Tophat v1.4.0 (Transcriptome Alignment)

[Jon_Keats]

Not sure how many people have tried this new version. I'm a big fan of the transcriptome alignment phase this new version includes (ie. Align to transcritopme, then genome, then split remaining reads to attempt alignment to genome) as most mRNAseq reads (60+ percent) will align to known transcriptome references using bwa or your favorite aligner in a very short amount of time. So that is the good news...

The bad news, at least on my first test run, the sample I tested had 71% alignment using Tophat v1.3.2 and now with the new version, which I expected to increase the percent alignment it dropped to 62%. Oddly the 62% is almost exactly the alignment frequency (63%) I see when I align to transcriptome reference using bwa. Has anyone else noted this odd behavior or percent aligned decreasing with this version?

[rnaseek]

Hi

I am wondering whether you can share the tophat command that worked successfully for transcriptome only alignment using "-T" option. I am trying use TopHat 1.4.0 with -T option, but could not make it to work. I think I am not using the "-T" option ciorrectly as I get the error "IOError: [Errno 2] No such file or directory: '-T'".

Thanks

[westerman]

I think I am not using the "-T" option ciorrectly as I get the error "IOError: [Errno 2] No such file or directory: '-T'".

Posting your unsuccessful command line might be useful since the rest of us could debug it for you. I suspect that you are using an option that requires additional information (e.g., the --GTF option) and that option is 'swallowing up' the '-T' as its additional input.

[rnaseek]

Sorry about that and I think you are right. Here is the command that used
tophat -p 8 -o top_out -T -G <path2GTF> <path2BowTieIndex> --transcriptome-index=transcriptome_data/known <path2FASTQfile>

Thanks

[pzumbo]

I observe better alignment rate with tophat 1.4.1 vs 1.3.1.

On the same read set, I ran tophat-1.3.1 vs tophat-1.4.1.

tophat-1.3.1
~/bin/tophat-1.3.1/tophat -g 1 --segment-length 25 --segment-mismatch 2 -G /home/paz2005/bin/ppbs/references/hg19/annotation/ref/gtf/refFlat.gtf -o /tmp/pz/test/tophat131 hg19 /tmp/pz/test/test.fastq.gz &

samtools flagstat ./tophat131/accepted_hits.bam
2504664 + 0 mapped (100.00%:nan%)


tophat-1.4.1
~/bin/tophat-1.4.1/tophat -g 1 --segment-length 25 --segment-mismatch 2 -G /home/paz2005/bin/ppbs/references/hg19/annotation/ref/gtf/refFlat.gtf -o /tmp/pz/test/tophat141 hg19 /tmp/pz/test/test.fastq.gz &

samtools flagstat ./tophat141/accepted_hits.bam
2873734 + 0 mapped (100.00%:nan%)


So, tophat-1.4.1 managed to map 369,070 more reads than tophat-1.3.1.

[Jon_Keats]

For the transcriptome only alignment you should upgrade to Tophat 1.4.1 as they fixed a bug regarding this limited alignment in the new release.

Pzumbo,

Watch using samtools flagstats as you don't know how many of the 369,070 extra aligned reads are unique alignment events. If a read maps to two regions tophat reports those two alignments as two independent events. Picard alignment metrics will give you the unique alignment events compared to the total alignment events reported by samtools

[vyellapa]

We were consistently getting better alignment % for shorter read lengths on the same dataset using tophat1.4 ie. there was worse alignment % for 100bp reads compared to 50bp clipped reads from the same dataset. We believe that -N/--initial-read-mismatches option whose default is 2 should be increased for longer read lengths.

Im trying the tophat1.4.1 for -N/--initial-read-mismatches values =2,3,4 & 5 to align 101bp paired end reads.
Everything except N=4 aligned without errors.

I'm getting the following error which looks like a Python error. Did anybody come across this error?

"[Sat Feb 18 20:48:47 2012] Building Bowtie index from GRCh37_E64_1kg.fa
Traceback (most recent call last):
File "/home/vyellapantula/local/bin/tophat", line 3063, in ?
sys.exit(main())
File "/home/vyellapantula/local/bin/tophat", line 3029, in main
user_supplied_deletions)
File "/home/vyellapantula/local/bin/tophat", line 2501, in spliced_alignment
m2g_left_maps, m2g_right_maps = mapped_gtf_list
ValueError: need more than 1 value to unpack"