All Your Base: a fast and accurate probabilistic approach to base calling

Table 5 Assessment of quality score accuracy across all data sets for the three base callers under a variety of criteria; see text for details.

		${\bar{S}}_{p}$			${\bar{S}}_{p}$ Max				RMS		Q _tot, ×10⁹
	Bustard	Ibis	AYB	AYB*	Bustard	Ibis	AYB	Bustard	Ibis	AYB	Bustard	Ibis	AYB
ϕX174 L2	-35.0	23.4	31.5		26.1	24.7	32.1	12.4	3.6	2.3	1.0	1.0	1.2
ϕX174 L4	9.5	22.3	30.1		25.8	23.0	31.0	8.8	2.4	2.5	1.6	1.5	1.9
ϕX174 L6	-5.6	23.2	31.5		26.8	24.0	32.2	10.1	2.9	2.3	1.3	1.2	1.6
Ibis Test	N/A	23.6	30.7		N/A	24.0	32.0	N/A	1.7	3.2	N/A	0.2	0.3
B. pert./1	1.3	21.3	12.0	22.1	20.3	22.5	25.5	8.1	1.8	5.9	1.9	2.8	3.6
... trimmed	-1.5	17.0	23.0	24.5	23.2	20.3	27.0	9.1	3.1	4.4	3.8	3.5	4.5
B. pert./2	4.9	16.4	23.9	24.2	22.4	20.3	26.6	9.6	1.4	3.7	2.5	3.1	4.4
... trimmed	-3.7	14.0	23.6	24.5	22.3	19.9	26.7	9.3	2.6	4.0	3.1	3.0	3.9
BGI/1	-29.0	21.6	0.0	22.9	22.7	22.2	24.0	11.3	2.5	8.7	8.6	8.5	8.8
BGI/2	-9.8	20.6	1.0	22.3	22.3	21.2	23.6	11.1	2.5	8.5	8.2	7.9	8.4
Illumina/1	-1.4	22.8	1.3	23.8	24.0	23.2	24.9	9.8	1.9	8.8	16.7	14.8	16.7
Illumina/2	3.4	22.4	3.1	23.7	23.9	22.8	24.9	9.7	1.9	8.6	16.4	15.7	16.5
HiSeq/1	-0.3	24.9	11.6	25.6	25.9	25.8	26.6	9.6	2.7	7.5	17.4	17.3	17.7
HiSeq/2	1.4	N/A	12.4	24.9	25.5	N/A	26.5	9.2	N/A	7.3	16.1	N/A	16.7

Note that higher values of ${\bar{S}}_{p}$ and ${\bar{S}}_{p}$ Max and lower values of RMS are better. The scores for AYB were calculated using a calibration function with parameters derived from the ϕX174 L2 data set; the scores reported under AYB* are those achieved using run-specific calibration on the larger data sets (see text for details). Quality scores were not available for Bustard on the Ibis Test data set and Ibis failed to train on the second end of the HiSeq data.

ISSN: 1474-760X