Figure 2

Correlation of PMD annotation and FoldX predictions. To gauge the best ΔΔG cutoff for annotating a mutation as neutral stability-wise we correlated PMD binary annotations with FoldX predictions. Overall correlation was relatively weak, pointing out the subjectivity of annotations, possible problems of curation, and FoldX prediction errors. The best correlation between binary annotation and FoldX predictions was at -0.5 kcal/mol for stabilizing mutations (open squares) and 2.5 kcal/mol for destabilizing ones (filled squares).