Fig. 3

Accuracy assessment and simulated error rate of LoMA. A. A comparison with raw reads (blue) and sequences by lamassemble (orange). Consensus sequences (CSs) by LoMA (green) showed a sharp drop in the error rate compared with raw reads and lamassemble. The standard deviations were 0.67, 4.09 and 0.72 for LoMA, lamassemble and raw reads, respectively, demonstrating that the dispersion of LoMA is smaller than of lamassemble and that the instability of sequences in raw data is reduced by LoMA. B. A comparison of CSs constructed by LoMA with sequences by Sanger sequencing. The sequences from 121 insertions of NA18943 showed high accuracy (at least 99.7%). Individually, 66 out of 121 (54.5%) regions constructed by LoMA were error-free compared with poly(A) compressed sequences. C. The average error rate of simulated data (n = 100). Simulated reads were generated for one hundred randomly selected genomic regions and analyzed by LoMA. The average error rates were 2.6% (10×), 0.29% (20×), 0.076% (30×), 0.041% (40×), and 0.034% (50×) for each mean coverages from 10 to 50