Cuffmerge and cuffdiff on bam file generated by STAR alignment

Question

[oriolebaltimore]

Dear group,

I have BAM files generated by STAR alignment and subsequently, I used this file as input to cufflinks. I followed the protocol 8 described by Dobin et al titled "Mapping RNA-seq Reads with STAR" (pmid - 26334920). cufflinks -p 12 --library-type fr-firststrand \ Aligned.sortedByCoord.out.bam

I want to use cuffmerge on the gtf files from cufflinks.

Typically I use -s/–ref-sequence <seq_dir>/<seq_fasta> and -b/–frag-bias-correct <genome.fa> for cuffmerge and cuffdiff respectively for fragment bias correction. I point the bowtie2 index directory from where cufflinks pick up the fasta files.

Since I am using STAR (indexes were downloaded from STAR site), I dont see .fa files. Is there any way i can point the fasta files that were used to generate the index files. Moreover, do you observe a significant difference in output if fasta files are not used for cuffdiff for correcting fragment bias. I did not do any testing. Wondering if anyone has some experience to share.

thanks a lot. Adrian

[Brian Bushnell]

I recommend that you not use any part of the Tuxedo pipeline for anything; it tends to be unstable, and gives inferior results compared to other differential expression software like DESeq and edgeR.