Skip to main content

Table 3 Timing profile of variational EM algorithm when median depth is 3,089×

From: Variational inference for rare variant detection in deep, heterogeneous next-generation sequencing data

  

E-step

M-step

 

Computation

Region

Optimize

Optimize

Update

Optimize

Optimize

Optimize

Update

Total

resource

length

γ

δ

ELBO

μ 0

M 0

M

ELBO

time (s)

Single processor

100

617.7 (63%)

4.232

10.42

0.264

0.159

332.8 (34%)

10.29

976.0

 

200

1124 (65%)

8.936

18.64

0.418

0.256

570.0 (33%)

18.37

1741

 

300

1728 (65%)

13.27

27.81

0.445

0.400

851.5 (32%)

27.65

2649

 

400

2433 (66%)

17.99

38.55

0.737

0.635

1176 (32%)

38.17

3705

60 processors

100

29.93 (41%)

0.2470

11.67

0.3070

0.1890

19.56 (26%)

11.98

73.89

 

200

44.69 (40%)

0.4170

22.14

0.5230

0.3040

24.04 (21%)

22.24

114.3

 

300

63.47 (40%)

0.7160

33.31

0.5620

0.5040

29.41 (18%)

33.24

161.2

 

400

94.66 (43%)

0.7270

42.78

0.8200

0.7060

40.04 (18%)

44.28

219.7

  1. Timing profile of 4 significant figures for one iteration of variational EM algorithm when median read depth is 3,089×. Single and multiple processors are both tested to estimate computational efficiency. Time for optimizing γ in the E-step and optimizing M in the M-step is highlighted in percentage
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us