Fig. 3: Validation on highly mismatched targets and independent HiTS-FLIP data.

a Validation data (upper-left triangle) for effective association constant (CHAMP) on block-mismatched targets, and model estimates (lower-right triangle). The two terminal mismatch positions in the block are marked on the axes. b Correlation plot between measured effective association constants and model predictions on block-mismatched targets. c Validation data (triangles) for association rates (HiTS-FLIP data set¹¹) on single-mismatch targets, and model estimates (line). d Validation data (upper-left triangle) for association rates on double-mismatch targets, and model estimates (lower-right triangle). e Correlation plot for all positive association rates, including moderately (1–2 mismatches, dark purple) and highly (3–20 mismatches, light purple) mismatched targets. f Validation data (triangles) for dissociation rates (HiTS-FLIP data set¹¹) on single-mismatch targets, and model estimates (line). The missing mismatch-averaged dissociation rates in the seed are negative. g Validation data (upper-left triangle) for dissociation rates on double-mismatch targets, and model estimates (lower-right triangle). h Correlation plot for all positive dissociation rates, including moderately (1–2 mismatches, dark green) and highly (3–20 mismatches, light green) mismatched targets. Mismatch-averaged rates dominated by negative scores are excluded from the analysis, and all data is averaged over mismatch type (see Methods and Supplementary Data 1). The quoted correlation coefficients are Pearson-correlation coefficients, and correlation plots are displayed with log-scales to show the quantitative agreement also for weak targets. The dashed lines in the correlation plots correspond to perfect quantitative prediction.