Dihedral angles that could neither be assigned cis nor trans are labeled as non-planar. highlight the enormous improvements in the field, but we also aim to YK 4-279 increase consciousness that protein structure models, and in particular antibody models, may suffer from structural inaccuracies, namely incorrect cis-amide bonds, wrong stereochemistry or clashes. We show that these inaccuracies impact biophysical house predictions such as surface hydrophobicity. Therefore, we stress the importance of cautiously reviewing protein structure models before investing further computing power and setting up experiments. To facilitate the assessment of model quality, we provide a tool TopModel to validate structure models. KEYWORDS:Antibodies, antibody structure, protein structure prediction, biophysical surface properties, structural inaccuracies == Breakthroughs in protein/antibody structure prediction == Predicting the three-dimensional (3D) structure of a protein based solely within the amino-acid sequence is one of the grand difficulties in the field of protein structure prediction.1Accurate prediction of the 3D structure of a protein is critical to understand its function, as the shape of the protein determines its properties and ultimately its function. To determine the state-of-the-art methods in protein structure prediction, the biennial community-based benchmarking experiment Critical Assessment of methods in protein Structure Prediction (CASP) was founded.24In CASP14 (2020), DeepMind showcased AlphaFold2, a program based on artificial intelligence (AI) that directly processes multiple sequence alignments.5Comparable accuracies in predicting protein structures can also be achieved with additional methods including RoseTTAFold,6and specialized tools for antibodies which include the recent advances.79Those tools are highly accurate based on global measures, often with root mean square deviations (RMSDs) to the crystal structure of less than 1 . However, there are often higher inaccuracies in specific parts of the protein that should be cautiously examined.10,11Post-translational modifications are omitted, but can sometimes be added afterwards.12Furthermore, the accuracy for multimers, such as antibodies, is still lower.13Additional challenges can arise for antibodies since VDJ recombination events do not follow the classical pathway of evolution.14 Antibodies are crucial components of the adaptive immune response.15Genetic recombination and somatic hypermutation events enable the adaptive immune system to produce a vast number of antibodies against a variety of pathogens.14To understand and optimize antigen recognition and to enable rational design of antibodies, accurate YK 4-279 structure models are essential.16Despite these recent advances, accurate structure prediction of antibodies remains challenging and still needs to be extensively validated. In particular, the flexible loops involved in realizing the antigen present a major challenge.17,18In comparison to additional protein superfamilies, the fold of antibodies is generally highly conserved.1921In particular, the framework YK 4-279 of the antigen-binding fragment (Fab) is structurally almost identical for those antibodies.22,23However, the area hardest to predict accurately is the six hypervariable Rabbit Polyclonal to CPZ loops that can form, together with several platform residues, the antigen-binding site, interesting with the respective epitope. These loops are also known as the complementarity-determining region (CDR) and provide the sequence and structure diversity essential to recognize a wide range of antigens. Five of the six loops tend to adopt canonical cluster folds based on their size and sequence composition. However, the third CDR loop of the weighty chain, the CDR-H3 loop, is the most varied in length, sequence and structure and therefore is the most demanding loop to forecast accurately. Various methodologies have been developed to improve and advance antibody structure prediction, by incorporating different homology model strategies or using the canonical cluster model,2428but recently numerous antibody-specific deep learning methods such as ABlooper, DeepAb and IgFold have significantly improved the CDR loop modeling accuracy.79,29The predicted structure models achieve related or better quality than methods that are able to predict all types of protein structures (including AlphaFold2).13,30Additionally, the CDR-H3 loop conformation is also strongly influenced from the relative interdomain orientation, as it is located in the center, directly in the interface between the heavy and the light chains. It has been demonstrated that, in addition to the CDR loops, the relative interdomain orientation takes on a crucial part in defining the.