How good bowtie can map SOLiD data?

[ypsd]

I tried the following command to map the attached fq file (single end, ChIP-seq).

bowtie -C -p $threads -n 3 -e 160 -a -m 1 --best --strata $BOWTIE2_INDEX/hg19_c s_6.fq -S s_6n.sam 2>&1 | tee s_6n.log

The mappable reads are of only 35%.

# reads processed: 2500
# reads with at least one reported alignment: 885 (35.40%)
# reads that failed to align: 1376 (55.04%)
# reads with alignments suppressed due to -m: 239 (9.56%)
Reported 885 alignments to 1 output stream(s)

The following command gives 38% (as -e is large)

bowtie -p 8 -C -m 10 -l 27 -n 3 -e 950 -k 1 $BOWTIE_INDEX/hg19_c s_6.fq -S s_6.sam 2>&1 | tee s_6.log
# reads processed: 2500
# reads with at least one reported alignment: 972 (38.88%)
# reads that failed to align: 1275 (51.00%)
# reads with alignments suppressed due to -m: 253 (10.12%)
Reported 972 alignments to 1 output stream(s)

But supposedly, SOLiD pipeline should be able map this file for over 50%. Does anybody know how good bowtie is to map SOLiD files? Are there any other parameters that can improve the results?

I see that people said, bfast is better than bowtie for default parameters. By tuning the parameter, can the results of bowtie be improved? Thanks!

http://seqanswers.com/forums/showpost.php?p=11328&postcount=3

[Jean]

We've mapped SOLiD RNAseq data with several mappers (Bowtie, BWA, bfast) and never got above 25-35% mapped. Others we have spoken to are getting similar results (or worse).

Would be curious to know if anyone has mapped better.

[iamli]

we mapped SOLiD with low mapped rate too!

we mapped SOLiD with low mapped rate too!
I used tophat to analyse it ! They are colorspace pair end reads.

We've mapped SOLiD RNAseq data with several mappers (Bowtie, BWA, bfast) and never got above 25-35% mapped. Others we have spoken to are getting similar results (or worse).

Would be curious to know if anyone has mapped better.

[kopi-o]

I have mapped 70-80% with BioScope for some SOLiD RNA-seq data (paired or not). But in many cases also much less. With Bowtie maybe 50% in the best cases.

[pmiguel]

I was talking to someone the other day who said they got >50% mapping with Novoalign. If I recall correctly. Actually they seemed to read SeqAnswers, so maybe they would chime in?

--
Phillip

[morellr]

I have 3 RNA-seq libraries generated on Solid5500. 64%-72% of the reads mapped with Lifescope.

[brentp]

I got best results with -v1 -m1 with bowtie on colorspace.

https://github.com/brentp/bowfast/tree/master/aligner-compare

[gringer]

We've been getting 30-50% mapping for SOLiD4 reads using Bowtie. It has been as low as ~9% for a very low-read sample (~400k reads multiplexed in with other samples totalling ~400M reads).

[hajime]

Based on my mapping results,the same RNA-seq sample mapped by tophat and lifescope got ~50% and ~80% mappable reads, respectively, even we have used the additional ECC primer.

I guess the Refcor module in lifescope may improve the mappable rate.

[gringer]

sure, the process used by Lifescope (which includes end clipping) increases the mappable reads. That doesn't necessarily mean it's a better result -- I can increase the sensitivity of a clinical test to 100% by always reporting a positive result.

[colindaven]

For human exome reads we generally get the following figures:

Bowtie ~35%
BWA 0.5.x ~40%
Clcbio ~50-60%
NovoalignCS ~55-65% *
Lifescope ~70-80% *



*use end clipping.
Iteratively trimmed reads are still very useful, provided you don't make them too short.

Bowtie1 is not that useful for long reads, as it only allows max 3 mismatches in the seed. Apparently more mismatches are necessary for useful mapping of Solid data as said by someone about using bwa successfully.

As discussed here before, % alignments isn't really the best quality criterion for an aligner. We prefer NovoalignCS to date based on clean alignments and SNPs called versus the others via a standard pipeline.

[carmeyeii]

So sad! I have been tweaking the parameters on Bowtie to get some huge runs to map decently, but so far the max I've gotten is 3% using -l 17 -n 3 !!!! Do you think it might be better to use -v ?

Seeded quality full-index search: 16:45:26
# reads processed: 60132812
# reads with at least one reported alignment: 1882083 (3.13%)
# reads that failed to align: 58250729 (96.87%)
Reported 4232170 paired-end alignments to 1 output stream(s)
Time searching: 16:45:30
Overall time: 16:45:30