Blind prediction performance of RosettaAntibody 3.0: Grafting, relaxation, kinematic loop modeling, and full CDR optimization

Abstract

Antibody Modeling Assessment II (AMA-II) provided an opportunity to benchmark RosettaAntibody on a set of eleven unpublished antibody structures. RosettaAntibody produced accurate, physically realistic models, with all framework regions and 42 of the 55 non-H3 CDR loops predicted to under an Ångström. The performance is notable when modeling H3 on a homology framework, where RosettaAntibody produced the best model among all participants for four of the eleven targets, two of which were predicted with sub-Ångström accuracy. To improve RosettaAntibody, we pursued the causes of model errors. The most common limitation was template unavailability, underscoring the need for more antibody structures and/or better de novo loop methods. In some cases, better templates could have been found by considering residues outside of the CDRs. De novo CDR H3 modeling remains challenging at long loop lengths, but constraining the C-terminal end of H3 to a kinked conformation allows near-native conformations to be sampled more frequently. We also found that incorrect VL–VH orientations caused models with low H3 RMSDs to score poorly, suggesting that correct VL–VH orientations will improve discrimination between near-native and incorrect conformations. These observations will guide the future development of RosettaAntibody.

abYsis: Integrated Antibody Sequence and Structure—Management, Analysis, and Prediction

“…Total missing from IMGT that are present in EMBLIG 35552 Rosetta Antibody [35,36], Kotai Antibody Builder [37] (kotaiab.org)…”

Improved prediction of antibody VL–VHorientation

“…Because the CDRs are attached to the framework of the V L and V H domains, any change in the relative orientation of the V L and V H domains will propagate to change the CDRs' relative orientation, and therefore, the shape of the paratope. Failing to account for the V L -V H orientation during CDR or paratope structure prediction dramatically hinders the quality of the output models, and recent evaluation found the V L -V H orientation to be a limiting factor in antibody structure prediction (Weitzner et al, 2014). Abhinandan and Martin (2010) were the first to codify a metric for measuring the V L -V H orientation.…”

“…RosettaAntibody is an application for blind prediction of antibody structure (Sivasubramanian et al, 2009;Weitzner et al, 2014). RosettaAntibody operates in two phases: (i) template selection and grafting, wherein known antibody structure fragments are combined to create a coarse-grained model, and (ii) structure refinement, which uses Monte Carlo perturbations with minimization to remodel the CDR H3 loop, refine all CDR loops, and redock the V L and V H domains.…”

“…During the Second Antibody Modeling Assessment (AMA-II), RosettaAntibody's orientation predictions were evaluated explicitly, comparing the packing angles of the Rosetta models to those of their corresponding crystal structures (Weitzner et al, 2014). RosettaAntibody compared favorably in most respects to the competing protocols, producing two sub-Ångstrom H3 models and achieving the best H3 model in four targets.…”

MoFvAb: Modeling the Fv region of antibodies

“…35 The different methodologies differ mainly in the approach of the initial template selection, the extent of de novo modeling in CDR-H3, and the preference for either knowledge-based or forcefield-based model refinement techniques. [36][37][38][39][40][41] The latter also applies to the problem of VH-VL orientation, which some try to tackle by protein-protein docking-like approaches 36 and sophisticated energybased refinement, 37 while others are relying more on identifying a suitable VH-VL orientation template structure. [38][39][40][41] Here, we present a novel antibody homology methodology that incorporates experiences and ideas that have arisen from applied antibody engineering in an industry environment.…”