Splitter/auto_data_71
Velvet hash value: 71
Roadmap file size: 241309783
Total number of contigs: 13803
n50: 240
length of longest contig: 989
Total bases in contigs: 3215066
Number of contigs > 1k: 0
Total bases in contigs > 1k: 0

##conclusion it show 71 is optimal

2] I also run it from 27 to 43, but it show similar kind of warning like above.

##conclusion it shows 43 is optimal.

Can anybody please explain what does this mean?

Your feedback will be much appreciated!

Thanks in advance.

[AdrianP]

For RNA Seq, dont do such a boad range. Run the assembler from 45 to 55, 57 to 65, 67 to 75. Or split it even in lower intervals. Then, you can merge those assemblies.

Read the oases manual on how to merge the assemblies.

[nareshvasani]

Hi Adrian

But i am just trying find optimal kmer value and n50 value.
Can you please provide me any better suggetion for fining optimal kmer value.
How to select best n50 value.

Thanks.

[AdrianP]

You are trying to optimize your kmer, for the highest n50, I understand that.

Many people will agree with me, that optimizing for a higher n50 is wrong, when it comes to genome assemblies.

You are not even doing a genome assembly, this is RNASeq data. Your goal (or better yet said, my goal if I was doing this), is to have as many mRNA in your contig file as possible, and for them to be as complete as possible.

n50 has nothing to do with RNA data. It can be computed, but it is meaningless.

[nareshvasani]

Thanks AdrianP,

Thanks for your suggestion. It make sense after your explaination. You are right i am not doing genome assembly.
Then on what basis I should select kmer value? Any suggestion.

[winsettz]

I would assess coverage of your assembled contigs by your reads (is the coverage good?), and validate against a putative reference database to see if you are assembling noise.

If you blast your largest unassembled reads, what do you get? Do you get full coverage of all of your reads to it? Or is it bits and pieces mapping to everything in blast?

[nareshvasani]

HI winsettz

I would assess coverage of your assembled contigs by your reads (is the coverage good?), and validate against a putative reference database to see if you are assembling noise.

If you blast your largest unassembled reads, what do you get? Do you get full coverage of all of your reads to it? Or is it bits and pieces mapping to everything in blast?

This is first time I am using velvet. I have not done any blast yet.
I think we need contig.fa file for running blast. I am right?


Thanks!

[AdrianP]

Is this Illumina technology that you have sequenced your reads with?

[winsettz]

Is this Illumina technology that you have sequenced your reads with?

The last thread he posted with mentioned Ion Torrent; but I'm not sure if it's the same libraries.

[nareshvasani]

Is this Illumina technology that you have sequenced your reads with?

It is from ion proton.

[nareshvasani]

The last thread he posted with mentioned Ion Torrent; but I'm not sure if it's the same libraries.

It is same libraries.