oktoberfest.data.Spectra.preprocess_for_machine_learning

Spectra.preprocess_for_machine_learning(include_intensities=True, include_additional_columns=None, ion_type_order=None, remove_decoys=False, search_engine_score_threshold=None, num_duplicates=None)

Filter and preprocess for machine learning applications and transform into a Parquet-serializable dataframe.

Parameters:

• include_intensities (bool) – Whether to include intensity (label) column

• include_additional_columns (Optional[list[str]]) – Additional column names that are not required by DLomix to include in output. Capitalization does not matter - internal column names are all uppercase, whereas returned column names are all lowercase.

• ion_type_order (Optional[list[str]]) – Ion type order in which to save output intensity values.

• remove_decoys (bool) – Whether to remove decoys

• search_engine_score_threshold (Optional[float]) – Search engine score cutoff for peptides included in output

• num_duplicates (Optional[int]) – Number of (sequence, charge, collision energy) duplicates to keep in output

Return type:

DataFrame

Returns:

Pandas DataFrame with column names and dtypes corresponding to those required by DLomix - modified_sequence (str) - precursor_charge_onehot (list[int]) - collision_energy_aligned_normal (int) - method_nbr (int) [- intensities_raw (list[float]) (if include_intensities == True)] [additional columns (if specified via include_additional_columns)]