Skip to main content
Fig. 1 | Genome Biology

Fig. 1

From: Performance difference of graph-based and alignment-based hybrid error correction methods for error-prone long reads

Fig. 1

Illustration of alignment-based and graph-based method; results for model fitness and accuracy gain on simulated data. a Schematic of alignment-based method. b is a certain base on the long read, and b′ is the corresponding base on the reference sequence. The C real short reads are aligned to the long read (with N of them being successfully aligned), and then the consensus is inferred at each base. b Relationship of the successful alignment probability for short reads τ with the mismatch rate p, lower threshold on perfect match k-mer size k and the upper threshold of mismatches m. In spite of the changes of k or/and m, τ is near to one when p < 5%, and is near to zero when p > 30%. This indicates that mismatch rate is the most dominant factor on τ. As m increases from 10 to 20, the curves move upper (from blue to red and green), implying that τ increases with m. Moreover, the divergence between the dashed and solid blue, red, and green lines also shows an increasing tendency, which means the effect of k on τ also increases with m. c Schematic of graph-based error correction method. DBG is built based on short reads. Solid k-mers are detected on the long reads. The fragment between two adjacent solid k-mers is then aligned with the correlated path on the DBG. The path is used to correct the fragment when certain criteria are satisfied. d Accuracy gain at each error rate for simulated long reads corrected by alignment-based method. The boxplots represent the accuracy gain distribution for long reads. The solid lines represent the theoretical values. The dashed gray lines (diagonal lines) correspond to perfect correction. e Proportion of simulated long reads with solid k-mer detected at each error rate level. The solid lines represent the theoretical values. The dashed lines represent the results on simulated long reads. f Accuracy gain at each error rate for simulated long reads corrected by graph-based method. L: long read length; k: size of perfectly matched seed or solid k-mer

Back to article page