A variant of QT clustering for the detection of feature groups. More...

#include <OpenMS/ANALYSIS/MAPMATCHING/QTClusterFinder.h>

Inheritance diagram for QTClusterFinder:

Public Member Functions
	QTClusterFinder ()
	Constructor.
virtual	~QTClusterFinder ()
	Destructor.
void	run (const std::vector< ConsensusMap > &input_maps, ConsensusMap &result_map)
	Runs the algorithm on consensus maps.
void	run (const std::vector< FeatureMap<> > &input_maps, ConsensusMap &result_map)
	Runs the algorithm on feature maps.
Static Public Member Functions
static const String	getProductName ()
	Returns the name of the product.
static BaseGroupFinder *	create ()
	Returns an instance of this class.
Protected Types
enum	{ RT = Peak2D::RT, MZ = Peak2D::MZ }
Private Types
typedef std::map< std::pair < GridFeature , GridFeature > , DoubleReal >	PairDistances
	Distances between pairs of grid features.
typedef HashGrid< GridFeature * >	Grid
Private Member Functions
DoubleReal	getDistance_ (GridFeature left, GridFeature right)
	Calculates the distance between two grid features.
bool	compatibleIDs_ (QTCluster &cluster, const GridFeature *neighbor)
	Checks whether the peptide IDs of a cluster and a neighboring feature are compatible.
void	setParameters_ (DoubleReal max_intensity, DoubleReal max_mz)
	Sets algorithm parameters.
void	makeConsensusFeature_ (std::list< QTCluster > &clustering, ConsensusFeature &feature)
	Generates a consensus feature from the best cluster and updates the clustering.
void	computeClustering_ (Grid &grid, std::list< QTCluster > &clustering)
	Computes an initial QT clustering of the points in the hash grid.
template<typename MapType >
void	run_ (const std::vector< MapType > &input_maps, ConsensusMap &result_map)
	Runs the algorithm on feature maps or consensus maps.
Private Attributes
Size	num_maps_
	Number of input maps.
bool	use_IDs_
	Consider peptide identifications for grouping?
DoubleReal	max_diff_rt_
	Maximum RT difference.
DoubleReal	max_diff_mz_
	Maximum m/z difference.
FeatureDistance	feature_distance_
	Feature distance functor.
PairDistances	distances_
	Distance map.

Detailed Description

A variant of QT clustering for the detection of feature groups.

The algorithm accumulates all features from all input maps, then applies a variant of QT clustering to find groups of corresponding features. In more detail, every feature from every input map is considered as a potential cluster center. For every center, its nearest neighbors from the other input maps are detected and added to the potential cluster. Iteratively, the cluster with the highest quality is extracted and the clustering is updated.

Properties affecting the grouping

To be included in a particular cluster, a feature has to fulfill the following conditions:

differences in RT and m/z from the cluster center must be below user-defined thresholds (distance_RT:max_difference and distance_MZ:max_difference),
the charge state must match that of the cluster center (unless ignore_charge is set),
if use_identifications is set and both the feature and the cluster center are annotated with peptide identifications, the identifications have to match.

Every cluster contains at most one feature from each input map - namely the feature closest to the cluster center that meets the criteria and does not belong to a better cluster.

The notion of "closeness" for features is defined by the distance function implemented in FeatureDistance, the parameters of which can be set by the user.

The quality of a cluster is computed from the number of elements in it and their distances to the cluster center. For more details see QTCluster.

Optimization

This algorithm includes a number of optimizations to reduce run-time:

two-dimensional hashing of features,
a look-up table for feature distances,
a variant of QT clustering that requires only one round of clustering.

See also:: FeatureGroupingAlgorithmQT

Parameters of this class are:

Name	Type	Default	Restrictions	Description
use_identifications	string	false	true, false	Never link features that are annotated with different peptides (only the best hit per peptide identification is taken into account).
ignore_charge	string	false	true, false	Compare features normally even if their charge states are different
distance_RT:max_difference	float	100	min: 0	Maximum allowed difference in RT in seconds
distance_RT:exponent	float	1	min: 0	Normalized RT differences are raised to this power (using 1 or 2 will be fast, everything else is REALLY slow)
distance_RT:weight	float	1	min: 0	RT distances are weighted by this factor
distance_MZ:max_difference	float	0.3	min: 0	Maximum allowed difference in m/z (unit defined by 'unit')
distance_MZ:unit	string	Da	Da, ppm	Unit of the 'max_difference' parameter
distance_MZ:exponent	float	2	min: 0	Normalized m/z differences are raised to this power (using 1 or 2 will be fast, everything else is REALLY slow)
distance_MZ:weight	float	1	min: 0	m/z distances are weighted by this factor
distance_intensity:exponent	float	1	min: 0	Differences in relative intensity are raised to this power (using 1 or 2 will be fast, everything else is REALLY slow)
distance_intensity:weight	float	0	min: 0	Distances based on relative intensity are weighted by this factor

Note:

If a section name is documented, the documentation is displayed as tooltip.
Advanced parameter names are italic.

Member Typedef Documentation

typedef HashGrid<GridFeature *> Grid [private]

typedef std::map<std::pair<GridFeature *, GridFeature *>, DoubleReal> PairDistances [private]

Distances between pairs of grid features.

Member Enumeration Documentation

anonymous enum [protected]

Enumerator:

RT
MZ

Constructor & Destructor Documentation

QTClusterFinder ( )

Constructor.

virtual ~QTClusterFinder ( ) [virtual]

Destructor.

Member Function Documentation

bool compatibleIDs_	(	QTCluster &	cluster,
		const GridFeature *	neighbor
	)			`[private]`

Checks whether the peptide IDs of a cluster and a neighboring feature are compatible.

A neighboring feature without identification is always compatible. Otherwise, the cluster and feature are compatible if the best peptide hits of each of their identifications have the same sequences.

void computeClustering_	(	Grid &	grid,
		std::list< QTCluster > &	clustering
	)			`[private]`

Computes an initial QT clustering of the points in the hash grid.

static BaseGroupFinder* create ( ) [inline, static]

Returns an instance of this class.

DoubleReal getDistance_	(	GridFeature *	left,
		GridFeature *	right
	)			`[private]`

Calculates the distance between two grid features.

The distance is looked up in the distance map and only computed (and stored) if it's not already available.

static const String getProductName ( ) [inline, static]

Returns the name of the product.

void makeConsensusFeature_	(	std::list< QTCluster > &	clustering,
		ConsensusFeature &	feature
	)			`[private]`

Generates a consensus feature from the best cluster and updates the clustering.

void run	(	const std::vector< FeatureMap<> > &	input_maps,
		ConsensusMap &	result_map
	)

Runs the algorithm on feature maps.

Exceptions:

Exception::IllegalArgument

is thrown if the input data is not valid.

void run	(	const std::vector< ConsensusMap > &	input_maps,
		ConsensusMap &	result_map
	)			`[virtual]`

Runs the algorithm on consensus maps.

Exceptions:

Exception::IllegalArgument

is thrown if the input data is not valid.

Implements BaseGroupFinder.

void run_	(	const std::vector< MapType > &	input_maps,
		ConsensusMap &	result_map
	)			`[private]`

Runs the algorithm on feature maps or consensus maps.

void setParameters_	(	DoubleReal	max_intensity,
		DoubleReal	max_mz
	)			`[private]`

Sets algorithm parameters.

Member Data Documentation

PairDistances distances_ [private]

Distance map.

To compute it only once, the distance between two features is accessible by searching for a pair where the first position is the smaller pointer value.

FeatureDistance feature_distance_ [private]

Feature distance functor.

DoubleReal max_diff_mz_ [private]

Maximum m/z difference.

DoubleReal max_diff_rt_ [private]

Maximum RT difference.

Size num_maps_ [private]

Number of input maps.

bool use_IDs_ [private]

Consider peptide identifications for grouping?

OpenMS / TOPP release 1.10.0

Documentation generated on Thu Mar 7 2013 09:42:49 using doxygen 1.7.1