MOTIVATION: Antibody-antigen complex modelling is an important step in computational workflows for therapeutic antibody design. While experimentally determined structures of both antibody and the cognate antigen are often not available, recent advances in machine learning-driven protein modelling have enabled accurate prediction of both antibody and antigen structures. Here, we analyse the ability of protein-protein docking tools to use machine learning generated input structures for information-driven docking. RESULTS: In an information-driven scenario, we find that HADDOCK can generate accurate models of antibody-antigen complexes using an ensemble of antibody structures generated by machine learning tools and AlphaFold2 predicted antigen structures. Targeted docking using knowledge of the complementary determining regions on the antibody and some information about the targeted epitope allows the generation of high-quality models of the complex with reduced sampling, resulting in a computationally cheap protocol that outperforms the ZDOCK baseline. AVAILABILITY AND IMPLEMENTATION: The source code of HADDOCK3 is freely available at github.com/haddocking/haddock3. The code to generate and analyse the data is available at github.com/haddocking/ai-antibodies. The full runs, including docking models from all modules of a workflow have been deposited in our lab collection (data.sbgrid.org/labs/32/1139) at the SBGRID data repository.

SUMMARY: In this article, we introduce ABodyBuilder3, an improved and scalable antibody structure prediction model based on ABodyBuilder2. We achieve a new state-of-the-art accuracy in the modelling of CDR loops by leveraging language model embeddings, and show how predicted structures can be further improved through careful relaxation strategies. Finally, we incorporate a predicted Local Distance Difference Test into the model output to allow for a more accurate estimation of uncertainties. AVAILABILITY AND IMPLEMENTATION: The software package is available at https://github.com/Exscientia/ABodyBuilder3 with model weights and data at https://zenodo.org/records/11354577.

Hundreds of technical, special interest Internet weblogs are already generating thousands of niche articles worldwide, and many institutions are starting to create internal blogs for team collaboration. As this style of communication becomes more pervasive in the lives of employees and researchers, the difficulty of finding relevant information only grows with the number of authors and articles. To reduce the load, we propose using implicit group messaging (IGM) to automatically deliver relevant content to readers grouped by shared characteristics or interests. In this paper, we outline a context-aware application suited to special interest messaging and describe three alternative delivery models including our peer-to-peer (P2P) design called SPICE and a broker-based design. We investigate the advantages and disadvantages of each approach through detailed simulations driven by realistic data and actual national/global network topologies. We find that although a broker-based design is generally the most network efficient and lowest latency, a structured P2P system can offer exceptionally low and fair loading across peers and network links without relying on specialized broker nodes.

We propose a decoupled distribution paradigm for connecting large numbers of content publishers and consumers called implicit group messaging (IGM). Unlike traditional multicast or publish/subscribe messaging, IGM delivers content to "implicit groups" of consumers matching combinations of attributes specified by the publisher at the time of publication. The contributions of this paper include a characterisation of IGM encompassing the types of groups that must be supported, and an innovative distributed P2P implementation that we compare via simulation to a functionally equivalent client/server approach. Our results show the P2P model to efficiently support a range of implicit groups with acceptable delivery times and vastly less maximum peer and link stress than a client/server approach