Title:	Denoising diffusion probabilistic models for computationally-optimised immunotherapy
DNr:	Berzelius-2025-142
Project Type:	LiU Berzelius
Principal Investigator:	Luca Panconi <luca.panconi@scilifelab.se>
Affiliation:	Stockholms universitet
Duration:	2025-04-29 – 2025-11-01
Classification:	10610
Keywords:

Abstract

During a healthy immune response, antigens bind to T cell receptors (TCR) on adjacent lymphocytes and trigger TCR oligomerisation. It is well-documented that clustering of TCR amplifies the signal from antigen-presenting cells and serves as a fundamental precursor of the immune response. The success of immunotherapy depends largely on expression of neoantigens, such as programmed death-ligand 1 (PD-L1), and nanoscale organisation of T cell receptors (TCR), which aggregate to promote signal digitisation and activation. Despite exhibiting the hallmarks of an immunogenic cancer, immunotherapy often fails to induce TCR clustering in triple negative breast cancer (TNBC). However, therapeutic mediators can directly elicit aggregation of T cell receptors, inducing immunological activation in the presence of cancer cells, to target and destroy tumours. Such therapeutic mediators could target many cancer antigens, but this process would differ on a cell-to-cell and antigen-to-antigen basis. Immunotherapeutic treatment strategies depend heavily on nanoscale receptor organisation in pre- and post-activated T cells and neoantigen prevalence in cancerous targets. In vitro testing is expensive, time-consuming, and limited to mediator availability – the only viable solution is in silico therapeutic design. Computational optimisation for immunotherapy has been hindered by a lack of receptor visualisation, but with the advent of super-resolution, we now have unique insights into TCR protein maps at nanoscale resolution. Protein aggregation dynamics (PAD) simulators, founded on classical agent-based modelling (ABM) approaches, show promise in simulating transmembrane receptor motility. However, they are limited by slow runtimes, under-fitting, and an inability to generalise to unseen distributions. By constructing simulations, learned via diffusive denoising from protein localisation data, it is possible to track receptor dynamics, determine requirements for clustering, and predict signal digitisation. As such, the goal of this project is to develop in silico methods, built upon diffusion-based neural networks, of quantifying immunotherapeutic efficiency and optimising treatment parameters for eliciting T cell activation. Here, we will take TNBC as a specific case study, with intent to generalise towards other hormone-negative cancers. Fundamentally, this framework will be used to optimise nanotherapeutics for immunotherapy in treating TNBC and other immunogenic cancers.