[Kennels]

Hi boetsie

Is there a version for an Itanium machine (linux IA64)? Or anyway to compile it for such?

cheers,
kennels

[rskr]

Can you comment on what part of SSPACE you are having trouble, is it with contig extension or scaffolding?

In the SSPACE premium version i've made the read-in and Bowtie mapping multithreaded. Also the results are immmediately read-in into SSPACE, instead of writing the results to files and reading of those files. Extension has also speeded-up by immediately read-in of the results and faster way of finding overlap of the reads on the contigs.

Regards,
Boetsie

The basic version, but one that I have modified to read bwa files instead.

[boetsie]

What is the run-time complexity for SSPACE? It seems that we are having trouble getting it to scale, and what was taking days on a smaller set is now taking several weeks on a larger set. Also is there anyway to get the alignments to be done in parallel?

Can you comment on what part of SSPACE you are having trouble, is it with contig extension or scaffolding?

In the SSPACE premium version i've made the read-in and Bowtie mapping multithreaded. Also the results are immmediately read-in into SSPACE, instead of writing the results to files and reading of those files. Extension has also speeded-up by immediately read-in of the results and faster way of finding overlap of the reads on the contigs.

Regards,
Boetsie

[rskr]

What is the run-time complexity for SSPACE? It seems that we are having trouble getting it to scale, and what was taking days on a smaller set is now taking several weeks on a larger set. Also is there anyway to get the alignments to be done in parallel?

[colindaven]

Hi Sspace users,

just a bug report and solution we experienced here on a Suse 64 bit system.

We had this error

"
=>Thu Sep 8 13:45:01 2011: Building Bowtie index for contigs (/working/colin/seq/cloneC/sspace/C40A_454AllContigs.fna)

Bowtie-build error; -1 at (directory)"

There was an issue with bowtie-build.

We solved by making sure the bowtie executables were all executable, ie. chmod a+x on the files in the SSPACE/bowtie subdirectory. We also did chmod g+w though that likely didn't do much.

cheers

[boetsie]

Hi Boetsie,

Yes, this was caused by the fact that each contig was present together with its reverse complementary. After leaving only one contig from each pair everything works great.

Thank you for your help and for designing SSPACE! =)

Best,
Sasha

Ah, great! Good that the problem is solved. I already thought something like that was happening.

Thanks for the support and the nice comments!

Boetsie

[kulikov]

Could you check two things;
- are there are any readfiles generated for this project in folder 'reads'?
- If there are, can you try to run it on a subset of your contigs (2 contigs for example) and see how if any reads align.

If a read can map to more than one location (reverse complement and normal) it is not used for scaffolding.

Regards,
Boetsie

Hi Boetsie,

Yes, this was caused by the fact that each contig was present together with its reverse complementary. After leaving only one contig from each pair everything works great.

Thank you for your help and for designing SSPACE! =)

Best,
Sasha

[boetsie]

Hi Boetsie,
Thank you for your suggestions, I will try it. Can you may be show the parameters you used for the panda.
Thank you,
Cheers,
Ashu

For the Panda genome I did not do any extension. For scaffolding I used

-x = 0
-k = 3
-a = 0.7
-n = 15

Regards,
Boetsie

[Ashu]

Hi Boetsie,
Thank you for your suggestions, I will try it. Can you may be show the parameters you used for the panda.
Thank you,
Cheers,
Ashu

[boetsie]

Hi Ashu,

I know it takes a lot of time to extend the contigs, especially with large genomes. Can you tell me what part takes up the most time?

I've speeded up some things of contig extension in the SSPACE premium, you can give it a shot. One thing is the mapping of the reads against the contigs, which is now multithreaded. Another is the search for partial overlap of the reads on the contigs, which uses a faster method.

I've never ran SSPACE extension on a eukaryotic genome myself, sorry i can't tell you.

Regards,
Boetsie

[Ashu]

Hi Boetsie,
I was trying to use test the with extension x -1 , and it takes a lot of time and memory, I dont have a memory problem, but time. What is your experience with a mouse genome, around 75x coverage
Thank you in advance,
cheers,
Ashu

[boetsie]

Hi Boetsie,

No, contigs1 is not a mix. This is just a raw set of contigs. contigs2 are obtained from contigs1 after repeat resolving stage. I guess some parts of contigs may be present in other contigs (in contigs1), but it is still strange that no pair is found on contigs.

Btw, what happens if not only a read, but also its reverse complementary is mapped somewhere?

Could you check two things;
- are there are any readfiles generated for this project in folder 'reads'?
- If there are, can you try to run it on a subset of your contigs (2 contigs for example) and see how if any reads align.

If a read can map to more than one location (reverse complement and normal) it is not used for scaffolding.

Regards,
Boetsie

[kulikov]

Hi Kulikov,

Are the contigs of summary1.txt a mix of the two assemblies? In other words; are parts of the contigs present in other contigs? Because what i think has happened, is that reads could map to multiple contigs. SSPACE does not allow reads to map to multiple contigs.

Boetsie

Hi Boetsie,

No, contigs1 is not a mix. This is just a raw set of contigs. contigs2 are obtained from contigs1 after repeat resolving stage. I guess some parts of contigs may be present in other contigs (in contigs1), but it is still strange that no pair is found on contigs.

Btw, what happens if not only a read, but also its reverse complementary is mapped somewhere?

[boetsie]

Hi Boetsie,

Thanks for the update, I've started using it in spades for a number of datasets and its great.

In one project, I have a total of around 770 million 100nt-long PE reads across 7 lanes. Unfortunately I am quite limited in computing capacity (only 16Gb RAM - yes I have been critiqued before in other posts to get better specs, but we all have our circumstances ) till we get access to a better one, so as expected an analyses pretty much stopped at the stage of reading the unmapped reads into memory. I currently just want to extend a small number of separate contigs as much as possible, and it would be great to consider all reads at once.

1. I'm just wondering if it is possible to overcome the memory limitation - is the Premium version using a different way to store/access the data?
2. Should I split my inputs into smaller libraries - does sspace free up memory after reading a library before going on to the next (but i think not)?
3. Or should I carry out separate analyses of sspace - but i'm afraid of losing possibilities to extend contigs by not considering all data at once.

Sorry if the questions are naive.

Cheers,
kennels

Hi Kennels,

good that it works great!

To start; it is important to filter your reads on quality, especially with such large read length. For extension the whole read is mapped to the contigs, if not mapped it will be used for contig extension. If the quality of the reads (or part of the reads) are bad, the read will not map and it will be used for contig extension.

1.
In SSPACE premium;
- you can run bowtie with gaps, allowing up to 3 gaps, thereby reducing the number of unmapped reads and thus the number of reads stored in memory.
- A different method of storing the unmapped reads is used compared with Basic version, saving 25% of memory.
- extension is faster.

2.
For contig extension all libraries are used at once, so all reads are used. Memory is thus not freed after each library.

3.
You will loose coverage if you split the libraries. If you have sufficient coverage for one library, I should give it a go.

Regards,
Boetsie

[boetsie]

Hello Boetsie,

I'm now playing with SSPACE and I'm getting some strange output. I have two files with contigs -- say, contigs1.fasta and contigs2.fasta. They were output by the same assembler on the same data set (E.coli reads). The second file have some contigs from the first file glued together. For some reasons, SSPACE successfully scaffolds contigs2, but fails to find a single read pair on contigs1. Could you please help with this? I'm attaching the two summary files.

Hi Kulikov,

Are the contigs of summary1.txt a mix of the two assemblies? In other words; are parts of the contigs present in other contigs? Because what i think has happened, is that reads could map to multiple contigs. SSPACE does not allow reads to map to multiple contigs.

Boetsie