Using the nucleotide substitution rate matrix to detect horizontal gene transfer

Hamady, Micah; Betterton, M D; Knight, Rob

doi:10.1186/1471-2105-7-476

BMC Bioinformatics

Table 1 Statistics used to detect changes in the rate matrix.

From: Using the nucleotide substitution rate matrix to detect horizontal gene transfer

#	Type	Sequences	Normalized	Centered	SV measure	Centered SVD method	Rank	%
1	SVD	2	No	Yes	a	Correlation	44	0.4
2	SVD	2	No	Yes	b	Correlation	40	0.5
3	SVD	2	No	Yes	c	Correlation	12	2.5
4	SVD	2	No	Yes	a	Covariance	41	0.5
5	SVD	2	No	Yes	b	Covariance	42	0.5
6	SVD	2	No	Yes	c	Covariance	37	0.6
7	SVD	2	No	No	a	_	36	0.6
8	SVD	2	No	No	b	_	39	0.5
9	SVD	2	No	No	c	_	35	0.8
10	Mean(d)	2	No	-	-	-	34	0.9
11	Var(d)	2	No	-	-	-	22	1.8
12	SVD	2	Yes	Yes	a	Correlation	23	1.8
13	SVD	2	Yes	Yes	b	Correlation	43	0.5
14	SVD	2	Yes	Yes	c	Correlation	32	1.2
15	SVD	2	Yes	Yes	a	Covariance	14	2.2
16	SVD	2	Yes	Yes	b	Covariance	45	0.3
17	SVD	2	Yes	Yes	c	Covariance	18	2.0
18	SVD	2	Yes	No	a	_	28	1.5
19	SVD	2	Yes	No	b	_	21	1.9
20	SVD	2	Yes	No	c	_	31	1.4
21	Mean(d)	2	Yes	-	-	-	24	1.7
22	Var(d)	2	Yes	-	-	-	33	1.2
23	SVD	3	No	Yes	a	Correlation	7	3.9
24	SVD	3	No	Yes	b	Correlation	26	1.6
25	SVD	3	No	Yes	c	Correlation	20	1.9
26	SVD	3	No	Yes	a	Covariance	6	3.9
27	SVD	3	No	Yes	b	Covariance	19	1.9
28	SVD	3	No	Yes	c	Covariance	2	6.9
29	SVD	3	No	No	a	_	8	3.8
30	SVD	3	No	No	b	_	30	1.5
31	SVD	3	No	No	c	_	13	2.3
32	Mean(d)	3	No	-	-	-	3	6.8
33	Var(d)	3	No	-	-	-	10	3.3
34	SVD	3	Yes	Yes	a	Correlation	5	5.1
35	SVD	3	Yes	Yes	b	Correlation	15	2.2
36	SVD	3	Yes	Yes	c	Correlation	4	5.7
37	SVD	3	Yes	Yes	a	Covariance	1	7.7
38	SVD	3	Yes	Yes	b	Covariance	29	1.5
39	SVD	3	Yes	Yes	c	Covariance	16	2.1
40	SVD	3	Yes	No	a	-	9	3.5
41	SVD	3	Yes	No	b	-	27	1.6
42	SVD	3	Yes	No	c	-	38	0.6
43	Mean(d)	3	Yes	-	-	-	17	2.1
44	Var(d)	3	Yes	-	-	-	25	1.7
45	Var(GC)	-	-	-	-	-	11	3.2

Table of 45 statistics derived from collections of rate matrices. Columns are: #: number of each method, used in the text for reference. Type: type of statistic, either SVD, mean distance of each Q_ifrom the average of all Q_i(Mean(d)), variance of the distances of the Q_ifrom the average of all Q_i(Var(d)), or variance in GC content (Var(GC)). Sequences: either 2, for pairwise sequence comparisons that assume a time-reversible substitution model, or 3, for three-way sequence comparisons that do not assume a time-reversible model. Normalized: either No, for unnormalized Q_i, Yes, for Q_inormalized to a trace of 1 (eliminating the contribution of time to the inferred matrix), or N/A, where normalization was not applicable. Centered: Yes, for CSVD, No, for USVD, or N/A, where SVD was not used. SV measure: statistic for characterizing singular values, either a, for ratio of the two largest singular values, or b, for ratio of the largest singular value to the sum of all singular values, or c, for ∑_iln(1 + σ_i), where the σ_iare the singular values (this is the SVD version of the statistic introduced by Weiss et al. [43]), or N/A, where not applicable. CSVD method: Covariance if the covariance matrix was used for CSVD, Correlation if the correlation matrix was used, N/A if not applicable (i.e., if the technique was not centered SVD). Rank: rank of the statistic in contributing to overall classification accuracy using random forests [75]. %: percentage contribution of the statistic to overall classification accuracy using random forests. The statistic used by Devauchelle et al. [42] corresponds to statistic 7; that used by Weiss et al. [43] corresponds to statistic 9.

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com