BMC Bioinformatics

Table 11 SNP calling on the M. musculus dataset with and without compression

From: QualComp: a new lossy compressor for quality scores based on rate distortion theory

				One cluster
R	MSE	T.P.	F.P	F.N.	Selectivity (%)	Sensitivity (%)	Size (MB)
0	143.0	11217	2033	1810	84.66	86.11	0.019
0.20	19.16	12585	1159	442	91.57	96.61	16.17
0.33	16.46	12602	1120	380	91.84	97.07	26.68
0.66	12.94	12669	998	358	92.70	97.25	53.34
1.00	10.01	12656	875	371	93.53	97.15	80.82
2.00	4.58	12733	594	294	95.54	97.74	161.62
				Two clusters
R	MSE	T.P.	F.P	F.N.	Selectivity (%)	Sensitivity (%)	Size (MB)
0	37.58	12086	1534	941	88.73	92.77	0.039
0.20	16.42	12644	1184	383	91.44	97.06	16.19
0.33	14.39	12655	1107	372	91.95	97.14	26.70
0.66	11.00	12669	985	358	92.78	97.25	53.36
1.00	8.59	12687	830	340	93.85	97.39	80.84
2.00	3.76	12751	606	276	95.46	97.88	161.64
				Three clusters
R	MSE	T.P.	F.P	F.N.	Selectivity (%)	Sensitivity (%)	Size (MB)
0	27.49	12048	1219	979	90.81	92.48	0.050
0.20	14.89	12638	1108	389	91.93	97.01	16.21
0.33	13.09	12645	1070	382	92.19	97.06	26.98
0.66	9.82	12646	909	381	93.29	97.07	53.91
1.00	7.35	12685	776	342	94.23	97.37	80.85
2.00	3.12	12730	554	297	95.83	97.72	161.65

We compare the SNPs detected by Samtools with the original FASTQ file and those obtained with the compressed files, using QualComp with one, two and three clusters and different rates. In all cases, reads were aligned first using the BWA algorithm. T.P., F.P. and F.N. stand for true positive (detected both with the original FASTQ file and the reconstructed one), false positive (detected only with the reconstructed FASTQ file) and false negative (detected only with the original FASTQ file), respectively. The selectivity parameter is computed as T.P./(T.P. + F.P.), and sensitivity as T.P./(T.P. + F.N.). Note that already for R = 0.2 the sensitivity is above 96% and the selectivity is above 91%.

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com