Any issues with the SOLiD platform?

[biotech.startup]

Hello everyone,

My company is very interested in purchasing a SOLiD 3 machine. Before we pull the trigger, I am interested in hearing user feedback on issues, library construction or emulsion pcr issues, number of reads that you don't end up using after some threshold, etc on previous versions. Also, how do users typically analyze the data that comes off the machine? Do you mainly stick to the software that ABI provides or do you integrate the open source pipeline code? Do you have a separate server running 64 bit linux to run the pipeline analysis off of the SOLiD servers or do you run analysis on the servers provided? If you run analysis on the servers provided, does this interfere with an ongoing run?

Thanks for any feedback or input that you can give.

[westerman]

I am on the computational side. Thus nothing about the library construction aside that the techs mutter about the difficulty of it. But maybe they are just in the process of getting use to it.

As for the software, we have stuck to the ABI 'corona lite' programs. They seem to do a good job in mapping the reads to the reference and determining SNPs. We have not explored the indel portion very much since it is newly released. I am still leery about any de-novo assembly using the short reads from the SOLiD.

We off-load the processing from the SOLiD cluster to our own set of computers -- 64 x86 cores running Solaris (not Linux) with an aggregate 180 GB memory. We generally do a small mapping of the run on the SOLiD cluster (say mapping to a small chromosome or mitochondria) just to make sure that the run looks good. Such mapping helps us quickly spot potential problems. Mapping the whole genome of an organism at 35 bp with 3 mismatches will take about a day on our set of computers. Mate pairing about the same. SNP calling likewise. In other words 3 days if everything goes well. My most recent project has taken about 8 days to process but I made some typing mistakes along the way (e.g., transposing the expected pairwise distance from 1800 - 4200 to 1800 - 2400 bp -- no wonder I was getting so many fewer paired reads than expected!)

[biotech.startup]

Thanks for the response westerman, that was informative on the informatics side. Do you usually just plug a USB Hard drive into the SOLiD cluster to offload the data and then transfer it to your Solaris cluster or do you just do a network transfer of the data? From what I gather from the informational spec sheets off of ABI's website, it seems as though a second run shouldn't happen until primary analysis is complete or you have a back up of the images off of the first run to perform primary analysis on a different cluster. Any insight to this?

Thanks again.

[swbarnes2]

Library prep is a pain for all the current platforms. The thing about Solid is that the reads are in color space, and there aren't as many programs designed to use that information properly. With the other two platforms, there are lots and lots of aligners, which maybe is a plus, maybe a minus.

Speed-wise, the faterst alignment algorithms are the Burrows-Wheeler transforms, and I don't know if there is one for color-space data, though there's probably someone working on it.

[westerman]

We do a network transfer. Gbit to Gbit doesn't take too long. Minutes but not hours.

As for waiting for the primary analysis to be complete. Yes, one needs to do that. At the end of the primary analysis (image files to colorspace) you end up with a ~20 GByte Colorspace fasta file, a bigger quality file (which does not get used!) and a stats file. I transfer these over to our Solaris cluster and fire up corona lite to do the mapping, pairing, snp calling.

[westerman]

And replying to swbarnes' post, I agree that there are not many programs which work properly in colorspace. But all you need is one! I hope that the ABI programs are working correctly but given the mismash of perl, python, and scripting code within them ... well, sometimes I have doubts. The speed is not much of an issue since the physical run takes about a week, two weeks for paired data. My side of the process can be much less in comparison.

[BioWizard]

Wow, you need a huge cluster to do alignment in a day ?

Applied Biosystem licenses ISAS for their own scientists to do alignment.
100 million 25mers in color space on one computer in 25 minutes, while their customers are wasting money, time and electricity running 100 computers with the free junk software. Its so unfair. With SOLiD 3 you'll start getting massive amounts of sequence data. It will take you weeks to do alignment with the junk free software. We could do it in under 3 hours. Why ABI won't release ISAS to its customers is a travesty. Before you buy SOLiD 3, maybe you have a chance - tell the ABI salesman you will only buy it if they give you a "free" (or maybe they can add a token couple dollars to the invoice) license to use Imagenix Sequence Alignment system. Otherwise - NO DEAL. Make them go back to their scientists and DEMAND they give you a license. Once you buy the system, they'll never give you the good alignment software.

[ECO]

Wow, you need a huge cluster to do alignment in a day ?

Applied Biosystem licenses ISAS for their own scientists to do alignment.
100 million 25mers in color space on one computer in 25 minutes, while their customers are wasting money, time and electricity running 100 computers with the free junk software. Its so unfair. With SOLiD 3 you'll start getting massive amounts of sequence data. It will take you weeks to do alignment with the junk free software. We could do it in under 3 hours. Why ABI won't release ISAS to its customers is a travesty. Before you buy SOLiD 3, maybe you have a chance - tell the ABI salesman you will only buy it if they give you a "free" (or maybe they can add a token couple dollars to the invoice) license to use Imagenix Sequence Alignment system. Otherwise - NO DEAL. Make them go back to their scientists and DEMAND they give you a license. Once you buy the system, they'll never give you the good alignment software.

Hi BioWizard,

Please check your PM box before posting about ISAS again. Thanks.

-=ECO

[BioWizard]

OK - noted !

I moved the discussion to a new thread in the BioInformatics forum.