search

gagneurlab / drop

Pipeline to find aberrant events in RNA-Seq data, useful for diagnosis of rare disorders
MIT License
128 stars 43 forks source link

Does FRASER make any comparisons between genes, or only comparing splicing within a given gene (between sample and controls)? #520

Closed dissakov closed 3 weeks ago

dissakov commented 4 months ago

Hello! Does FRASER do any correction/normalization between genes or make any assumptions about the number of causal genes in a sample? I'm trying to run a sample that has a few thousand splicing aberrations, but FRASER is only returning a few candidates. Is there an assumption being made that would be violated by a sample like this? Thank you!

vyepez88 commented 4 months ago

Hi, FRASER's autoencoder does normalize between co-expressed junctions, not exactly of genes. It does not take into account causal genes. Your sample has few thousand aberrant junctions or genes? In any case it is a lot and might violate the assumptions of the model in which all samples come from the same populations and all junctions follow a Beta Binomial distribution. In that sense, FRASER is able to detect a few outliers per sample, but if there are too many (in the thousands), then the modeling might not work.