fragcomp¶

The fragment competition module contains functionality to maintain the exclusive assignment of signal to identifications.

The sweep itself lives in Rust (alphadia_search_rs.FragmentCompetition); what is left here is the dataframe bookkeeping it needs.

alphadia.fragcomp.fragcomp.compete_for_fragments(psm_df: DataFrame, frag_df: DataFrame, cycle: ndarray, rt_tol_seconds: float = 3.0, mass_tol_ppm: float = 15.0) → DataFrame[source]¶

Remove PSMs that share fragments with other PSMs.

PSMs only compete against others in the same DIA window and within rt_tol_seconds, and the lower proba wins. Row order does not matter and is preserved.

Parameters:

psm_df (pd.DataFrame) – The PSM dataframe.
frag_df (pd.DataFrame) – The fragment dataframe.
cycle (np.ndarray) – DIA cycle.
rt_tol_seconds (float) – The retention time tolerance in seconds.
mass_tol_ppm (float) – The mass tolerance in ppm.

Returns:

The PSM dataframe, reduced to the PSMs that won their fragments.

Return type:

pd.DataFrame

Utility methods for fragment competition.

alphadia.fragcomp.utils.add_frag_start_stop_idx(psm_df: DataFrame, frag_df: DataFrame) → DataFrame[source]¶

The fragment dataframe is indexed by the precursor index.

This function adds the start and stop indices of the fragments to the PSM dataframe.

Parameters:

psm_df (pd.DataFrame) – The PSM dataframe.
frag_df (pd.DataFrame) – The fragment dataframe.

Returns:

The PSM dataframe with the start and stop indices of the fragments.

Return type:

pd.DataFrame

alphadia.fragcomp.utils.candidate_hash(precursor_idx: ndarray, rank: ndarray) → ndarray[source]¶

Create a 64 bit hash (np.ndarray[np.uint64]) from precursor_idx and rank.

The precursor_idx is the lower 32 bits. The rank is the next 8 bits.

Note: the explicit casting is important to ensure it returns a 64-bit integer.