LocARNA::ConfusionMatrix Class Reference

Compare RNA secondary structure by their confusion matrix. More...

#include <confusion_matrix.hh>

List of all members.

Public Member Functions
	ConfusionMatrix (const std::string &ref, const std::string &pred, const bool slide, const bool conflict, const BasePairFilter::Filter &filter=BasePairFilter::BPMinLoopSize(3))
	ConfusionMatrix (const RnaStructure &ref, const RnaStructure &pred, const bool slide, const bool conflict, const BasePairFilter::Filter &filter=BasePairFilter::BPMinLoopSize(3))
size_t	tp () const
	True positives.
size_t	tn () const
	True negatives.
size_t	fp () const
	False positives.
size_t	fn () const
double	ppv () const
double	sens () const
double	spec () const
double	f1_score () const
double	mcc () const
Protected Member Functions
size_t	count_common_bps (const RnaStructure &s1, const RnaStructure &s2)
	Count common base pairs.
size_t	count_tps (const RnaStructure &pred, const RnaStructure &ref)
	Count true positive base pairs.
size_t	count_conflicting_base_pairs (const RnaStructure &s1, const RnaStructure &s2)
	Count conflicting base pairs (including common bps)
size_t	count_potential_base_pairs (size_t length)
	Count potential base pairs.
size_t	count_base_pairs (const RnaStructure &s)
	Count base pairs in a structure.

---

Detailed Description

Compare RNA secondary structure by their confusion matrix.

Computes confusion matrix and Matthews' correlation coefficient

Base pair filters provide a generic way to count only a subset of potential base pairs, e.g. only canonical base pairs, or base pairs in a certain range l..u

Implements slide rule and conflict rule; see Paul special

Slide rule: tolerate shift of one base pair end by 1 position (affects true positives and false negatives)

Conflict rule: predicted base pairs are false only if they conflict with reference; two base pair conflict iff they share one end (affects "false positives")

---

Constructor & Destructor Documentation

LocARNA::ConfusionMatrix::ConfusionMatrix	(	const std::string &	ref,
		const std::string &	pred,
		const bool	slide,
		const bool	conflict,
		const BasePairFilter::Filter &	filter = BasePairFilter::BPMinLoopSize(3)
	)

Construct with reference and predicted structure (given as dot-bracket strings)

Parameters:

ref	reference structure
pred	predicted structure
slide	use slide rule [default off]
conflict	use conflict rule [default off]
filter	base pair filter

LocARNA::ConfusionMatrix::ConfusionMatrix	(	const RnaStructure &	ref,
		const RnaStructure &	pred,
		const bool	slide,
		const bool	conflict,
		const BasePairFilter::Filter &	filter = BasePairFilter::BPMinLoopSize(3)
	)

Construct with reference and predicted structure (given as base pair sets)

Parameters:

ref	reference structure
pred	predicted structure
slide	use slide rule [default off]
conflict	use conflict rule [default off]
filter	base pair filter

---

Member Function Documentation

size_t LocARNA::ConfusionMatrix::count_base_pairs

(

const RnaStructure &

s

)

[protected]

Count base pairs in a structure.

Parameters:

structure

RNA structure

Returns:

number of base pairs

size_t LocARNA::ConfusionMatrix::count_common_bps	(	const RnaStructure &	s1,
		const RnaStructure &	s2
	)		[protected]

Count common base pairs.

Parameters:

s1	first structure
s2	second structure

Returns:

number of base pairs in the first structure that are as well in the second (optionally according to slide rule)

Note:

due to slide rule this is not symmetric

size_t LocARNA::ConfusionMatrix::count_conflicting_base_pairs	(	const RnaStructure &	s1,
		const RnaStructure &	s2
	)		[protected]

Count conflicting base pairs (including common bps)

Parameters:

s1	first structure
s2	second structure

Two base pairs (i,j) and (i',j') conflict if and only if they share exactly one end.

Returns:

number of base pairs in s1 that conflict with s2

size_t LocARNA::ConfusionMatrix::count_potential_base_pairs

(

size_t

length

)

[protected]

Count potential base pairs.

Parameters:

length

length of RNA

Returns:

number of potential base pairs

size_t LocARNA::ConfusionMatrix::count_tps	(	const RnaStructure &	pred,
		const RnaStructure &	ref
	)		[protected]

Count true positive base pairs.

Parameters:

pred	predicted structure
ref	reference structure

Returns:

number of true positive base pairs

double LocARNA::ConfusionMatrix::f1_score

(

)

const

F1 score (aka F-score, F-measure)

harmonic mean of PPV and SENS

Returns:

F1 = PPV*SENS / (PPV+SENS), if PPV+SENS!=0; 0, otherwise

size_t LocARNA::ConfusionMatrix::fn

(

)

const [inline]

False negatives

Returns:

number of false negative base pairs

Note:

A false negative is pair annotated to basepair but not predicted.

size_t LocARNA::ConfusionMatrix::fp

(

)

const [inline]

False positives.

Returns:

number of false positive base pairs

Note:

A false positive is a pair predicted to basepair but not annotated.

double LocARNA::ConfusionMatrix::mcc

(

)

const

Matthews' correlation coefficient

Returns:

MCC = (TP*TN - FP*FN) / sqrt( (TP+FP)*(TP+FN)*(TN+FP)*(TN+FN) )

double LocARNA::ConfusionMatrix::ppv

(

)

const

Positive prediction value

aka precision

Returns:

PPV = TP/(TP+FP)

double LocARNA::ConfusionMatrix::sens

(

)

const

Sensitivity

aka recall

Returns:

SENS = TP/(TP+FN)

double LocARNA::ConfusionMatrix::spec

(

)

const

Specificity

Returns:

SPEC = TN/(TN+FP)

size_t LocARNA::ConfusionMatrix::tn

(

)

const [inline]

True negatives.

Returns:

number of true negative base pairs

Note:

A true negative is a pair which is neither annotated nor predicted to basepair

size_t LocARNA::ConfusionMatrix::tp

(

)

const [inline]

True positives.

Returns:

number of true positive base pairs

Note:

A true positive is a pair that is basepaired in both prediction and annotation; this is softened with slide rule, see ().

---

The documentation for this class was generated from the following files:

• src/LocARNA/confusion_matrix.hh
• src/LocARNA/confusion_matrix.cc