Fig. 4
The evaluation of the large-deletion detection capability of various TIDE analysis tools. A The detection rates of the 85 bp deletion dsDNA at various ratios are plotted on the horizontal axis. The initially expected detection rates are plotted on the vertical axis. PtWAVE all and backstep modes uses NNLS for modeling. The figure for PtWAVE all and backstep modes are identical to that in Fig. 3C. The default setting for TIDE is specified as"left boundary of alignment window = 100, decomposition window: 115–685, indel size range = 10,"and the long deletion setting for TIDE is"left boundary of alignment window = 1, decomposition window: 1–700, indel size range = 50."A linear approximation curve was drawn using the linear_ model LinearRegression. fit function in the scikit-learn module. The linear relationship was evaluated using CoD, which can reach a maximum value of 1 and may take negative values. Correlations were assessed using R, and the p value from the no-correlation test was noted. The absence of R and p values indicated that the calculation was impossible. B Table presents the accuracy, recall, precision, F1 score, and Matthews correlation coefficient (MCC) for detecting the presence of large deletions (−85 bp) in empty and large-deletion samples, as analyzed by TIDE tools that validated the samples. These metrics are calculated separately for all samples and for those with R2 > 0.8. The numbers of empty and large-deletion samples were counted only if the corresponding tools successfully analyzed them. Red text indicates the results from our PtWAVE. Orange-, green-, and purple-highlighted cells denote scores of 1.000, > 0.8, and 0.000, respectively