Table 2.

Robustness of barcode error correction, measured through simulated data

Edit distance		Positive (%)	Negative (%)
1	True	99.999	67.281
	False	0.001	32.721
2	True	99.983	74.971
	False	0.017	25.033
3	True	99.710.	77.972
	False	0.290.	22.033

The barcodes of 400,000 simulated duplex reads were corrected with Du Novo 2.15 with three different edit distance thresholds: 1, 2, and 3. Corrected and uncorrected barcodes were compared to the original, true barcode sequences. For each corrected barcode, if the correction assigned it to its true family, this was counted as a true positive. Otherwise it was a false positive. Uncorrected barcodes which were not one of the original, true barcodes were counted as false negatives. The rest of the uncorrected barcodes were counted as true negatives. Each family was counted once, rather than each raw read