Table 2.

Summary of alignments inconsistent between the BWA-SW and SSAHA2 on real data

Condition	Count	BWA-SW			SSAHA2
		avgLen	avgDiff	avgMapQ	avgLen	avgDiff	avgMapQ
BWA-SW≥20; SSAHA2 unmapped	0	–	–	–	–	–	–
BWA-SW≥20 plausible; SSAHA2<20	1188	398.2	1.3%	178.4	198.3	13.1%	3.9
BWA-SW≥20 questionable	40	183.0	7.8%	41.2	280.3	9.4%	2.4
SSAHA2≥20; BWA-SW unmapped	946	–	–	–	75.4	6.3%	51.2
SSAHA2≥20 plausible; BWA-SW<20	47	129.0	9.3%	2.5	200.5	8.8%	34.4
SSAHA2≥20 questionable	185	400.2	1.7%	13.4	399.2	2.9%	216.4

A total of 137 670 454 reads uniformly selected from SRR003161 were mapped against the human genome with BWA-SW and SSAHA2, respectively. A read is said to be aligned inconsistently if the leftmost coordinates of the BWA-SW and SSAHA2 alignment differs by over 355 bp, the average read length. A score, which equals to the number of matches minus three multiplied by the number of differences (mismatches and gaps) in the aligned region, is calculated for each alignment. A BWA-SW alignment is said to be plausible if the score derived from the BWA-SW alignment minus the one derived from the SSAHA2 alignment of the same read is greater than or equal to 20 (i.e. the BWA-SW alignment is sufficiently better); otherwise the BWA-SW alignment is said to be questionable. Plausible and questionable SSAHA2 alignments are defined in a similar manner. In the table, ‘BWA-SW≥20’ denotes the BWA-SW alignments with mapping quality higher than 20. In all, BWA-SW misses 993 (=946 + 47) alignments which SSAHA2 aligns well, while SSAHA2 misses 1188; 40 BWA-SW Q20 alignments and 185 SSAHA2 Q20 alignments are possibly wrong.