compare GSMapper and BLAT

[seq_GA]

We did a small homework on comparing 454FLX reads captured using nimblegen Seq Cap technology. We used GSMapper as well as BLAT and the results differ. Any one compared the results like this? Please let me know what algorithm u have used for mapping.

My next question is regarding gSelAdapter. Since our software version doesnot allow us to trim the adapter, any other software available to achieve this? Still we haven't upgraded to version 2 which has the facility to trim gSel Adapter.

Regards

[bioinfosm]

Hi

Could you share how the results differ? what parameters were used for blat?

[hlu]

I believe the latest version of GSMapper (2.0 software for Titanium platform) support nimblegen Seq Cap out of box. I have seen presentation, but I have never used it myself.

I have used blat/gsMapper in the past, not for numblegen data, just for trancriptome profiling and others.

Both are useful, and doing slightly different. It is probably best if 2 can be combined to generate useful results.

[doxologist]

can you guys be a little more specific? what does slightly different mean? 5% 10% totally different hits or the target sizes just are a little different?

thanks a bunch

[hlu]

can you guys be a little more specific? what does slightly different mean? 5% 10% totally different hits or the target sizes just are a little different?

thanks a bunch

All fasta mapping algorithm are heuristic as my understanding. blat is super-fast algorithm similar to blast, but less sensitive. It works at close related sequences with small SNP and difference. This works for numblegen type of mapping, which is exactly with human ref available.

454 gsMapper is also blast similar type of mapping tool, which is heuristic.

They should provide similar mapping results, but I do not expect them to be identical in the results due to heuristic nature of mapping algorithm. Further more, changing mapping cut off will also generate different level of hits.

One big difference is treat of repeat reads on blat vs gsMapper:

In the past I used both blat and gsMapper, I have found 454 gsMapper typically exclude those repeats reads, and only generate uniquely mapping reads results in coverage or stats counting. The result layout is nice as all the reference stats are there.

In blat, all mapping hits are included, regardless of uniquely mapping or repeat (meaning one reads mapping to 2 locations of reference). Just this difference is huge.

For example, if you need to find where those repeats reads in 454 gsMapper hits on reference, 454 gsMapper only report read ID, no more information. I found it useful in the past to extract those repeat reads, and use blat to map to reference.

But gsMapper is convenient as it reports how much reads hittting which reference sequence. particularly the next gsMapper software release will much better support Nimblegen data out of box for Titanium data set, which should provide better handling on interpretation of Nimblegen data set.

[joa_ds]

i use BLAT as it just gives you more flexibility. Using gsMapper, i miss a lot insertions and deletions that are known to be in the sequences. If i use BLAT with the -fine parameter and i combine multiple hits data, i can extract even CNVs and 1-30nuc deletions, whereas gsMapper tends to miss certain deletions or insertion/deletion combinations.