From the 11th until the 16th of May, we (Gemma and Henriette) attended PEGS Boston 2025. First of all, we will share some tips for preparing to attend a conference. We will give some general feedback on the conference and share some general highlights. Lastly, we will mention some of the talks we found interesting.  

Preparing to go to a conference

• Budget your trip. Conference fees, travel, accommodation, etc can get expensive quickly. Make sure you budget the trip and check with your funding sources how much you are allowed to spend on the individual categories (and where you are allowed to book them).  
• Submit your work and sign up. This is obvious but make sure you are signed up for the conference. If you do this early on you might benefit from early registration discount codes. Bring your own work to the conference to share your research and ideas with a broader audience. Make sure your supervisor approves the submission.  
• Print your poster. For (almost) all conferences you are supposed to bring your own poster. Printing on fabric might be easier for transporting the poster to the conference location. If you are in Oxford have a look at: https://www.physics.ox.ac.uk/about-us/our-facilities-and-services/media-services  
• Get your visa. Check if you need to get a visa/ESTA to go to the country where conference will be held. Be aware that for some you need to provide contact details of someone in that country and information on where you are staying. 
• Get travel insurance. Check if you need travel insurance and what is in line with the University policies. If you are in Oxford have a look here: https://finance.admin.ox.ac.uk/travel-insurance. 
• Plan your conference. PEGS and many other conferences have multiple tracks running at the same time. Use the conference app to make your personal calendar and plan which talks/posters you want to attend. Although you can easily swap between tracks during the session, try to minimise this such that you can focus on the talk instead of where to go next.  

General feedback

• Antibody focus. Although PEGS is a ‘Protein Engineering Summit’, talks pretty much exclusively focus on antibodies and immune proteins for applications in therapeutics/drug discovery.  
• Strong attendance by companies. PEGS is strongly attended by various lab and computational-based companies who are there to promote their work/service. As a PhD student, visting their booths in the exhibition hall can be a good opportunity to find out more about working in industry, or at least find yourself a fun new pair of socks. 
• Good to gauge areas of interest/challenges in the field. The strong attendance by companies also means that there talks tend not to go into a lot of detail on methods. However, PEGS is very useful to get a general interest and challenges in the field (see next section). For PhD students attending PEGS, it is therefore probably most valuable if you are far enough into the journey to know what you are doing, but also have some time left to work on new ideas gained from the conference.  
• Valuable poster sessions. The poster sessions at PEGS are organised into smaller clusters outside the exhibition hall and presentation rooms, placed where the most relevant people would find us. We found that this meant we had lots of visitors to our posters with questions and useful feedback that we could work on later.  
• OPIG tools are highly used. It is always good to hear that people use and are happy with the open-source tools provided by your research group. The Observed Antibody Space (OAS) and ABodyBuilder2 were the most often mentioned OPIG database/tool. 
• Well organised conference. PEGS Boston is a very well organised conference. The various tracks cover many aspects of protein and antibody engineering allowing us to learn more about the wider field while all within the same niche. The registration, the calendar, and the map of the venue was clearly communicated. The venue is very nice with comfortable chairs that allow you to sit all day without pain, and enough food and drinks are provided throughout the day. The conference is attended by various interesting and inspiring researchers.  

General highlights

• Antibody formats. Multispecific antibodies are of high interest and people get very creative with linking single or paired antibody chains with various other proteins. Antibody-drug conjugates were also often discussed. 
• Affinity / avidity. Experimentally improving and computational predicting antibody affinity/avidity is still challenging. 
• Developability. Again, developability was discussed from both the experimental and computational perspective. The best practices for experimentally measuring developability, as well as how to improve it, were discussed in various tracks. Towards the end of the conference, interest shifted to computationally predicting developability. Various tools were mentioned for predicting and optimising individual properties or paired/multiple properties in one go.  
• Antibody – antigen modelling. The current state of the art of accurately modelling of the antibody – antigen complex was discussed as well as the best ways to generate training and evaluation sets for this task. 
• Data generation and storage. As computational prediction methods are often limited by the amount and biased datasets available, designing experimental datasets for machine learning was a big topic. Various presenters mentioned how in their companies they tried to maximise this interplay. 
• Benchmark sets. There was a strong emphasis on the importance of benchmark sets and the need to compare against a simple baseline. 

Challenges/competitions

To get back to that latest point about benchmark set, benchmarking competitions/challenges were presented aiming to compare models on an independent test set.  

Dr. Andrew Bradbury and M. Frank Erasmus from Specifica presented The AIntibody Challenges, a set of antibody affinity challenges aimed to understand the potential of AI in real world antibody discovery. The challenges included in silico antibody affinity maturation, in silico affinity rank prediction, and NGS inspired computational design. Initial results indicated that computational methods rarely outperform experimental methods, but that better designs were generated by computational tools in a much shorter time frame. The challenge also indicated that many computational designed antibodies failed various developability assays.  

Another interesting challenge was presented by Justin Barton from Xaira Therapeutics. IMMREP is a TCR-Epitope binding challenge which aims to provide a standardised benchmark. Barton presented the main takeaways from the various years in which they run the challenge (IMMREP22, IMMREP12, and IMMREP23). Increasing difficulty of the competition highlighted the challenge of out of distribution prediction, both to novel TCRs and peptides.  

Sharks

A highlight of the conference was Prof. Helen Dooley’s (University of Maryland) talk on using VNARs, shark nanobodies, to develop novel multispecific antibodies. Her group focuses primarily on comparative immunology, looking to use the evolutionary distance between human and shark immune systems for therapeutic applications. Two pieces of work were discussed, both involving the engineering of multispecifics with VNAR domains: the first was multispecific antibodies to downregulate PD-L1, a common target for cancer therapies, and the second to target SARS-CoV-2 and sarbecoviruses. These studies both demonstrate the huge potential of VNARs in therapeutics, but naturally a large portion of the discussion was whether the greater genetic distance for shark-derived immune proteins may prompt a larger immunogenic response.