Using an ensemble of statistical metrics to quantify large sets of plant transcription factor binding sites

Table 1 Statistical metrics

Metric	Equation	Output range	Reference
Confidence (CF)	m a x(P(G_a\|t_i),P(t_i\|G_a))	0…1	[26]
Cosine (CO)	$\frac{P (t_{i}, G_{a})}{\sqrt{P (t_{i}) P (G_{a})}}$	$0 \dots \sqrt{P (t_{i}, G_{a})} \dots 1$	[27]
Jaccard (JAC)	$\frac{P (t_{i}, G_{a})}{P (t_{i}) + P (G_{a}) - P (t_{i}, G_{a})}$	0…1	[28]
Kappa coefficient (K)	$\frac{P (t_{i}, G_{a}) + P (\bar{t_{i}, G_{a}}) - P (t_{i}) P (G_{a}) - P (\bar{t_{i}}) P (\bar{G_{a}}))}{1 - P (t_{i}) P (G_{a}) - P (\bar{t_{i}}) P (\bar{G_{a}})}$	−1…1	[29]
Laplace Correction (LP)	$max (\frac{NP (t_{i}, G_{a}) + 1}{NP (t_{i}) + 2}, \frac{NP (t_{i}, G_{a}) + 1}{NP (G_{a}) + 2})$	0…1	[30]
Lift (LI)	$\frac{P (t_{i}, G_{a})}{P (t_{i}) P (G_{a})}$	0…∞	[31]
Phi coefficient (PHI)	$\frac{P (t_{i}, G_{a}) - P (t_{i}) P (G_{a})}{\sqrt{P (t_{i}) P (G_{a}) (1 - P (t_{i})) (1 - P (G_{a}))}}$	−1…1	[32]

Given a group, G_a, and a TFBS, t_i, magnitude of TFBS over-representation can be determined using a set of statistical metrics.

ISSN: 1746-4811