Constantin Schneider is an immunoinformatics research scientist at Exscientia working on ways of predicting antibody amino acid sequences, an important step in the process of antibody therapeutics design.

Constantin’s talk “Inverse folding for antibody sequence design using deep learning” describes how the information obtained from antibody backbone structures can be used to train machine learning models to predict their amino acid sequences from structure (inverse folding). The aim of the machine learning model is to be able to take structural information from an antibody of interest and design its amino acid sequence. This inverse folding approach can potentially improve biotherapeutic drug discovery by predicting sequences with good binding and designability characteristics.

Machine learning (ML) has the potential to transform many aspects of drug discovery and development by enabling research teams to extract hidden value from their data, understanding key relationships within chemical and biological systems that might otherwise remain undiscovered. It can build models that can be used to predict behaviour and to design experiments so that targets are achieved much more rapidly. But there are also a range of practical problems that constrain the use of ML with real pharmaceutical R&D datasets.

Data generated for drug discovery and development often has inherent characteristics that limit analysis. One example is that datasets may record many independent factors but relatively few actual experimental data points. Datasets with these sorts of dimensions pose many issues for ML - they can make it much harder for ML models to find and understand critical relationships within the data. Sparsity of data is another challenge. Not all factors and responses are measured for all experiments. ML models typically don’t like null or empty values and tend to discard the whole instance of that experiment if one value is missing. Finally, it can be difficult to distinguish ‘noise’ in the data - which values are actually significant and which are just background signal and biological variability.

Thomas’s talk “Harnessing AI for R&D in Drug Discovery” uses case studies to show how AI can get around these issues with source data and provide meaningful outputs

Lee's talk is titled "Skills for augmenting R&D with AI"

Prokarium is, a biotech company aiming to develop sustainable therapeutics at the convergence of synthetic biology and immunology.

The oncology field is remarkable innovations, but challenges such as unsustainability, cost, restricted efficacy, and side effects. The toxicity is, in part, caused as tumour cells can ‘cloak’ or ‘hide’ themselves from the immune system, making the tumour environment hard to target and penetrate.

Prokarium is introducing a novel approach to address these challenges: Living Cures. The platform provides a way to target delivery of therapeutic cargo, both in time and to a specific location. Such accurate targeting will both improve efficacy and reduce off-target side effects. Livija's talk is titled "Reengineering evolution into a synthetic biology platform for novel immunotherapies".

DJS Antibodies was founded in 2015 and was successfully acquired by AbbVie in 2022. 

David explains how his scientific curiosity caused him to take a different route from perhaps the more well-known academic or industrial career after his PhD. The spark was wanting to take an idea about how to produce antibodies against challenging targets and translate it into reality.

 His talk “Turning an idea into a successful company” charts the requirements of fundamental research, an understanding of the therapeutic needs, building a capable team and identifying commercial opportunities.

Daniel’s talk “Bicycles: An entirely new approach to targeting disease” describes bicyclic peptides and discusses their potential to be developed into effective therapeutic agents.

Bicyclic peptides (Bicycles) are short peptides, approximately 9-20 residues long, that have been chemically constrained by a central scaffold, improving binding affinity to the target protein and resistance to degradative pathways.

Bicycles combine the pharmacological properties normally associated with a biologic with the manufacturing and PK advantages of a small molecule. Bicycles are chemically synthesised, making them easier to scale for manufacture and be used as building blocks to form large complexes.