Bypass sleuth loader - Githubissues

bartongroup / RATS

Relative Abundance of Transcripts: An R package for the detection of Differential Transcript isoform Usage.

MIT License

32 stars 1 forks source link

Bypass sleuth loader #26

Closed fruce-ki closed 7 years ago

fruce-ki commented 7 years ago

Using sleuth to get scaled estimated counts requires creating a huge sleuth object in memory, in the order of 10 times larger than the actual data, even when no DTE models or multiple conditions are fitted. This is a huge overhead for loading Salmon/Kallisto quantifications into the much more light-weight RATs.

[x] Input bootstrapped abundances directly from .h5 files
[x] Apply TPM scaling

fruce-ki commented 7 years ago

est_counts in the bootstraps table are not normalised. They are simply the read counts assigned by kallisto/salmon. Normalised TPMs and est_counts are available elsewhere in the sleuth object. But cross-smaple normalisation is not necessary for RATs.

What RATs would prefer over est_counts, is TPMs scaled to remove the -PM part of their name.

fruce-ki commented 7 years ago

Having sleuth as a dependency is causing R CMD check namespace errors. Seems to be importing something that is not explicitly declared in the namespace, but it doesn't really say what.

The sooner I can import data independently, the sooner I can drop the sleuth dependency.