This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Can I use HTSeq-count with a merged.gtf files I created with Cuffmerge?

Hi, I am running some RNAseq analyses, but I have the problem of not having a .gff/.gtf file. Here is my analyses pipeline:

  • mapping with Bowtie2 (without providing .gff reference)
  • assembly with Cufflinks
  • merging with Cuffmerge the transcript.gtf files of all my samples (output of Cufflinks)
  • counting with HTSeq-count using the files.SAM (output of Bowtie2) and the merged.gtf (output of Cuffmerge) proceeding with DESeq then.

I would like to use HTSeq/DESeq and this is the only way I found to overcome my problem of not having a suitable .gff/.gtf file. Is it this an acceptable pipeline?

rna-seq htseq cuffmerge gtf
• 4,082 views • link
written by claudia.rivetti85 • claudia.rivetti85

I would suggest you to use tophat2 over bowtie for alignment. For DEG genes, as you have merged.gtf file, you can use cuffdiff as well and it's very well acceptable pipeline.

• 0 views • link
written by Chirag Parsania Chirag Parsania

3 answers

I used Bowtie2 as TopHat requires a .gff file that don't have available. Besides, I mapped to the trascriptome so maybe there is no need to use TopHat (for what I understand it is meant when you align against the genome..)

• 0 views • link
written by claudia.rivetti85 • claudia.rivetti85

Tophat is Bowtie with spliced alignment implemented. If you align on the transcriptome, you'll have the same results. If you align on the genome, you'll have way better results with Tophat. Also, Tophat doesn't want a GFF file by default, but you can provide one to improve mapping. Therefore, you can use it!

I personally recommend the last algorithm from their team, which is HISAT2. They claim it's up to 50 times faster than Tophat and Bowtie.

• 0 views • link
written by Matteo Schiavinato Matteo Schiavinato

To answer your question entirely: yes you can use the same GFF/GTF file. The pipeline is acceptable imho, and I personally would substitute Bowtie with HISAT2 and Cufflinks / Cuffmerge with Stringtie. Those two are the newest algorithms of the same team that developed Bowtie and Cufflinks.

• 0 views • link
written by Matteo Schiavinato Matteo Schiavinato

Thank you Macspider! I will try that! What about STAR aligner compared to Bowtie2?

• 0 views • link
written by claudia.rivetti85 • claudia.rivetti85

As far as I know, STAR is not the best choice if you want gapped alignment, which is often the case with RNA Seq reads. Moreover, HISAT2 is as fast if not even faster than STAR, I would choose that one in any case! I still have doubts about StringTie being better than Cufflinks, but it's worth a try!

• 0 views • link
written by Matteo Schiavinato Matteo Schiavinato

Infact with STAR I have frequently observed better alignment percentages.Can be used with or without gff3/gtf. Its extremely fast. RAM requirement may be high when the genome is quite big.Very good aligner IMHO.

• 0 views • link
written by Jeffin Rockey Jeffin Rockey

Log in to answer this question.

More posts like this