Evaluating deterministic motif significance measures in protein databases

Ferreira, Pedro Gabriel; Azevedo, Paulo J

doi:10.1186/1748-7188-2-16

Algorithms for Molecular Biology

Table 1 List of the motif significance measures.

From: Evaluating deterministic motif significance measures in protein databases

Symbol	Measure	Formula	Range	Type
Sn	Sensitivity	$S n (M) = \frac{T_{P}}{T_{P} + F_{N}}$	[0,1]	C
Sp	Specificity	$S p (M) = \frac{T_{N}}{T_{N} + F_{P}}$	[0,1]	C
PPV	Positive Predicted Value	$P P V (M) = \frac{T_{P}}{T_{P} + F_{P}}$	[0,1]	C
Fpr	False Positive Rate	$F p r (M) = \frac{F_{P}}{F_{P} + T_{N}}$	[-1,1]	C
F	F-Measure	$F (M) = \frac{2 \times S e n s i t i v i t y \times P P V}{S e n s i t i v i t y + P P V} = \frac{2 \times T_{P}}{2 \times T_{P} + F_{N} + F_{P}}$	[0,1]	1
Corr	Correlation	$C (M) = \frac{T_{P} \times T_{N} - F_{P} \times F_{N}}{\sqrt{(T_{P} + F_{N}) (T_{P} + F_{P}) (T_{N} + F_{P}) (T_{N} + F_{N})}}$	[-1,1]	C
Dp	Discrimination Power	$D p (M) = \frac{T_{P}}{\| C \|} - \frac{F_{P}}{\| \bar{C} \|}$	[-1,1]	C
IG	Information Gain	$\begin{array}{l} I G (M) = I n f o (M) \times [S u p p o r t (M) - 1] \\ where I n f o (M) = - l o g_{\| Σ \|} p (M) \end{array}$	[0, + ∞[	IT
Pratt	Pratt Measure	$\begin{matrix} P r a t t (M) = \sum_{i}^{n} I^{'} (A_{i}) - c \cdot \sum_{k = 1}^{n - 1} (q_{k} - p_{k}) \\ where I^{'} (A_{i}) = - \sum_{a_{i} \in A_{i}} (P (a_{i}) \times l o g (P (a_{i}))) + \sum_{a_{i} \in A_{i}} (\frac{P (a_{i})}{P (A_{i})} \times l o g (\frac{P (a_{i})}{P (A_{i})})) \\ and P (A_{i}) = \sum_{a_{i} \in A_{i}} p (a_{i}) \end{matrix}$	]- ∞, + ∞[	IT
LogOdd	LogOdd	$Logodd (M) = l o g (\frac{\frac{S u p p o r t (M)}{N u m S e q s}}{P (M)})$	]- ∞, + ∞[	IT
ZScore	Z-Score	$\begin{matrix} Z s c o r e (M) = \frac{S u p p o r t (M) - E (M)}{N (M)} \\ where E (M) = N_{r e s i d} \times P (M) and N (M) = \sqrt{N_{r e s i d} \times P (M) \times (1 - P (M))} \end{matrix}$	]- ∞, + ∞[	IT
J	J-Measure	$\begin{matrix} J (C; M) = P (M) \times j (C; M) \\ where j (C; M) = P (C \| M) \times l o g_{2} \frac{P (C \| M)}{P (C)} + (1 - P (C \| M)) \times l o g_{2} \frac{(1 - P (C \| M))}{(1 - P (C))} \end{matrix}$	[0, + ∞[	H
I	Mutual Information	$\begin{matrix} I (Q; M) = H (Q) - H (Q \| M) where H (Q) = - \sum_{q \in {C, \bar{C}}} P (q) \times l o g_{2} P (q) \\ and H (Q \| M) = - P (M) \times \sum_{q \in {C, \bar{C}}} P (q \| M) \times l o g_{2} P (q \| M) \end{matrix}$	[0, 1]	H
S	Surprise Measure	$S (M) = I n f o (M) \times P (C \| M) = I n f o (M) \times \frac{S u p p o r t (M \in C)}{S u p p o r t (M)}$	[0, + ∞[	H

Description of the fourteen significance measures according to the respective type (C = Class based; IT = Information-Theoretic based; H = Hybrid). For each measure the abbreviation symbol used throughout the paper, the formula and the respective range.

Back to article page

ISSN: 1748-7188

Contact us

General enquiries: journalsubmissions@springernature.com