search.selection

The selection package handles candidate selection for peak groups using various algorithms and strategies.

alphadia.search.selection.selection	Main candidate selection implementation for DIA data analysis.
alphadia.search.selection.config_df	Configuration DataFrames for selection parameters.
alphadia.search.selection.fft	Fast Fourier Transform operations for signal processing.
alphadia.search.selection.kernel	Kernel-based operations for candidate selection.
alphadia.search.selection.utils	Utility functions for candidate selection operations.

Main candidate selection implementation for DIA data analysis.

class alphadia.search.selection.selection.CandidateSelection(dia_data: TimsTOFTranspose | AlphaRawBase | MzML | Sciex | Thermo, precursors_flat: DataFrame, fragments_flat: DataFrame, config: CandidateSelectionConfig, rt_column: str, mobility_column: str, precursor_mz_column: str, fragment_mz_column: str, fwhm_rt: float = 5.0, fwhm_mobility: float = 0.012)

Bases: object

__init__(dia_data: TimsTOFTranspose | AlphaRawBase | MzML | Sciex | Thermo, precursors_flat: DataFrame, fragments_flat: DataFrame, config: CandidateSelectionConfig, rt_column: str, mobility_column: str, precursor_mz_column: str, fragment_mz_column: str, fwhm_rt: float = 5.0, fwhm_mobility: float = 0.012) → None

Select candidates for MS2 extraction based on MS1 features

Parameters:

• dia_data (DiaData) – dia data object

• precursors_flat (pd.DataFrame) – flattened precursor dataframe

• fragments_flat (pd.DataFrame) – flattened fragment dataframe

• config (CandidateSelectionConfig) – config object

• rt_column (str) – name of the rt column in the precursor dataframe

• mobility_column (str) – name of the mobility column in the precursor dataframe

• precursor_mz_column (str) – name of the precursor mz column in the precursor dataframe

• fragment_mz_column (str) – name of the fragment mz column in the fragment dataframe

• fwhm_rt (float, optional) – full width at half maximum in RT dimension for the GaussianKernel, by default 5.0

• fwhm_mobility (float, optional) – full width at half maximum in mobility dimension for the GaussianKernel, by default 0.012

Configuration DataFrames for selection parameters.

class alphadia.search.selection.config_df.CandidateContainer(*args, **kwargs)

Bases: CandidateContainer

class_type = jitclass.CandidateContainer#75e5dd9f7d10<precursor_idx:array(uint32, 1d, C),rank:array(uint8, 1d, C),score:array(float32, 1d, C),scan_center:array(uint32, 1d, C),scan_start:array(uint32, 1d, C),scan_stop:array(uint32, 1d, C),frame_center:array(uint32, 1d, C),frame_start:array(uint32, 1d, C),frame_stop:array(uint32, 1d, C)>

class alphadia.search.selection.config_df.CandidateSelectionConfig

Bases: JITConfig

__init__()

Base class for creating numba compatible config objects.

validate()

Validates the config object. Note that this class is not meant to be instantiated. Classes inheriting from JITConfig must implement their own validate method.

class alphadia.search.selection.config_df.CandidateSelectionConfigJIT(*args, **kwargs)

Bases: CandidateSelectionConfigJIT

Numba compatible config object for the HybridCandidate class. Please see the documentation of the CandidateSelectionConfig class for more information on the parameters and their default values.

candidate_count: int64

center_fraction: float64

class_type = jitclass.CandidateSelectionConfigJIT#75e5dd9f4fd0<rt_tolerance:float64,precursor_mz_tolerance:float64,fragment_mz_tolerance:float64,mobility_tolerance:float64,isotope_tolerance:float64,peak_len_rt:float64,sigma_scale_rt:float64,peak_len_mobility:float64,sigma_scale_mobility:float64,candidate_count:int64,top_k_precursors:int64,top_k_fragments:int64,exclude_shared_ions:bool,kernel_size:int64,f_mobility:float64,f_rt:float64,center_fraction:float64,min_size_mobility:int64,min_size_rt:int64,max_size_mobility:int64,max_size_rt:int64,group_channels:bool,use_weighted_score:bool,join_close_candidates:bool,join_close_candidates_scan_threshold:float64,join_close_candidates_cycle_threshold:float64,feature_std:array(float64, 1d, C),feature_mean:array(float64, 1d, C),feature_weight:array(float64, 1d, C)>

exclude_shared_ions: bool

f_mobility: float64

f_rt: float64

feature_mean: Array(float64, 1, 'C', False, aligned=True)

feature_std: Array(float64, 1, 'C', False, aligned=True)

feature_weight: Array(float64, 1, 'C', False, aligned=True)

fragment_mz_tolerance: float64

group_channels: bool

isotope_tolerance: float64

join_close_candidates: bool

join_close_candidates_cycle_threshold: float64

join_close_candidates_scan_threshold: float64

kernel_size: int64

max_size_mobility: int64

max_size_rt: int64

min_size_mobility: int64

min_size_rt: int64

mobility_tolerance: float64

peak_len_mobility: float64

peak_len_rt: float64

precursor_mz_tolerance: float64

rt_tolerance: float64

sigma_scale_mobility: float64

sigma_scale_rt: float64

top_k_fragments: int64

top_k_precursors: int64

use_weighted_score: bool

class alphadia.search.selection.config_df.PrecursorFlatContainer(*args, **kwargs)

Bases: PrecursorFlatContainer

candidate_start_idx: Array(uint32, 1, 'C', False, aligned=True)

candidate_stop_idx: Array(uint32, 1, 'C', False, aligned=True)

charge: Array(uint8, 1, 'C', False, aligned=True)

class_type = jitclass.PrecursorFlatContainer#75e5dd9f5750<precursor_idx:array(uint32, 1d, C),frag_start_idx:array(uint32, 1d, C),frag_stop_idx:array(uint32, 1d, C),candidate_start_idx:array(uint32, 1d, C),candidate_stop_idx:array(uint32, 1d, C),charge:array(uint8, 1d, C),rt:array(float32, 1d, C),mobility:array(float32, 1d, C),mz:array(float32, 1d, C),isotopes:array(float32, 2d, C)>

frag_start_idx: Array(uint32, 1, 'C', False, aligned=True)

frag_stop_idx: Array(uint32, 1, 'C', False, aligned=True)

isotopes: Array(float32, 2, 'C', False, aligned=True)

mobility: Array(float32, 1, 'C', False, aligned=True)

mz: Array(float32, 1, 'C', False, aligned=True)

precursor_idx: Array(uint32, 1, 'C', False, aligned=True)

rt: Array(float32, 1, 'C', False, aligned=True)

alphadia.search.selection.config_df.candidate_container_to_df(candidate_container: CandidateContainer) → DataFrame

Convert a CandidateContainer to pd.DataFrame.

Fast Fourier Transform operations for signal processing.

exception alphadia.search.selection.fft.NumbaContextOnly

Bases: Exception

alphadia.search.selection.fft.convolve_fourier(dense, kernel)

Numba helper function to apply a gaussian filter to a 2d or 3d dense matrix.

Parameters:

• dense (np.ndarray) – Array of shape (…, n_scans, n_frames)

• kernel (np.ndarray) – Array of shape (i, j)

Returns:

Array of shape (…, n_scans, n_frames) containing the filtered dense stack.

Return type:

np.ndarray

Kernel-based operations for candidate selection.

class alphadia.search.selection.kernel.GaussianKernel(dia_data: TimsTOFTransposeJIT | AlphaRawJIT, fwhm_rt: float = 10.0, sigma_scale_rt: float = 1.0, fwhm_mobility: float = 0.03, sigma_scale_mobility: float = 1.0, kernel_height: int = 30, kernel_width: int = 30)

Bases: object

__init__(dia_data: TimsTOFTransposeJIT | AlphaRawJIT, fwhm_rt: float = 10.0, sigma_scale_rt: float = 1.0, fwhm_mobility: float = 0.03, sigma_scale_mobility: float = 1.0, kernel_height: int = 30, kernel_width: int = 30)

Create a two-dimensional gaussian filter kernel for the RT and mobility dimensions of a DIA dataset. First, the observed standard deviation is scaled by a linear factor. Second, the standard deviation is scaled by the resolution of the respective dimension.

This results in sigma_scale to be independent of the resolution of the data and FWHM of the peaks.

Parameters:

• dia_data (DiaDataJIT) – dia_data jit object.

• fwhm_rt (float) – Full width at half maximum in RT dimension of the peaks in the spectrum.

• sigma_scale_rt (float) – Scaling factor for the standard deviation in RT dimension.

• fwhm_mobility (float) – Full width at half maximum in mobility dimension of the peaks in the spectrum.

• sigma_scale_mobility (float) – Scaling factor for the standard deviation in mobility dimension.

• kernel_size (int) – Kernel shape in pixel. The kernel will be a square of size (kernel_size, kernel_size). Should be even and will be rounded up to the next even number if necessary.

determine_mobility_sigma(mobility_resolution: float)

Determine the standard deviation of the gaussian kernel in mobility dimension. The standard deviation will be sclaed to the resolution of the raw data.

Parameters:

mobility_resolution (float) – Resolution of the mobility dimension in 1/K_0.

Returns:

Standard deviation of the gaussian kernel in mobility dimension scaled to the resolution of the raw data.

Return type:

float

determine_rt_sigma(cycle_length_seconds: float)

Determine the standard deviation of the gaussian kernel in RT dimension. The standard deviation will be sclaed to the resolution of the raw data.

Parameters:

cycle_length_seconds (float) – Cycle length of the duty cycle in seconds.

Returns:

Standard deviation of the gaussian kernel in RT dimension scaled to the resolution of the raw data.

Return type:

float

static gaussian_kernel_2d(size_x: int, size_y: int, sigma_x: float, sigma_y: float)

Create a 2D gaussian kernel with a given size and standard deviation.

Parameters:

• size (int) – Width and height of the kernel matrix.

• sigma_x (float) – Standard deviation of the gaussian kernel in x direction. This will correspond to the RT dimension.

• sigma_y (float) – Standard deviation of the gaussian kernel in y direction. This will correspond to the mobility dimension.

Returns:

weights – 2D gaussian kernel matrix of shape (size, size).

Return type:

np.ndarray, dtype=np.float32

get_dense_matrix(verbose: bool = True) → ndarray

Calculate the gaussian kernel for the given data set and parameters.

Parameters:

verbose (bool) – If True, log information about the data set and the kernel.

Returns:

Two-dimensional gaussian kernel.

Return type:

np.ndarray

alphadia.search.selection.kernel.multivariate_normal(x: ndarray, mu: ndarray, sigma: ndarray)

Multivariate normal distribution, probability density function

Most likely an absolutely inefficient implementation of the multivariate normal distribution. Is only used for creating the gaussian kernel and will only be used a few times for small kernels.

Parameters:

• x (np.ndarray) – (N, D,)

• mu (np.ndarray) – (1, D,)

• sigma (np.ndarray) – (D, D,)

Returns:

array of shape (N,) with the density at each point

Return type:

np.ndarray, float32

Utility functions for candidate selection operations.

alphadia.search.selection.utils.amean1(array)

alphadia.search.selection.utils.assemble_isotope_mz(mono_mz, charge, isotope_intensity)

Assemble the isotope m/z values from the precursor m/z and the isotope offsets.

alphadia.search.selection.utils.astd1(array)

alphadia.search.selection.utils.find_peaks_1d(a: ndarray, top_n: int = 3) → tuple[ndarray, ndarray, ndarray]

Accepts a dense representation and returns the top three peaks

alphadia.search.selection.utils.find_peaks_2d(a: ndarray, top_n: int = 3) → tuple[ndarray, ndarray, ndarray]

Accepts a dense representation and returns the top three peaks

alphadia.search.selection.utils.slice_manual(inst, slices)

alphadia.search.selection.utils.symetric_limits_2d(a, scan_center, dia_cycle_center, f_mobility=0.95, f_rt=0.95, center_fraction=0.01, min_size_mobility=3, max_size_mobility=20, min_size_rt=1, max_size_rt=10)

alphadia.search.selection.utils.wrap0(value, limit)

alphadia.search.selection.utils.wrap1(values, limit)