Tophat2 with GTF annotation file made from blast results

Question

[Anton Shekhov]

Hi!

The goal of my subproject is to find transcripts upregulated upon treatment in the larvae of Spodoptera littoralis. On the current step I have an assembly, and the tblastx results (assembly against the huge database from NCBI). In order to continue with annotation I transformed blast results into gtf file (wrote a python script that also does cutoff and then write results into gtf format). The gtf file looks like this:

Slitt_C1    tblastx exon    4225    5697    667 +   .   gene_id "gi|827554818|ref|XM_004929801.2|"; transcript_id "gi|827554818|ref|XM_004929801.2|";

After that aligned the reads to the assembly by bowtie2, and tried to run tophat2.

~/bin/tophat2 -G bowtie_result/Slitt.gtf -o tophat_with_annotation/ -p 16 bowtie_result/Slitt ../02_trim/A_R1_P.fq ../02_trim/A_R2_P.fq

And I got the error:

[2016-07-28 11:05:14] Beginning TopHat run (v2.1.1)
-----------------------------------------------
[2016-07-28 11:05:14] Checking for Bowtie
                  Bowtie version:        2.2.8.0
[2016-07-28 11:05:15] Checking for Bowtie index files (genome)..
[2016-07-28 11:05:15] Checking for reference FASTA file
[2016-07-28 11:05:15] Generating SAM header for bowtie_result/Slitt
[2016-07-28 11:05:15] Reading known junctions from GTF file
[2016-07-28 11:05:20] Preparing reads
         left reads: min. length=59, max. length=99, 4805446 kept reads (1341 discarded)
        right reads: min. length=59, max. length=99, 4795101 kept reads (11686 discarded)
[2016-07-28 11:08:38] Building transcriptome data files transcriptome_index/Slitt
[2016-07-28 11:08:42] Building Bowtie index from Slitt.fa
        [FAILED]
Error: Couldn't build bowtie index with err = 1

After looking into the log file, the last thing tophat was trying to do was to run bowtie2-build on the Slitt.fa from the temp folder.

I already checked the names in the assembly and annotation, it is all the same, so mistake is coming from something else. I would appreciate any tips how to get blast results into the expression level file (of course I can run a script for assigning the regions of scaffolds to the specific annotations, but it will take a lot of time).

Thank you!

[Anton Shekhov]

Even though it won't fix this problem I am thinking to use STAR.

[WouterDeCoster]

Would be a good idea, although it might satisfy you to solve the problem ;-)

[GenoMax]

Is the Slitt.fa file in the same directory as your bowtie index (bowtie_result)? If not try putting a copy in there.

[Anton Shekhov]

yep, I guess I tried all the standard solutions there are in the internet.

[GenoMax]

You may want align against the genome (rather than the transcriptome) so as to avoid forcing the aligner to mis-align reads, especially in this case where you don't have a well defined transcriptome.

[Anton Shekhov]

In the perfect world I would. However, there is no genome for this or even close related species.

[GenoMax]

TopHat is deprecated to some extent so perhaps using STAR or HISAT2 (if you want to stay with the same family) may be better option as you have already indicated.

I would recommend that you try BBMap. If you do use it then remember to add flag sam=1.3 since the default SAM flags are v. 1.4 which are not understood by featureCounts/HTSeq-count.

Answer 1

[EagleEye]

• Have you first tried to build Bowtie index on your genome file before running Tophat ? We might get some some clue if you try.

• Also check Chromosome names, whether it matches between GTF and your genome FASTA file.

[Anton Shekhov]

Bowtie2-build and bowtie worked perfectly before tophat2 step. It only did not work during tophat process. The names of the chromosomes are the same everywhere, I checked it several times.