Erratum: Protein Structure Prediction and Design in a Biologically Realistic Implicit Membrane (Biophysical Journal (2020) 118(8) (2042–2055), (S000634952030237X), (10.1016/j.bpj.2020.03.006))

(Biophysical Journal 118, 2042–2055; April 21, 2020) To compute the ΔΔG of mutation results (Test 2, presented in Figure 3, Supplementary Figure 8B, and Supplementary Figure 13), the score function used an fa_water_to_bilayer weight value of 1.5 instead of the reported value of 0.5. In addition, the ΔΔG results were erroneously computed allowing all residues within 8 Å of the host site to change identities (that is, to be designed). The figures were annotated with R² values, but the numbers shown are in fact the correlation coefficients, R. The script has been corrected to restrict these residues to repacking only. The updated data and correlation coefficients (fa_water_to_bilayer weight of 1.5) are presented in the updated Figure 3 below and in updated Supporting Material, in which Supplemental Tables 4 and 8 and Supplemental Figures 4–6, 8B, and 13 have been corrected and replaced. Supplemental Table 10 also has been replaced to correct a typographical error. [Figure presented] The corrected analysis preserves the original observation that, relative to the M07 and M12 models, the M19 energy function improves the correlation of predicted PagP ΔΔG values with the experimental values (R = 0.830, 0.822, 0.403, and 0.919 for the M07, M12, R15M, and M19 models, respectively). The improvement is more pronounced than previously reported since noise arising from inadvertent design is now removed. The conclusion also holds for the contribution of polar side chains to the correlation coefficient (R = 0.778, 0.755, 0.484, and 0.920 for the M07, M12, R15M, and M19 models, respectively). The amino acids G, T, V, Y, and L remain the residues with the largest deviation from the experimental values. The other three benchmark tests were carried out at the originally reported weight of 0.5. As the ΔΔG test is independent from the fitting and benchmarking of the energy function, this correction does not affect any other results or the conclusions of the paper. We thank Johanna Tiemann, Kresten Lindorff-Larsen, and Amelie Stein of the University of Copenhagen for identifying the error and checking our correction. Rituparna Samanta of Johns Hopkins University debugged issues and recalculated the new data.