Adaptable probabilistic mapping of short reads using position specific scoring matrices

BMC Bioinformatics

Table 3 Analysis of simulated biased single-end data

		Unfiltered		MapQ filtered
	Mapper	Sensitivity	PPV	Sensitivity	PPV	Time (s)
a) ART single-end length 36 / PAR-CLIP
	BWA-PSSM^PC	0.699	0.881	0.662	0.996	68.84
	BWA-PSSM	0.642	0.865	0.594	0.990	81.89
	BWA	0.628	0.866	0.582	0.996	56.05
	BWA-MEM	0.451	0.844	0.388	0.999	74.78
	Bowtie	0.689	0.870	*	*	32.21
	Bowtie2	0.594	0.845	0.419	0.992	25.91
	GEM	0.475	0.980	*	*	39.48
b) ART single-end length 55 / Ancient DNA
	BWA-PSSM^A	0.807	0.941	0.774	0.997	122.16
	BWA-PSSM	0.797	0.937	0.766	0.996	115.17
	BWA	0.743	0.935	0.703	0.998	90.33
	BWA-MEM	0.817	0.924	0.725	1.000	50.89
	Bowtie	0.807	0.934	*	*	28.55
	Bowtie2	0.788	0.916	0.665	0.999	57.76
	GEM	0.691	0.993	*	*	30.62
c) ART single-end length 100 / *P. falciparum*
	BWA-PSSM	0.899	0.975	0.886	1.000	86.60
	BWA	0.332	0.974	0.325	0.999	59.20
	BWA-MEM	0.824	0.840	0.786	0.868	34.56
	Bowtie	0.925	0.976	*	*	12.91
	Bowtie2	0.832	0.972	0.712	1.000	34.33
	GEM	0.726	1.000	*	*	13.84

Comparison of sensitivity, positive predictive value (PPV) and run time using BWA-PSSM, BWA, BWA-MEM, Bowtie, Bowtie2 and GEM on data sets simulated using the ART_illumina [37] program with the default error profile for the given read length. The PAR-CLIP and Ancient DNA data sets cover a random 1% of the human genome, while the P. falciparum data set covers 42.9% of the P. falciparum genome corresponding to 100,000 reads. The data simulated for the PAR-CLIP and Ancient DNA data was further mutated to simulate the bias introduced by the experiment and natural degradation (see the respective Results and discussion sections). The symbol ’*’ indicates that the mapper does not provide MapQ scores.

ISSN: 1471-2105