PeptideProteinResolution Class Reference

Resolves shared peptides based on protein scores. More...

#include <OpenMS/ANALYSIS/ID/PeptideProteinResolution.h>

Public Member Functions
	PeptideProteinResolution (bool statistics=false)
void	buildGraph (const ProteinIdentification &protein, const std::vector< PeptideIdentification > &peptides)
void	resolveGraph (ProteinIdentification &protein, std::vector< PeptideIdentification > &peptides)
ConnectedComponent	findConnectedComponent (Size &root_prot_grp)
void	resolveConnectedComponent (ConnectedComponent &conn_comp, ProteinIdentification &protein, std::vector< PeptideIdentification > &peptides)

Private Types
typedef std::map< Size, std::set< Size > >	IndexMap_

Private Attributes
IndexMap_	indist_prot_grp_to_pep_
	mapping indist. protein group indices -> peptide identification indices More...
IndexMap_	pep_to_indist_prot_grp_
	mapping indist. protein group indices <- peptide identification indices More...
std::map< String, Size >	prot_acc_to_indist_prot_grp_
bool	statistics_
	log debug information? More...

Detailed Description

Resolves shared peptides based on protein scores.

Resolves connected components of the bipartite protein-peptide graph based on protein probabilities/scores and adds them as additional protein_groups to the protein identification run processed. Thereby greedily assigns shared peptides in this component uniquely to the proteins of the current best indistinguishable protein group, until every peptide is uniquely assigned. This effectively allows more peptides to be used in ProteinQuantifier at the cost of potentially additional noise in the peptides quantities. In accordance with most state-of-the-art protein inference tools, only the best hit (PSM) for a peptide ID is considered. Probability ties are currently resolved by taking the first occurring protein of the component.

Implement probability tie resolution.

Improvement:

The class could provide iterator for ConnectedComponents in the future. One could extend the graph to include all PeptideHits (not only the best). It becomes a tripartite graph with larger connected components then. Maybe extend it to work with MS1 features. Separate resolution and adding groups to output.

Member Typedef Documentation

IndexMap_

typedef std::map<Size, std::set<Size> > IndexMap_

private

Constructor & Destructor Documentation

PeptideProteinResolution()

PeptideProteinResolution

(

bool

statistics = false

)

Constructor

Parameters

statistics

Specifies if the class stores/outputs info about statistics

Member Function Documentation

buildGraph()

void buildGraph	(	const ProteinIdentification &	protein,
		const std::vector< PeptideIdentification > &	peptides
	)

Initialize and store the graph (= maps)

Parameters

protein	ProteinIdentification object storing IDs and groups
peptides	vector of ProteinIdentifications with links to the proteins

findConnectedComponent()

ConnectedComponent findConnectedComponent

(

Size &

root_prot_grp

)

Does a BFS on the two maps (= two parts of the graph; indist. prot. groups and peptides), switching from one to the other in each step.

Parameters

root_prot_grp

Starts the BFS at this protein group index

Returns

Returns a Connected Component as set of group and peptide indices.

resolveConnectedComponent()

void resolveConnectedComponent	(	ConnectedComponent &	conn_comp,
		ProteinIdentification &	protein,
		std::vector< PeptideIdentification > &	peptides
	)

Resolves connected components based on Fido probabilities and adds them as additional protein_groups to the output idXML. Thereby greedily assigns shared peptides in this component uniquely to the proteins of the current BEST INDISTINGUISHABLE protein group, ready to be used in ProteinQuantifier then. This is achieved by removing all other evidence from the input PeptideIDs and iterating until each peptide is uniquely assigned. In accordance with Fido only the best hit (PSM) for an ID is considered. Probability ties are _currently_ resolved by taking the first occurrence.

Parameters

conn_comp	The component to be resolved
protein	ProteinIdentification object storing IDs and groups
peptides	vector of ProteinIdentifications with links to the proteins

resolveGraph()

void resolveGraph	(	ProteinIdentification &	protein,
		std::vector< PeptideIdentification > &	peptides
	)

Applies resolveConnectedComponent to every component of the graph and is able to write statistics when specified. Parameters will both be mutated in this method.

Parameters

protein	ProteinIdentification object storing IDs and groups
peptides	vector of ProteinIdentifications with links to the proteins

Member Data Documentation

indist_prot_grp_to_pep_

IndexMap_ indist_prot_grp_to_pep_

private

mapping indist. protein group indices -> peptide identification indices

pep_to_indist_prot_grp_

IndexMap_ pep_to_indist_prot_grp_

private

mapping indist. protein group indices <- peptide identification indices

prot_acc_to_indist_prot_grp_

std::map<String, Size> prot_acc_to_indist_prot_grp_

private

represents the middle layer of an implicit tripartite graph: consists of single protein accessions and their mapping to the (indist.) group's indices

statistics_

bool statistics_

private

log debug information?