Alignment Questions

[ssing]

Dear all,

I am working with 454 sequence from a transcriptome of an organism 2 million years diverged from the rat. De novo assembly using Newbler was very disappointing, so we are trying other software that uses the reference genome as a guide....however, most of these alignment software are designed for shorter reads (<60 bp) or don't feed well into an assembly viewing software like EagleViewer/Consed.

MOSAIK is our best bet right now, but we are having challenges getting it to run on our machine (64 bit vs 32 bit issues). Any other suggestions?

Thanks!

[swbarnes2]

I think the original Euler will work on longer reads. Euler SR is the new version, for short reads. I don't know about visualizing contigs after, though.

[hlu]

GS De NOVO Assembler (newbler) has its issue with repeat rich transcriptome reads.

MIRA claims to use repeat region and support EST for 454 data.

I have never used MIRA and I am looking at this myself. Give it try.

[sklages]

You have mentioned Consed. Why not using it for alignment of your 454 data to a reference sequence? This is possible since version 17.0. There is a version 18 freshly out.
Consed uses cross_match (in a new version as well) for aligning the reads.

But keep in mind that none the programs can handle splice events! That will cause problems.
Mapping ESTs may be done ESTmapper or similar programs.

[kmcarr]

We do a lot of de novo transcriptome assembly from 454 data and have worked up the following pipeline.

First you have to clean the heck out of the raw data; we use a combination of cross_match, SeqClean (see link below) and fuzznuc (part of the EMBOSS package). Most of our researchers provide us with cDNA which has been prepared using the Clontech SMART cDNA construction kit so you have to make sure to get all remnants of the adapter sequence off the reads.

Then we assemble the reads using the TGI Clustering tools. Both SeqClean and TGICL were originally developed at TIGR but are now maintained at the DFCI (http://compbio.dfci.harvard.edu/tgi/software). TGICL is itself a pipeline which first clusters the reads based on on similarity scores and then assembles the clusters using CAP3. We use a fairly stringent set of parameters here.

We then take the contigs produced from the step above and reassemble the whole group using CAP3. For this round we use less stringent parameters.

The assemblies produced are not going to be perfect. There will be considerable redundancy in your final set of contigs (that is, multiple contigs which likely represent the same gene), indels (particularly at homopolymer runs) and some chimeras.

If you want to try reference guided assembly have a look at PASA (http://pasa.sourceforge.net). I used it recently on a project combining 454 and Sanger reads and it worked quite well. Of course I had a draft genome of the organism which the EST came from. I'm not sure how well it would work in your case. It would really depend on how well you are able to align the 454 reads to the Rat genome.

[mchaisso]

I think the original Euler will work on longer reads. Euler SR is the new version, for short reads. I don't know about visualizing contigs after, though.

euler-sr is a bit of a misnomer since although it was designed for short reads, you can input a file of chromosomes as reads. I've had little experience with titanium reads, but it does a decent job at assembling/clustering other 454 reads. At one point in time I was working on a method to visualize contig coverage, but shelved the project until more people cared.

[bioinfosm]

We do a lot of de novo transcriptome assembly from 454 data and have worked up the following pipeline.

First you have to clean the heck out of the raw data; we use a combination of cross_match, SeqClean (see link below) and fuzznuc (part of the EMBOSS package). Most of our researchers provide us with cDNA which has been prepared using the Clontech SMART cDNA construction kit so you have to make sure to get all remnants of the adapter sequence off the reads.

Then we assemble the reads using the TGI Clustering tools. Both SeqClean and TGICL were originally developed at TIGR but are now maintained at the DFCI (http://compbio.dfci.harvard.edu/tgi/software). TGICL is itself a pipeline which first clusters the reads based on on similarity scores and then assembles the clusters using CAP3. We use a fairly stringent set of parameters here.

We then take the contigs produced from the step above and reassemble the whole group using CAP3. For this round we use less stringent parameters.

The assemblies produced are not going to be perfect. There will be considerable redundancy in your final set of contigs (that is, multiple contigs which likely represent the same gene), indels (particularly at homopolymer runs) and some chimeras.

If you want to try reference guided assembly have a look at PASA (http://pasa.sourceforge.net). I used it recently on a project combining 454 and Sanger reads and it worked quite well. Of course I had a draft genome of the organism which the EST came from. I'm not sure how well it would work in your case. It would really depend on how well you are able to align the 454 reads to the Rat genome.

I heard of a reference guided de-novo type alignment tool coming out for short reads, but not sure where to follow up..
Any idea of such reference guided assembler for short reads .. bacterial and higher genomes?

[avilella]

Hi ssing,

I've recently been working on a method to cluster transcriptome reads and reconstruct cDNA models for species that don't have a reference genome, using existing gene family information as cross-species reference.

If you are still looking at ways to analyse your data, I could give a try. Can you contact me at avilella - ebi - ac - uk?

Cheers

Dear all,

I am working with 454 sequence from a transcriptome of an organism 2 million years diverged from the rat. De novo assembly using Newbler was very disappointing, so we are trying other software that uses the reference genome as a guide....however, most of these alignment software are designed for shorter reads (<60 bp) or don't feed well into an assembly viewing software like EagleViewer/Consed.

MOSAIK is our best bet right now, but we are having challenges getting it to run on our machine (64 bit vs 32 bit issues). Any other suggestions?

Thanks!