Fig. 4From: Comprehensive benchmark and architectural analysis of deep learning models for nanopore sequencing basecallingPerformance comparison on different tasks. Comparison between the human and global task on the top 10 best performing model combinations: a alignment event rates, b homopolymer error rates, and c PhredQ scoring. d Comparison between the cross-species (green) and global task (blue). Each boxplot contains a set of species split according to the cross-species task. Darker colors indicate increasing genomic distance from the species in the training set. The darkest colors contain lambda phage and human data. e Comparison between the human, global and cross-species task only on human dataBack to article page