Last week marked a major milestone in small-molecule drug discovery with the first FDA approval of a proteolysis targeting chimera (PROTAC). After a modest but successful phase 3 clinical trial demonstrated a 2.9 month improvement in median progression free survival1 for a type of advanced breast cancer1, the FDA has approved Veppanu (vepdegestrant), co-developed by Arvinas and Pfizer, as the first PROTAC protein degrader therapy2. So what is a PROTAC?
Continue readingCuriosity might not kill the cat
Unlike most members of OPIG, I don’t work on small molecules, antibodies, or protein structure; I use hypergraph representations of protein complexes to predict gene essentiality and drug targets. I have also had an unconventional route to get here, and on the way, discovered my love for learning and research.
Friends and family had noticed I jumped around with my interests, so much so that when we used to meet up, they took great delight in teasing me about what my current adventure was – ‘you don’t settle do you!’, ‘when are you going to find what you’re looking for?’, ‘why can’t you just stick to something’. Looking back, there was a pattern, I just couldn’t see it yet.
Read more: Curiosity might not kill the catOn paper, I looked like a proper nut job. Started off as a data analyst, moved to France and had two kids, returned to the UK and volunteered as a National Trust gardener, set up my own garden design and build business, started a chicken farm with 2000 birds, café and shop, joined my accountant to grow his business, submitting tax returns and running his accountancy practice…
You can perhaps see what they were getting at, and I was beginning to question my own motivation and sanity! At this point, my kids left home and suddenly I was free, as all the choices I’d made had always been with my husband and kids’ needs first. But was I supposed to stay as a practice manager and perhaps learn how to become an accountant? What was I looking for? Why couldn’t I find something and stick to it?
After a chat with my human resources friend, she took me through a coaching exercise and suddenly I knew – I wanted to try and help discover something! Very cliché but true. But how do I do that? She pointed out that although it was great fun teasing me about my apparent lack of resilience, it could be that I was a learner, someone who is driven by curiosity. So, even though I already had a BSc in Electronic and Computer engineering, I started back at the beginning with an Open University Maths degree. Two years later, I did a PGCE (secondary school teaching qualification) and taught teenagers how to learn maths… maybe I’m supposed to help kids discover something?
Then I noticed AI, prompted by discussions with my brother and the AlphaGo documentary, and I was hooked. I devoured everything I could, applied to Aberystwyth University to complete an AI masters and got in! Couldn’t believe it, me in my 50s going back to uni – my friends had a field day! And then it dawned on me, what happens when I finish? How am I going to get a job at my age?
I knew I needed to start making connections so rather than doing my dissertation at the uni, I applied to institutes and companies to take me on for the 3 months and complete a project for them. And I got in! One of the most positive, lovely people I’ve ever met works at the Rosalind Franklin Institute and he said yes and showed me what life could be like if you become a researcher – learning as a job! AND, I might help discover something! AND, a PhD position became available to start that year as someone had dropped out! So I applied. I didn’t get in.
As I was making the application, I explored different types of PhD routes and applied to Oxford. And I got in! So here I am at OPIG, learning every day with other people who love to learn too. Society teaches us to think learning is step one before you can ‘do your job’; it took me 50 years to realise that learning can be the end goal. I’m at the age where most of my peers are thinking about retiring, but I don’t want to take it easy and go travelling. My foot is firmly pressed to the floor, it’s just my vehicle is a bit slower, perhaps a golf cart. But I’m fine with that; when you’re not racing, you might notice things others miss.
Speeding up python through profiling
Python is a shockingly slow language. A test on a raspberry pi of simply “turn this pin on and off as fast as you can” gave the results below.
| System | Library | Speed |
|---|---|---|
| Shell | /proc/mem access | 2.8 kHz |
| Shell / gpio utility | WiringPi gpio utility | 40 Hz |
| Python | RPI.GPIO | 70 kHz |
| Python | wiringPi2 bindings | 28 kHz |
| Ruby | wiringPi bindings | 21 kHz |
| C | Native library | 22 MHz |
| C | BCM 2835 | 5.4 MHz |
| C | wiringPi | 4.1 – 4.6 MHz |
| Perl | BCM 2835 | 48 kHz |
Pitfalls of AI-Generated Reviews: Case Study of a Frontiers in Microbiology Review on Anti-Influenza A bnAbs
In the last five or so years, large language models (LLMs) have transformed from a novel regurgitator of haphazardly stitched together sentences to an almost ‘human’ personality standing by our side as we tackle life. Whilst the perceived humanity of these models is the topic for perhaps a future blogpost, it is almost undeniable to understate the impact of LLMs in our daily lives. Do you need someone to proofread your essay you’ve spent hours drafting? GPT (or one of its many counterparts) has you covered. Need help drafting an email from scratch? No problem. Want to write and/or heavily edit an entire academic article which would typically require days, if not weeks, of research? Surely just needs a push of a button… right?
Despite tremendous advances in LLMs, key issues mean they are not yet a fully dependable addition to our writing endeavours. They are known to fail when asked to generate new content with only a basic prompt. Some of these failures have made headlines 1. Some of the scariest instances are those of hallucinated information 2–4 . This refers to the phenomenon where AI tools generate convincing information which is factually inaccurate or simply fabricated 2 . In Belgium, the Ghent university rector came under fire for citing quotes, supposedly from influential thinkers, which were later found to be AI-hallucinations 1.
Whilst there are numerous examples of the poorly cited and often AI-hallucinated papers falling through the cracks of the peer-review process, today we focus on a Frontiers in Microbiology reviewtitled ‘Broadly neutralizing monoclonal antibodies against influenza A viruses: current insights and future directions’ 5. This paper attempts to provide an overview of the current landscape of monoclonal antibodies (mAbs) which are being developed to confer protection against influenza A, highlighting ‘technological advances, clinical performance, and scalability’. However, it contains many of the hallmarks of text that has been created or edited with generative AI, despite the generative AI statement stating ‘The author(s) declared that Generative AI was not used in the creation of this manuscript.’
Continue readingAnalyzing AlphaFold 3’s Diffusion Trajectory
A useful way to understand AlphaFold 3’s sampling behavior is to look not only at the final predicted structure, but at what happens along the reverse diffusion trajectory itself. If we track quantities such as the physical energy of samples, noise scale, and update magnitude over time, a very clear pattern emerges: structures remain physically imperfect for most of sampling, and only take proper global shape in the final low-noise steps.
This behavior is a result of the diffusion procedure implemented in Algorithm 18, Sample Diffusion, which follows an EDM-style sampler with churn. Rather than simply marching monotonically from noise to structure, the sampler repeatedly perturbs the current coordinates, denoises them, and then takes a Euler-like update step. Because of the churn mechanism, AlphaFold 3 deliberately injects additional noise during part of the trajectory, which encourages exploration but also delays local geometric convergence. This mechanism is shown in step 4 -7 of the Sample Diffusion Algorithm from Alphafold3 Supplementary Information.
No Pretraining, No Equivariant Architecture – Learning MLIPs without Explicit Equivariance
Machine-learned interatomic potentials (MLIPs) have become a cornerstone of modern computational chemistry, enabling simulations that approach quantum accuracy at a fraction of the cost of traditional methods such as density functional theory (DFT). However, a central challenge in designing MLIPs lies in respecting the fundamental symmetries of molecular systems, especially rotational and translational invariance, while maintaining scalability and flexibility.
In our recent work, we introduced TransIP, a novel framework that formulates how symmetry is incorporated into molecular models by learning symmetry directly in the latent space of an atomic transformer model, in which we treat atoms as tokens, instead of hard-coding equivariance into the neural network architecture.
At the core of TransIP is a simple yet powerful idea: instead of enforcing SO(3) equivariance through specialized layers, the model is trained with a contrastive objective that aligns representations of rotated molecular configurations. A learned transformation network maps latent embeddings under rotations, encouraging the model to discover symmetry-consistent representations implicitly. This design preserves the flexibility and scalability of standard Transformers while still capturing the geometric structure of molecular systems.
Continue readingBiologic Summit 2026
This year we (Fabian and Henriette) were invited to speak at the Biologic Summit. The conference took place January 20-22 in San Diego. Fabian presented his work on conformational ensembles of antibodies [1] in the “Data Strategies and the Future of AI Models” track. Henriette presented her work on LICHEN [2], a tool to generate an antibody light sequence for a specific heavy sequence, in the “ML/AI for Biologics Developability, Optimization and de novo Design” track. Below we give some general highlights of the conference, and some talks we enjoyed. We would like to thank the organisers for the opportunity to discuss our research and hear about the latest developments in harnessing ML for design and optimisation of antibodies.
General feedback
- Industry focused conference. Biologic Summit is strongly attended by industry, making this conference an excellent opportunity to promote your tools/databases, and to connect with companies. The conference is attended by both start-ups as big pharma companies.
- Medium size conference. With approximately 250 attendees, the Biologic Summit provides a good opportunity to connect with researchers from a wide range of disciplines. Held concurrently with Protein Science and Production Week (PepTalk) and sharing the same venue, the event further benefited from a diverse mix of scientific backgrounds and expertise.
- Panel and table discussions. Throughout the three days there where various table discussions and panel discussions organised. These are good places to learn about general interest and challenges in the field.
- Well-organised conference. The conference is well-organised with a clear schedule and enough breaks to recharge and connect. Most talks are scheduled for 30 minutes with around 4 talks per block.
Not-Proteins in Parliament: Part II
Following on in Clare’s footsteps, I similarly took a three-month break from proteins and small molecules to work in the Parliamentary Office of Science and Technology (POST). While at POST, I researched the impact of AI on employment in the UK, covering AI adoption, job creation and loss, and effects on working conditions. You can read our literature review here. Many thanks to Lydia, who was an amazing supervisor while I was at POST.
SigmaDock: untwisting molecular docking with fragment-based SE(3) diffusion
Molecular docking sits at the heart of structure-based drug discovery. If we can reliably predict how a small molecule binds in a protein pocket, we can prioritize compounds faster, reason about interactions more clearly, and build better pipelines for hit discovery and lead optimization. But in practice, docking is still a difficult problem: classical methods are often robust but imperfect, while recent deep learning approaches have sometimes looked promising on headline metrics without consistently producing chemically plausible poses.
SigmaDock was built to address exactly that gap. Instead of treating docking as a problem of directly diffusing on torsion angles or unconstrained atomic coordinates, SigmaDock represents ligands as collections of rigid fragments and learns how to reassemble them inside the binding pocket using diffusion on . In plain English: rather than trying to “wiggle” every flexible degree of freedom in a tangled way, SigmaDock breaks the ligand into chemically meaningful rigid pieces and learns where those pieces should go, and how they should reorient, to recover a valid bound pose.
Fragment-to-Lead Successes in 2024 – 10th Anniversary Edition
In what I have to admit is now becoming an annual tradition ([2023] [2019]), I’d like to highlight the 2024 edition of the fragment-to-lead success stories, published in J. Med. Chem. at the end of 2025 [Paper].
Continue reading