Fig. 2.

INC-Seq evaluation with simulated reads. a 100 representative reference sequences were selected from a customized database. Simulated reads were generated for each reference sequence, as well as artificial RCA products (that is, INC-Seq libraries). Synthetic INC-Seq reads significantly improved the proportion of correctly mapped reads over raw ONT 2D reads at both species and reference level (p-value <10^-9; one-sided paired Wilxcon test). b Species-level classification F1 score (that is, the harmonic mean of precision and recall) using simulated raw, CANU corrected and INC-Seq reads from 100 reference sequences and with different identity thresholds for classification (each curve represents the average across 10 replicates)