Oktoberfest: Rescoring and Spectral Library Generation for Proteomics.

API

Import Oktoberfest using

import oktoberfest as ok

Data: `data`

The data submodule provides access to PSMs, predictions, and metadata.

data.Spectra(*args, **kwargs)

Main to init spectra data.

Preprocessing: `pp`

Generating libraries

`pp.digest`(fasta, digestion, ...)	Digest a given fasta file with specific settings.
`pp.generate_metadata`(peptides, ...[, ...])	Create metadata about peptides for a spectral library.
`pp.gen_lib`(input_file)	Generate a spectral library from a given input.
`pp.merge_spectra_and_peptides`(spectra, search)	Merge peptides with spectra.
`pp.annotate_spectral_library`(psms[, ...])	Annotate all specified ion peaks of given PSMs (Default b and y ions).

Spectra preprocessing

`pp.list_spectra`(input_dir, input_format)	Return a list of all spectra files of a given format.
`pp.convert_raw_to_mzml`(raw_file, output_file)	Convert raw to mzML format.
`pp.convert_d_to_hdf`(d_dir, output_file)	Convert d to hdf format.
`pp.load_spectra`(filenames[, parser, ...])	Read spectra from a given file.

Peptide preprocessing

`pp.convert_search`(input_path, search_engine)	Convert search results to Oktoberfest format.
`pp.load_search`(input_file)	Load search results.
`pp.split_search`(search_results, output_dir)	Split search results by spectrum file.
`pp.convert_timstof_metadata`(input_path, ...)	Convert metadata for timstof to Oktoberfest format.
`pp.split_timstof_metadata`(timstof_metadata, ...)	Split timstof metadata by spectrum file.
`pp.filter_peptides`(peptides, min_length, ...)	Filter search results using given constraints.
`pp.filter_peptides_for_model`(peptides, model)	Filter search results to support a given peptide prediction model.

Predicting: `pr`

Access to functions that interface either a Koina server to retrieve predictions from various prediction models, or DLomix to serve & refinement-learn pre-trained models locally.

High-level prediction runner

pr.Predictor(predictor, model_name)

Abstracts common prediction operations away from their actual implementation via the DLomix or Koina interface.

Koina interface

pr.Koina(*args, **kwargs)

Extension of the Koina GRPC class in koinapy, to add required logic for Oktoberfest.

DLomix interface

`pr.DLomix`(model_type, model_path, ...[, ...])	A class for interacting with DLomix models locally for inference.
`pr.create_dlomix_dataset`(libraries, output_dir)	Transform one or multiple spectra into Parquet file that can be used by DLomix.
`pr.refine_intensity_predictor`(...[, ...])	Perform refinement/transfer learning on a baseline intensity predictor.

Rescoring: `re`

`re.generate_features`(library, search_type, ...)	Generate features to be used for percolator or mokapot target decoy separation.
`re.merge_input`(tab_files, output_file)	Merge spectra file identifier specific tab files into one large file for combined percolation.
`re.rescore_with_mokapot`(input_file[, ...])	Rescore using mokapot.
`re.rescore_with_percolator`(input_file[, ...])	Rescore using percolator.

Plotting: `pl`

`pl.plot_score_distribution`(target, decoy, ...)	Generate histogram of the score distribution for targets and decoys.
`pl.joint_plot`(prosit_target, prosit_decoy, ...)	Generate joint plot to compare rescoring with and without peptide property predictions.
`pl.plot_gain_loss`(prosit_target, ...)	Generate venn barplots to show lost, common and shared targets below 1% FDR attributed to peptide property predictions.
`pl.plot_mean_sa_ce`(sa_ce_df, filename)	Generate dotplot for spectral angle distribution over range of collision energies used for fragment intensity prediction.
`pl.plot_violin_sa_ce`(sa_ce_df, filename)	Generate violinplot for spectral angle distribution over range of collision energies used for fragment intensity prediction.
`pl.plot_pred_rt_vs_irt`(prosit_df, ...)	Generate scatterplot to compare predicted indexed retention time against (aligned) experimentally observed retention time.
`pl.plot_sa_distribution`(prosit_df, ...)	Generate spectral angle distribution for targets and decoys.
`pl.plot_mirror_spectrum`(spec_pred, mzml, ...)	Generate a mirror plot comparing an experimental and predicted MS/MS spectrum.
`pl.plot_all`(data_dir, config)	Generate all plots after a rescoring run.