Towards novel treatments for SLE with Digital Twins of Cells
Systemic Lupus Erythematosus (SLE), often referred to as lupus, is a chronic autoimmune disease characterised by widespread inflammation and production of autoantibodies that attack the body's own tissues. Lupus is a complex and challenging condition that can cause a wide range of debilitating symptoms. Managing lupus is particularly challenging because there is no cure, and existing treatment focuses on symptom control and preventing disease flares. Leveraging the latest advances in generative AI, DeepLife has spearheaded the development of Digital Twins of Cells. These digital twins enable organisations to simulate a wide range of cell states, advancing disease and biological understanding, as well as pinpointing novel drug targets. In this case study, we discuss how DeepLife’s Digital Twin can be used in the pursuit of new drug targets for SLE.
It has been well established that B cells play a central role in SLE pathogenesis. Under normal conditions, B cells serve to recognise foreign antigens and produce antibodies in response, which target the foreign material for destruction. SLE is characterised by loss of self-tolerance, wherein B cells cannot distinguish between antigens that are foreign and those that are endogenous. This results in the production of autoantibodies, antibodies that react with self-antigens, which cause widespread inflammation and damage to the body’s own tissues and organs. As a result of their role in disease aetiology, B cells have been a key focus for development of SLE therapies. In broad terms, the goal of such therapeutic development efforts to date has been to reduce the amount of reactive B cells circulating in the blood, which could be done via direct depletion of B cells, inhibition of B cell survival factors, or preventing maturation, differentiation, or activation of B cells.
Although some effective therapies for SLE have been identified, the clinical heterogeneity observed in SLE means that not all patients respond well to existing treatments, particularly sufferers of the more severe form of the disease, refractory SLE. This motivates further research to identify novel drug targets in the pursuit of new therapies to overcome these present shortcomings. In order to do so, we begin with a fundamental question - what does a B cell look like in SLE patients? At DeepLife, we can ask such questions by leveraging our Digital Twin of Cells.
DeepLife's Digital Twin of Cells functions as a numerical representation of cells, able to simulate cellular processes to guide scientific understanding and decision-making. Built on DeepLife's OMICS Catalog and Cell Atlas, which provide an extensive repository of single-cell RNA sequencing data covering a plethora of cell types and tissues, this Digital Twin utilises state-of-the-art large language models (LLM) to learn and simulate a wide variety of cellular contexts. In the case of SLE, for example, we can design a prompt to simulate the transcriptional signature of a B cell in SLE. Since the model has also been trained on data from non-disease contexts, we can also simulate what a normal B cell would look like, simply by defining a normal cell condition in our prompt. In fact, the Digital Twin model, being a foundation model, can simulate a wide variety of cell contexts. This is thanks to training it on a vast corpus of data, covering many different diseases, organs, and cell types; which has been meticulously curated by the DeepLife team.
Utilising the digital-twin-generated cell signatures, we can construct a Gene Regulatory Network for SLE, or in more simple terms, group functionally relevant groups of genes together. Since we have the transcriptional signature of both normal B cells and SLE B cells, we can filter our network for genes in B cells that are altered in SLE. Using graph-based algorithms, we can then cluster this network of genes into communities. This enables us to identify key gene “modulators”, or in other words, key regulatory driver genes that may underlie the perturbed expression of differentially expressed genes we see in SLE B cells, which could play a role in SLE pathogenesis.
Interestingly, when we applied these techniques to identify novel drug targets in SLE, we identified both well-known targets as well as novel targets. For example, the top candidate gene target for SLE identified in our approach was TNFSF13B, which encodes BAFF, or B-cell activating factor. BAFF is essential for the development and survival of B cells, and is the target of Belimumab, which inhibits B cell survival and maturation in SLE. The second ranked candidate gene is STAT2, which also has been identified as an SLE target gene in several reports. STAT2 encodes a transcription factor that is activated by Interferon (IFN) signalling. Notably, aberrant IFN signalling is a feature of SLE pathology, and several studies have shown promising results by inhibiting IFN signalling in SLE. The fact that we have identified gene targets already implicated in SLE confirms that our approach is robust.
What is perhaps more interesting than the known targets identified in our analysis is the discovery of novel targets in SLE. In our analysis, we found 8 previously unreported targets associated with SLE, including WARS1. WARS1 is a gene involved in protein synthesis, specifically, it ligates tryptophan to tRNAs during protein synthesis. This is interesting because perturbed tryptophan catabolism has been commonly reported in SLE, further implicating that WARS1 may play a key role in the manifestation of the disease. Not only that, but interrogation of DeepLife’s Cell Blueprint reveals that WARS1 is downstream of IFN signalling, which as pointed out earlier, is heavily implicated in SLE. Taken together, these findings suggest that WARS1 may play a key role in SLE pathogenesis, and represents a key candidate drug target to consider in future drug discovery research.
In this work, we have not only reaffirmed known SLE targets but also unearthed novel ones not yet reported in the literature, such as WARS1. With DeepLife’s Digital Twin of Cells and Cell Blueprint, users now have access to cutting-edge tools for understanding the intricate complexities of disease biology, offering a new data-driven approach to therapy development. As our technology matures, the Digital Twin will continue to evolve, enhancing our grasp on cellular interactions and expediting therapy development. For more information and to explore potential partnerships, please contact us.