The ELIXIR 3D-BioInfo Community in structural bioinformatics held its third Annual General Meeting in November 2021. Taking place over three afternoons, the virtual gathering attracted 500 registrants in total and as many as 300 participants for some sessions.
The 3D-BioInfo Community works on activities in five areas, each of which organised a session (Box 1). These were a mix of invited speakers, themed discussions and talks selected from poster abstracts. In addition, a dedicated poster session was held along with a spirited social get-together.
Box 1: The five areas of interest of the 3D-BioInfo Community
- To develop the infrastructure for FAIR structural and functional annotations
- To create open resources for sharing, integrating and benchmarking software tools for modelling the proteome in 3D
- To help develop models for protein-ligand interactions
- To develop tools to describe, analyse, annotate, and predict nucleic acid structures
- To establish a Biostudies database of protein engineering results
Common themes throughout the event were the advances in protein structure prediction by DeepMind and the release of AlphaFold2 code and nearly 390,000 predicted 3D-models for 20 model organisms. These data are available from a new web portal hosted by the European Bioinformatics Institute (EMBL-EBI). The first session on FAIR annotation discussed the importance of flagging that AlphaFold is a prediction database, and the associated opportunities for training and awareness activities.
The proteome modelling session featured talks on the application of the AlphaFold2-monomer inference engine to the prediction of protein-protein and protein-peptide complexes. Encouraging results were reported on prediction performance, notably when used in combination with performant classical docking algorithms. There were lively discussions on the challenges arising and how these may be met in the future.
The protein-ligand interaction session highlighted future areas of interest such as the emergence of new targets and target classes, new binding sites, addressing fake pockets from less reliable regions of 3D models and post-translational modifications.
Talks from the nucleic acid session covered a broad spectrum of approaches to the structural bioinformatics of nucleic acids including the application of deep learning, machine learning, and AI approaches. The need for tool integration and benchmarking standards was stressed.
Finally, the protein engineering session featured the many different approaches available to engineering and design, along with an exploration of the structural and functional consequences of point mutations. A challenge will be to identify commonalities and connections between the conclusions reached by the different approaches.
To find out more about the event and speakers you can view the event programme, and more detailed event summary, as well as the 3D-BioInfo Community page. Additionally, the 3D-BioInfo Community is running a webinar series covering many of the topics discussed in the meeting, check back on the page and sign up for the next session/s.