Profile hidden Markov model sequence analysis can help remove putative pseudogenes from DNA barcoding and metabarcoding datasets

Table 2 Sensitivity and specificity for two pseudogene filtering methods

Experiment	Dataset	Type of mutations	Sensitivity (%)		Specificity (%)
Experiment	Dataset	Type of mutations	ORFfinder	ORFfinder + profile HMM analysis	ORFfinder	ORFfinder + profile HMM analysis
Artificial DNA barcoding dataset. COI genes and nuMTs from 10 species	Full length COI barcode and nuMT sequences	N/A	70	73	90	90
Perturbed community dataset	Full length COI barcode and simulated nuMTs	GC—> AT	31	27	99	~ 100
Perturbed community dataset	Full length COI barcode and simulated nuMTs	Frameshift	88	94	~ 100	~ 100
Perturbed community dataset	Short COI barcode and simulated nuMTs	GC—> AT	17*—50	6*—15	99	~ 100
Perturbed community dataset	Short COI barcode and simulated nuMTs	Frameshift	42*—58	61*—87	99	99—~ 100*
Perturbed community dataset	Full length COI barcode and twice as many nuMTs	GC—> AT	17	0	99	~ 100
Perturbed community dataset	Full length COI barcode and twice as many nuMTs	Frameshift	0	0	~ 100	~ 100
Perturbed community dataset	Full length COI barcode and half as many nuMTs	GC—> AT	39	36	95	96
Perturbed community dataset	Full length COI barcode and half as many nuMTs	Frameshift	95	98	96	99

Sensitivity refers to the true positive rate, our ability to correctly identify known or simulated nuMTs. Specificity refers to the true negative rate, our ability to correctly identify COI genes. * 5’ fragment. ** 3’ fragment

ISSN: 1471-2105