ANDES: Statistical tools for the ANalyses of DEep Sequencing

BMC Research Notes

Table 2 Time (user + system) in seconds for generating profiles with various partitions sizes on 2900 avian MP segments

Number of Partitions	1	2	4	10
Number of Sequences/Partition	2900	1450	725	290
Sequence Alignment Times:	3179.73	882.88	282.43	74.37
(with clustalw2 -quicktree)		896.14	284.10	70.70
			275.90	70.83
			274.19	70.46
				71.59
				71.06
				70.69
				71.02
				71.28
				71.04
Parallel Time (max)	3179.73	896.14	284.10	74.37
Serial Time (sum)	3179.73	1779.02	1116.62	713.04
Profile Generation Times:	11.50	7.61	2.76	1.11
		6.15	2.74	1.17
			2.93	1.09
			2.86	1.14
				1.13
				1.12
				1.16
				1.10
				1.17
				1.24
Parallel Time (max)	11.50	7.61	2.93	1.24
Serial Time (sum)	11.50	13.76	11.29	11.43
Profile Merge Times:	0.00	0.30	0.63	1.57
Parallel Time (max)	0.00	0.30	0.63	1.57
Serial Time (sum)	0.00	0.30	0.63	1.57
Total Parallel Times	*3191.23*	*904.05*	*287.66*	*77.18*
Total Serial Times	*3191.23*	*1793.08*	*1128.54*	*726.04*
Parallel Speed Up Factor	1.00	3.53	11.09	41.35
Serial Speed Up Factor	1.00	1.78	2.83	4.40

2900 avian MP segments were randomly selected without replacement from the available 2913 segments in Genbank. The sequences were then partitioned into 2, 4, and 10 subsets to demonstrate the effect of partitioning on profile generation time and accuracy. Individual compute times for sequence alignment, profile generation, and profile merge were recorded for each job. Parallel and serial compute times for sequence alignment and profile generation were computed by finding the maximum and sum, respectively, across individual job times. Total parallel and serial compute times for each partition were calculated by summing up the parallel and serial compute times, respectively, across sequence alignment, profile generation, and profile merge. When partitioned into 10 subsets, serial and parallel speedup factors of 4.4 and 41.35 were observed, respectively.

ISSN: 1756-0500