Comparison of mapping program performance for color-space reads?

[g_solid]

Hi all,
I realize that many people here may have strong preferences as to which mapping program they use, but I'm trying to find out if there is some more or less accepted comparison of the performance of available programs for mapping SOLiD reads in color-space. Perhaps something published in a paper or review.

I'm primarily interested in how the different programs compare in terms of sensitivity, specificity and speed.

Thanks, g_solid

[nilshomer]

Hi all,
I realize that many people here may have strong preferences as to which mapping program they use, but I'm trying to find out if there is some more or less accepted comparison of the performance of available programs for mapping SOLiD reads in color-space. Perhaps something published in a paper or review.

I'm primarily interested in how the different programs compare in terms of sensitivity, specificity and speed.

Thanks, g_solid

I wrote a software tool that works well with SOLiD data called BFAST. See its publication (along with comparisons): http://dx.doi.org/10.1371/journal.pone.0007767

[golharam]

I'm interested in this two and will probably do a benchmark using a dataset I have. I will need to get a test dataset as well. I'm just not finding Bioscope/corona to have any useful documentation for customization. I need something I have more control over but would like to see how the different tools line up.

[snetmcom]

to be honest, the mappers from AB are much better than what you'll find 3rd party. Sorry nils

it's pretty easy to get 60-70% mapping with their new maxmapper. If you are going to benchmark, try to get some test data into whatever their newest mapper is.

[nilshomer]

to be honest, the mappers from AB are much better than what you'll find 3rd party. Sorry nils

it's pretty easy to get 60-70% mapping with their new maxmapper. If you are going to benchmark, try to get some test data into whatever their newest mapper is.

I am quite surprised since I have not heard of poor BFAST mapping results from other independent testers on SOLiD data. Could you post the input data (simulated/real-world)? What settings did you use for BFAST? Mapping % is a poor indicator of mapping quality, as SNPs/indels wont affect that % and mapping accuracy (predictive positive value) is not assessed.

In my opinion, both have strengths and weaknesses, but obviously I am biased towards BFAST.

Nils

[KevinLam]

I'm just not finding Bioscope/corona to have any useful documentation for customization. I need something I have more control over but would like to see how the different tools line up.

AGREED.

There is also BWA for mapping http://bio-bwa.sourceforge.net/bwa.shtml

[Nitrogen-DNE-sulfer]

I'm very impressed with BFAST as its very clear and published on what its doing. Well documented as well.
Maxmappers seed and extend is fast but appears to be a little messy in low complexity regions of the genome thus its SVs and larger indels have lower concordance than previous AB tools. It does handle 25mers really well but I think BFAST is being retuned for this. Benchmarking should be done from csfasta to variant files with Indels IMHO.

I see most large centers moving to BFAST at the moment for DNA mappings. RNA is a different beast.

[snetmcom]

I am quite surprised since I have not heard of poor BFAST mapping results from other independent testers on SOLiD data. Could you post the input data (simulated/real-world)? What settings did you use for BFAST? Mapping % is a poor indicator of mapping quality, as SNPs/indels wont affect that % and mapping accuracy (predictive positive value) is not assessed.

In my opinion, both have strengths and weaknesses, but obviously I am biased towards BFAST.

Nils

ya got me and i agree, my tests were for my datasets and do not cover all real world applications.

[xguo]

I'm very impressed with BFAST as its very clear and published on what its doing. Well documented as well.
Maxmappers seed and extend is fast but appears to be a little messy in low complexity regions of the genome thus its SVs and larger indels have lower concordance than previous AB tools. It does handle 25mers really well but I think BFAST is being retuned for this. Benchmarking should be done from csfasta to variant files with Indels IMHO.

I see most large centers moving to BFAST at the moment for DNA mappings. RNA is a different beast.

Thanks for the insight. Could you elaborate your experience with RNA data? We're currently using BWA & samtools for transcriptome data. Is there anyone who has compared performance of BWA and BFAST?

[betty]

Hey guys, how about the "bowtie" program?

[nilshomer]

Thanks for the insight. Could you elaborate your experience with RNA data? We're currently using BWA & samtools for transcriptome data. Is there anyone who has compared performance of BWA and BFAST?

You could take a look at the BFAST paper.

Hey guys, how about the "bowtie" program?

It does not map with insertions or deletions (gaps).

[Haneko]

Just curious, how many bases of gaps can BFast allow for colorspace reads in alignment?

[nilshomer]

Just curious, how many bases of gaps can BFast allow for colorspace reads in alignment?

There is no maximum as BFAST does not perform enumeration to find indels, but uses spaced seeds to work around them, and the Smith Waterman to align them. In practice, BFAST is able to detect indels up to 20bp with 50bp reads, although it depends on the genome and technology etc.