canvasFPMatrix

This utility generates a pairwise similarity or distance matrix using binary or scaled fingerprints from one or two sets of molecules. See canvasFPMatrix Command Help for syntax and options.

The metrics are listed in Table 1, along with their type and formula. The quantities in the formula are defined in Table 2, except for the Tversky α and β parameters.

Table 1. Available metrics for the canvasFPMatrix command
Index	Name	Type	Formula
1	`buser`	similarity	(sqrt(cd)+c)/(sqrt(cd)+a+b-c)
2	`cosine`	similarity	c/sqrt(ab)
3	`dice`	similarity	2c/(a+b)
4	`dixon`	distance	(A+B)^2/(a+b-c)
5	`euclidean`	distance	sqrt(A+B)
6	`hamann`	similarity	(c+d-A-B)/N
7	`hamming`	distance	A+B
8	`kulczynski`	similarity	0.5(c/a + c/b)
9	`matching`	similarity	(c+d)/L
10	`mcConnaughey`	similarity	(c^2-(a-c)(b-c))/(ab)
11	`minmax`	similarity	sum{min(a,b)/max(a,b)}
12	`modifiedTanimoto`	similarity	α T1 + (1.0-α)T0
13	`patternDifference`	distance	AB/N^2
14	`pearson`	similarity	(cd-AB)/sqrt(ab(A+d)(B+d))
15	`petke`	similarity	c/max(a,b)
16	`rogersTanimoto`	similarity	(c+d)/(2(a+b)-3c+d)
17	`shape`	distance	(A+B)/N - ((A-B)/N)^2
18	`simpson`	similarity	c/min(a,b)
19	`size`	distance	((A-B)/N)^2
20	`soergel`	distance	(A+B)/(A+B+c)
21	`tanimoto`	similarity	c/(a+b-c)
22	`tversky`	similarity	c/(α(a-c)+β(b-c)+c)
23	`variance`	distance	(A+B)/(4N)
24	`yule`	similarity	(cd - AB)/(cd + AB)

Table 2. Variables used in metric formulae
Variable	Description
a	Number of bits that are on in structure 1
b	Number of bits that are on in structure 2
c	Number of bits that are on in both structure 1 and structure 2
d	Number of bits that are off in both structure 1 and structure 2
A	Number of bits that are on in structure 1 but not in structure 2. A = a - c
B	Number of bits that are on in structure 2 but not in structure 1. B = b - c
L	Total number of bits. L = a + b - c + d
N	Restricted total number of bits. N = a + b - c + min(d,10000)