Checking workflow for assembly of time series expt using Stringtie-Ballgown

Question

[nattzy94]

I have 2 timepoints (0h and 56h) and 5 replicates per timepoint.

Starting with 10 Bam files, my workflow is to :

1. Run Stringtie on each Bam individually, giving me a total of 10 GTF files.
2. I then merge the 5 individual GTF files of each timepoint so that I get 2 files: 0h_merged.gtf and 56h_merged.gtf.
3. Run Stringtie again on the aligned bam files but this time using the respective merged file as the reference and also setting the -B and -e flags to generate the input files for DE analysis in Ballgown.

However, after reading this post, I'm not sure if I'm doing things right. From what I understand, the correct method should be to merge all 10 GTFs from step 1 into a single all_merged.gtf followed by running Stringtie again using all_merged.gtf as reference. Hence the workflow should be:

1. Run Stringtie on each Bam individually, giving me a total of 10 GTF files.
2. Merge all 10 GTF files into a single all_merged.gtf.
3. Run Stringtie on the Bam files using all_merged.gtfas reference and -B and -e flags set.

Appreciate if someone could tell me which of these methods is correct?