Self-trained fragment ion likelihood for protein-nucleic acid cross-links (after ProSE's ion priors). More...

#include <OpenNuXL/include/NuXLIonLikelihood.h>

Classes
struct	Evidence
	Matched ions of a hit and of its noise hypothesis. More...

struct	Features
	Features of one hit. More...

struct	Outcomes
	Matched ions of one hypothesis (hit or its reversed sequence) More...

Public Types
enum	Group : std::uint8_t { LINEAR = 0 , XL_UNSHIFTED = 1 , XL_SHIFTED = 2 , XL_UNLOCALIZED = 3 , GROUPS = 4 }

using	Model = FragmentIonLikelihoodModel

Public Member Functions
	NuXLIonLikelihood (double tolerance, bool ppm, int max_fragment_charge=2)

void	setSpectra (const PeakMap &spectra)
	Compact copies of the spectra the hits refer to (by scan index)

Evidence	match (Size scan, const AASequence &peptide, int precursor_charge, const std::vector< double > &shifts, int site) const
	Match the ions of a hit and of its reversed sequence.

void	resetFold (Size fold)
	Clear the models of `fold`.

void	addTraining (Size fold, const Evidence &evidence)
	Add a confident hit to the models of `fold`.

void	finalize ()
	Finalize the models of both folds.

Features	score (Size fold, const Outcomes &outcomes) const
	Features of a hit scored by the models of `fold`.

Static Public Member Functions
static AASequence	reversedSequence (const AASequence &peptide)
	Reversed sequence of the noise hypothesis (all but the C-terminal residue)

Private Member Functions
Outcomes	matchHypothesis_ (Size scan, const AASequence &peptide, int precursor_charge, const std::vector< double > &shifts, int site) const

Private Attributes
double	tolerance_

bool	ppm_

int	max_fragment_charge_

Model::PeakLists	peaks_

TheoreticalSpectrumGenerator	tsg_

std::array< std::array< Model, GROUPS >, 2 >	models_

Detailed Description

Self-trained fragment ion likelihood for protein-nucleic acid cross-links (after ProSE's ion priors).

Every b and y ion of a hit is matched against its spectrum (closest peak within the fragment tolerance) and gets an outcome: absent, or intensity-rank bin x mass-error bin (see FragmentIonLikelihoodModel, RICH contexts). The ions fall into four groups, each with its own FragmentIonLikelihoodModel:

LINEAR: ions of a hit without nucleic acid adduct
XL_UNSHIFTED: ions of a localized cross-link that do not contain the cross-linked residue
XL_SHIFTED: ions that contain it; they carry one of the fragment adducts of the cross-linked nucleotide, and the best outcome over these adducts counts
XL_UNLOCALIZED: ions of a cross-link without localization; the best outcome over no shift and every fragment adduct counts

The noise hypothesis is the hit's sequence reversed except for its C-terminal residue (modifications move with their residues, so does the cross-linked residue), matched with the same shift alternatives against the same spectrum. Models are trained on the confident hits of one half of the spectra and score the other half.

Class Documentation

◆ OpenMS::NuXLIonLikelihood::Evidence

struct OpenMS::NuXLIonLikelihood::Evidence

Matched ions of a hit and of its noise hypothesis.

Class Members
Outcomes	noise	empty if the reversed sequence equals the sequence
Outcomes	signal

◆ OpenMS::NuXLIonLikelihood::Features

struct OpenMS::NuXLIonLikelihood::Features

Features of one hit.

Class Members
double	explained_presence = 0.0	sum of P(present) of the matched ions / sum over all ions
double	llr = 0.0	summed log-likelihood ratio of all ions
double	llr_shifted = 0.0	the part of llr from XL_SHIFTED and XL_UNLOCALIZED ions
double	top_predicted_observed = 0.0	fraction of the 6 ions with the highest P(present) that matched

◆ OpenMS::NuXLIonLikelihood::Outcomes

struct OpenMS::NuXLIonLikelihood::Outcomes

Matched ions of one hypothesis (hit or its reversed sequence)

Class Members
vector< uint8_t >	groups	Group of each ion.
vector< Ion >	ions

Member Typedef Documentation

◆ Model

using Model = FragmentIonLikelihoodModel

Member Enumeration Documentation

◆ Group

enum Group : std::uint8_t

Enumerator
LINEAR
XL_UNSHIFTED
XL_SHIFTED
XL_UNLOCALIZED
GROUPS

Constructor & Destructor Documentation

◆ NuXLIonLikelihood()

NuXLIonLikelihood	(	double	tolerance,
		bool	ppm,
		int	max_fragment_charge = `2`
	)

Parameters

tolerance	fragment tolerance,
ppm	whether it is in ppm,
max_fragment_charge	highest fragment charge

References DefaultParamHandler::getParameters(), DefaultParamHandler::setParameters(), Param::setValue(), and NuXLIonLikelihood::tsg_.

Member Function Documentation

◆ addTraining()

void addTraining	(	Size	fold,
		const Evidence &	evidence
	)

Add a confident hit to the models of fold.

References NuXLIonLikelihood::GROUPS, NuXLIonLikelihood::Outcomes::groups, NuXLIonLikelihood::Outcomes::ions, NuXLIonLikelihood::models_, NuXLIonLikelihood::Evidence::noise, and NuXLIonLikelihood::Evidence::signal.

◆ finalize()

void finalize ( )

Finalize the models of both folds.

References NuXLIonLikelihood::models_.

◆ match()

NuXLIonLikelihood::Evidence match	(	Size	scan,
		const AASequence &	peptide,
		int	precursor_charge,
		const std::vector< double > &	shifts,
		int	site
	)		const

Match the ions of a hit and of its reversed sequence.

Parameters

scan	Scan index of the hit's spectrum.
peptide	The peptide, with fixed and variable modifications but without nucleic acid adduct.
precursor_charge	Precursor charge of the hit.
shifts	Masses of the fragment adducts of the cross-linked nucleotide; empty for hits without cross-link.
site	Position of the cross-linked residue, -1 if not localized (ignored without `shifts`).

References NuXLIonLikelihood::matchHypothesis_(), NuXLIonLikelihood::Evidence::noise, NuXLIonLikelihood::reversedSequence(), NuXLIonLikelihood::Evidence::signal, and AASequence::size().

◆ matchHypothesis_()

NuXLIonLikelihood::Outcomes matchHypothesis_	(	Size	scan,
		const AASequence &	peptide,
		int	precursor_charge,
		const std::vector< double > &	shifts,
		int	site
	)		const