Fig. 2From: Comprehensive benchmark and architectural analysis of deep learning models for nanopore sequencing basecallingLatest basecallers benchmark on the human task. a Basecalling reads failure rates: pass, failed mapping, short alignment, and no prediction (from top top bottom). b Alignment event rates: match, mismatch, insertion, and deletion (from top to bottom). c Homopolymer error rates per base. d PhredQ scores distributions for correctly (light) and incorrectly (dark) predicted bases. e AUC of match rate sorted by average read PhredQ scoreBack to article page