AlphaFold 3 Fails to Predict D-peptide Chirality, Fold, and Binding Pose in Heterochiral Complexes

Due to their favorable therapeutic properties, including improved stability, bioavailability, and membrane permeability, D-peptides that bind biological L-proteins represent an important class of systems in computational drug design. A reliable in silico workflow for these systems must correctly preserve stereochemistry while predicting fold and binding pose. The AlphaFold 3 (AF3) model reported by Abramson et al. (2024) enforces a strict chirality violation penalty to maintain chiral centers from model inputs and is reported to have a low chirality violation rate of only 4.4% on a PoseBusters benchmark containing diverse chiral molecules. Herein, we report the results of 3,255 black-box experiments with AF3 to evaluate its ability to predict the fold, chirality, and binding pose of D-peptides in heterochiral complexes. Despite inputs specifying explicit D-stereocenters, we report that the AF3 chirality violation rate for D-peptide binders is much higher at 51% across all evaluated predictions; on average the model is as accurate as chance (random chirality choice, L or D, for each peptide residue). Increasing the number of seeds failed to improve this violation rate. The AF3 predictions exhibit incorrect folds and binding poses, with D-peptides commonly oriented incorrectly in the L-protein binding interface. Confidence metrics returned by AF3 also fail to distinguish predictions with low chirality violation and correct docking vs. predictions with high chirality violation and incorrect docking. We conclude that AF3 is a poor predictor of D-peptide chirality, fold, and binding pose and propose solutions to address these limitations. Significance StatementAlphaFold 3 (AF3) is a model trained to predict protein interactions. This algorithm is tuned to respect chiral centers (L and D). Changing the chirality of even one protein residue can significantly alter chemical properties such as binding and stability. Therefore, an algorithm should exhibit a chiral center error rate of 0%. Although the original AF3 authors reported a 4.4% chirality violation rate, we have found that the rate for D-peptides is much higher at[~] 50%. Our data reveal a crucial structural prediction error in AF3 and demonstrate that this widely used model is as accurate on average as chance (random chirality choice, L or D, for each peptide residue). These results indicate structure prediction of D-peptides is an outstanding problem.