rna_majiq.PsiCoverage

class rna_majiq.PsiCoverage(df, events)

Summarized raw and bootstrap coverage over LSVs for one or more experiments.

Summarized raw and bootstrap coverage over LSVs for one or more experiments as input for quantification. Coverage is a total readrate over all bins, excluding stacks and after any preceding batch correction steps, ready for quantification. Per-experiment coverage stored independently over “prefix” dimension, where prefixes originate as the prefix from BAM file names (i.e. foo/experiment1.bam -> experiment1). Coverage is accompanied by boolean array indicating whether an event is “passed for quantification” for each experiment (PsiCoverage.passed).

Provides functionality for combining and summarizing over experiments and multiple PsiCoverage objects. Functions and attributes enable computation of PSI posterior statistics under MAJIQ models for splicing quantification. Computations are performed over xarray objects. When loading PsiCoverage from Zarr files, data/computations will be loaded/performed lazily using Dask. Testing of these computations have been performed over local clusters using threads rather than processes (expensive computations generally release the GIL).

Generally, for point estimates of location, quantification with raw coverage should be preferred, as bootstrap estimates converge very closely to raw estimates as the number of bootstrap replicates becomes large. For estimates of variability, quantification with bootstrap coverage should be used to account for additional per-bin readrate variability that isn’t fully captured by the Bayesian model on its own.

Underlying coverage is stored as the total number of reads over the event and the proportion of reads per intron/junction. This requires twice the uncompressed memory vs the number of reads per intron/junction, but permits easier lazy computation with Dask over large datasets.

Parameters:

df (xarray.Dataset) – Required variables/coordinates as in EXPECTED_VARIABLES
events (xarray.Dataset) – dataset that can be loaded along with matching introns/junctions as Events

See also

PsiCoverage.from_sj_lsvs: Create PsiCoverage from SJExperiment and Events
PsiCoverage.from_events_coverage: Create PsiCoverage from EventsCoverage
PsiCoverage.from_zarr: Load PsiCoverage from one or more Zarr files
PsiCoverage.updated: Create updated PsiCoverage with updated arrays
PsiCoverage.sum: Summed PsiCoverage over current prefixes
PsiCoverage.mask_events: Create updated PsiCoverage passing only specified events
PsiCoverage.__getitem__: Get PsiCoverage for subset of prefixes

__init__(df, events)

Initialize PsiCoverage with specified xarray datasets

Parameters:

df (xarray.Dataset) – Required variables/coordinates as in EXPECTED_VARIABLES
events (xarray.Dataset) – dataset that can be loaded along with matching introns/junctions as Events

Methods

`__init__`(df, events)	Initialize `PsiCoverage` with specified xarray datasets
`approximate_cdf`(x, **indexer_kwargs)	Compute cdf of approximate/smoothed bootstrapped posterior
`approximate_discretized_pmf`([nbins, ...])	Compute discretized PMF of approximate/smoothed bootstrap posterior
`approximate_quantile`([quantiles])	Compute quantiles of approximate/smoothed bootstrapped posterior
`approximate_stats`(labels[, quantiles, ...])	Statistics on approximate posterior means and psisamples.
`bootstrap_cdf`(x, **indexer_kwargs)	Compute cdf of mixture of bootstrapped posterior distribution
`bootstrap_discretized_pmf`([nbins, ...])	Compute discretized PMF of bootstrap posterior mixture
`bootstrap_psi_mean_population_quantile`([...])	empirical quantiles over prefixes of bootstrap_psi_mean
`bootstrap_quantile`([quantiles])	Compute quantiles of mixture of bootstrapped posterior distributions
`bootstrap_stats`(labels[, quantiles, ...])	Statistics on bootstrap posterior means and psisamples.
`concat`(*objs[, override_args, update_kwargs])	Concatenate multiple instances of class into single one
`convert_sj_batch`(sjs, lsvs, path[, ...])	Load PsiCoverage from sj paths, save to single output path
`dataset`([properties, quantiles, psibins, ...])	Extract selected properties into single `xr.Dataset`
`downsample`(num_prefixes[, rng])	Get random subset with exactly num_prefixes prefixes
`drop_prefixes`(prefixes)
`events_to_zarr`(path, mode[, consolidated])	Save events information to specified path
`from_events_coverage`(events_coverage[, ...])	Create `PsiCoverage` from `EventsCoverage`
`from_sj_lsvs`(sj, lsvs[, minreads, minbins, ...])	Create `PsiCoverage` from `SJExperiment` and `Events`.
`from_zarr`(path[, ec_idx_nchunks, prefix_nchunks])	Load `PsiCoverage` from one or more specified paths
`get_events`(introns, junctions)	Construct `Events` using saved dataset and introns, junctions
`group`([save_zarr, tmp_base_dir, ...])	Create `PsiGroup` from coverage in self
`mask_events`(passed)	Return `PsiCoverage` passing only events that are passed in input
`mock_with_psi_and_total`(psi, total[, ...])	Returns PsiCoverage over binary events with specified psi/total
`passed_min_experiments`([min_experiments_f])	Return boolean mask array for events passing min_experiments
`plot_violins`(ec_idx[, nbins, ...])	Plot posterior distributions over groups of prefixes
`raw_psi_mean_population_quantile`([...])	empirical quantiles over prefixes of raw_psi_mean
`raw_stats`(labels[, use_stats])	Statistics on raw posterior means with respect to labels
`raw_total_population_quantile`([quantiles, ...])	empirical quantiles over prefixes of raw_total
`rename_prefixes`(prefixes)	Rename prefixes as specified
`split_prefixes`([rng])	Split class randomly into evenly sized parts
`subset_mask`(prefix_mask)	Subset class to selected prefixes (provided as boolean mask)
`sum`(new_prefix[, min_experiments_f])	Create aggregated `PsiCoverage` with sum coverage over prefixes
`to_zarr`(path[, consolidated, show_progress])	Save `PsiCoverage` to specified path
`to_zarr_slice`(path, prefix_slice)	Save `PsiCoverage` to specified path for specified slice on prefix
`to_zarr_slice_init`(path, events_df, ...[, ...])	Initialize zarr store for saving `PsiCoverage` over many writes
`updated`(bootstrap_psi, raw_psi, **update_attrs)	Create updated `PsiCoverage` with new values of psi

Attributes

`DIMS_BEFORE_PREFIX`
`EVENTS_EXPECTED_VARIABLES`
`EXPECTED_VARIABLES`
`alpha_prior`	array(ec_idx) alpha parameter of prior distribution on PSI for connection
`approximate_alpha`	array(prefix, ec_idx) alpha parameter of approximated bootstrap posterior
`approximate_alpha_plus_beta`
`approximate_beta`	array(prefix, ec_idx) beta parameter of approximated bootstrap posterior
`beta_prior`	array(ec_idx) beta parameter of prior distribution on PSI for connection
`bootstrap_alpha`	array(prefix, ec_idx, bootstrap_replicate) alpha parameter of bootstrapped posterior
`bootstrap_beta`	array(prefix, ec_idx, bootstrap_replicate) beta parameter of bootstrapped posterior
`bootstrap_coverage`	array(prefix, ec_idx, bootstrap_replicate) bootstrapped raw_coverage
`bootstrap_posterior_mean`	array(...) means of mixtures of bootstrapped posteriors
`bootstrap_posterior_std`	array(...) standard deviations of bootstrap posterior distribution
`bootstrap_posterior_variance`	array(...) variances of mixtures of bootstrapped posteriors
`bootstrap_psi`	array(prefix, ec_idx, bootstrap_replicate) bootstrapped raw_psi
`bootstrap_psi_mean`	array(...) means of bootstrap posterior distribution on PSI (alias)
`bootstrap_psi_mean_legacy`	array(...) median of means of bootstrapped posteriors
`bootstrap_psi_mean_population_median`	array(...) median over prefixes of bootstrap_psi_mean
`bootstrap_psi_std`	array(...) standard deviations of bootstrap posterior distribution (alias)
`bootstrap_psi_variance`	array(...) variances of bootstrap posterior distribution on PSI (alias)
`bootstrap_total`	array(prefix, ec_idx, bootstrap_replicate) bootstrapped raw_total
`event_passed`	array(prefix, ec_idx) indicating if event passed
`event_size`	array(ec_idx) total number of connections from same event
`lsv_idx`	array(ec_idx) index identifying event it belongs to
`lsv_offsets`	array(e_offsets_idx) offsets for events into ec_idx
`num_bootstraps`	Number of bootstrap replicates used for bootstraped coverage estimates
`num_connections`	Total number of connections over all events
`num_events`	Total number of events
`num_passed`	Number of prefixes for which an event was passed
`num_prefixes`	Number of independent experiments
`prefix_total`	array(prefix) of total number of reads over entire experiment
`prefixes`	Names of independent units of analysis
`raw_alpha`	array(prefix, ec_idx) alpha parameter of raw posterior
`raw_alpha_plus_beta`
`raw_beta`	array(prefix, ec_idx) beta parameter of raw posterior
`raw_coverage`	array(prefix, ec_idx) coverage for individual connection (psi * total)
`raw_posterior_mean`	array(...) means of raw posterior distribution on PSI
`raw_posterior_std`	array(...) standard deviations of raw posterior distribution
`raw_posterior_variance`	array(...) variances of raw posterior distribution on PSI
`raw_psi`	array(prefix, ec_idx) percentage of raw_total for connection
`raw_psi_mean`	array(...) means of raw posterior distribution on PSI (alias)
`raw_psi_mean_population_median`	array(...) median over prefixes of raw_psi_mean
`raw_psi_std`	array(...) standard deviations of raw posterior distribution (alias)
`raw_psi_variance`	array(...) variances of raw posterior distribution on PSI (alias)
`raw_total`	array(prefix, ec_idx) raw total reads over event
`raw_total_population_median`	array(...) median over prefixes of raw_total