new_majiq.DPsiPrior.empirical_update

DPsiPrior.empirical_update(psi1, psi2, minreads=30.0, min_experiments_f=0.5, min_lsvs=100, n_update_a=1, n_update_pmix=None, legacy=False, show_progress=False)

Use reliable binary events from psi1,2 to return updated prior

Use high confidence empirical deltapsi from input groups of experiments meeting criteria: + one junction only per event + binary events only + must have passed at least min_experiments events + must also pass additional minreads criteria (higher confidence, etc.)

Parameters:

• psi1, psi2 (rna_majiq.PsiCoverage) – psi coverage for two groups of experiments. Read evidence will be combined per group.

• minreads (float) – Additional criteria (beyond having previously passed) for being considered

• min_experiments_f (float) – Proportion (if < 1) or number of experiments that must pass all criteria in order to be considered

• min_lsvs (int) – If less than this many binary events meeting criteria, don’t attempt an update to prior

• n_update_a (int) – Number of iterations to update a during M step

• n_update_pmix (Optional[int]) – Optional number of iterations to update pmix during M step. If not specified, use 1 + n_update_a.

• legacy (bool) – If True, use old implementation in v2 that does hard binning of observed dpsi into hard-coded bins at differences of 5% and 30%.

• show_progress (bool) – Attempt to show progress on distributed cluster for Dask