Salmon validating best settings for use case

Question

[yryan]

Hi folks,

I've got some illumina based mouse transcriptomic data. I've done multiple runs with salmon so I can compare the effects of cDNA, to partial and full selective alignment with decoys for indexes. I'm also repeating each of those with --gcBias and --seqBias on and off.

However moving forward I'm unsure about what would be the best tests to validate which of these combinations is most accurate/useful and decide which selections to move forward with.

[ATpoint]

I wonder why you would make these comparisons. The most recent salmon preprint has done that already and chose default parameters accordingly. I would rather leave everything at default, add both the bias flags and then perform your downstream analysis. See if results make biological sense, try to validate important findings. If the results are rubbish you can still go back and change parameters, but I would only do this if results are indeed not good, but this is probably unlikely. Comparing different settings can be quite challenging especially if you do not have a ground truth set to benchmark against.