Table 1:

Results from the experimental validation and benchmarking of Chiron against three segmentation-based nanopore basecallers and Albacore v2, which is also a segmentation-free basecaller

Dataset	Basecaller	Deletion rate (%)	Insertion rate (%)	Mismatch rate (%)	Identity rate (%)	Error rate (%)
	Metrichor	8.93	2.38	4.57	86.50	15.88
	Albacore v1.1	6.35	3.82	4.69	88.96	14.86
	Albacore v2	6.19	3.38	3.98	89.82	13.55
Lambda	BasecRAWller	7.89	10.01	10.56	81.54	28.46
	Chiron	8.20	2.13	4.03	87.76	14.36
	Chiron-BS	6.20	2.13	4.20	89.60	12.53
	Metrichor	7.52	1.93	3.84	88.64	13.29
	Albacore v1.1	5.76	3.27	4.14	90.10	13.17
	Albacore v2	5.21	2.99	3.57	91.22	11.77
E. coli	BasecRAWller	7.16	10.40	10.30	82.54	27.86
	Chiron	6.36	1.81	3.07	90.57	11.24
	Chiron-BS	4.94	2.36	3.16	91.90	10.46
	Metrichor	7.63	2.40	4.35	88.02	14.38
	Albacore v1.1	6.12	3.57	4.68	89.19	14.37
	Albacore v2	5.05	3.58	4.05	90.90	12.68
M. tuberculosis	BasecRAWller	7.17	10.85	10.42	82.41	28.44
	Chiron	7.16	2.50	4.33	88.51	13.99
	Chiron-BS	5.84	3.05	4.50	89.66	13.39
	Metrichor	12.95	4.15	7.65	79.4	24.75
	Albacore v1.1	8.62	6.51	7.52	83.86	22.65
	Albacore v2	8.71	6.03	6.05	85.24	20.79
Human	BasecRAWller	8.41	10.28	10.10	81.49	28.79
	Chiron	9.13	5.14	9.33	81.54	23.60
	Chiron-BS	9.30	5.62	7.87	82.83	22.79

Deletion, insertion, and mismatch rates (%) are defined as the number of deleted, inserted, and mismatched bases divided by the number of bases in the reference genome (the lower the better). Identity rate (%) is defined as the number of matched bases divided by the number of bases in the reference genome for that sample (the higher the better; identity rate = 1 - deletion rate - mismatch rate). Error rate (%) is defined as the sum of deletion, insertion, and mismatch rates (the lower the better; error rate = deletion rate + insertion rate + mismatch rate). This statistic effectively summarizes the basecalling accuracy of the associated model. The best result in each category is indicated in bold.