Closed hmehlan closed 10 months ago
The change fixed a bug that resulted in failure to fit a joint beta distribution over all replicates of a sample type, whereupon we fell back to pooling samples from beta distributions based on inclusion and exclusion read counts of separate replicates (which generally happened for PSI close to 0 or 100). While we like the intended behaviour for conceptual reasons, the fix has now resulted in increased compute time.