AbAdapt: An adaptive approach to predicting antibody-antigen complex structures from sequence

概要

[目的(Purpose)]
Antibodies are a highly diverse class of immune receptors whose binding residues (paratopes) are selected to specifically recognize a given antigen at a given surface patch (epitope). Antibody structures can be predicted from sequences and the binding modes of antibody-antigen complexes can be sampled by existing protein docking methods. However, the scoring of antibody-antigen docked poses starting from unbound homology models has not been systematically optimized for a large and diverse set of input sequences.

〔方法ならびに成績(Methods/Resulls)〕
To address this need, we have developed AbAdapt, a web server that accepts antibody and antigen sequences, models their 3D structures, predicts epitope and paratope, and then docks the modeled structures using two established docking engines (Piper and Hex). Each of the key steps has been optimized by developing and training new machine-learning models. The sequences from a diverse set of 622 antibody-antigen pairs with known structure were used as inputs for leave-one-out cross validation. The final set of cluster representatives included at least one “Adequate” pose for 550/622 (88.4%) of the queries. The median (IQR) ranks of these “Adequate” poses were 22 (5 to 77). Similar results were obtained on a holdout set of 100 unrelated antibody-antigen pairs.

[総括(Conclusion)]
We have attempted to build a working pipeline out of available tools, assess what worked and what did not, and suggest directions for future improvement. Our efforts were generally encouraging, as indicated by the query coverage and improvement in True ranks across a large and diverse test set, along with the general agreement between LOOCV and holdout benchmarks. Although questions remain about the best balance between Piper and Hex poses, both the only-Piper and Piper-Hex AbAdapt pipelines produced better results than could be obtained by Piper alone. Also, the improvement in epitope prediction performance upon addition of antibody-specific features suggests a way of addressing this long-standing and important problem.