Objective
The overarching goal of this project is to create an advanced artificial intelligence (AI) model that can replicate the challenging conditions beta cells face after transplantation. This model will allow us to identify new targets for interventions to protect these cells, ultimately aiming to improve the success of beta cell replacement therapies for Type 1 Diabetes (T1D) patients. By helping transplanted cells survive and function better, this project has the potential to make cell replacement therapy a more effective treatment option for T1D.
The project is organized into four key aims, each building upon the other to achieve a comprehensive understanding of how to support beta cell health in a transplant setting. First, the team will develop an AI-assisted tool to create a "signature" or blueprint of the cellular changes that occur during the early engraftment phase after transplantation (Aim1). This signature will help identify critical pathways and potential interventions that could be used to boost beta cell survival.
Next, in Aim2, we will use these insights to conduct a large-scale experimental screen, testing a wide range of interventions under lab conditions that mimic the environment the cells encounter after transplantation. This screening process will allow us to quickly identify the most promising interventions for protecting beta cells in the challenging early stages post-transplant.
In Aim3, the top-performing interventions from the screening process will be further tested by transplanting both primary islets and stem cell-derived islets into mice. This in vivo testing will enable us to observe the effects of these interventions on cell survival and functionality in a living system, closely simulating the conditions that would be faced in human patients.
Finally, in Aim4, data gathered from both the laboratory and in vivo studies will be fed back into the AI model, refining its accuracy and predictive power. This "lab-in-the-loop" approach ensures the model continuously improves and becomes more precise, allowing it to make even better predictions about which interventions will have the strongest positive impact on beta cell survival.
Through these combined efforts, and the complementary expertise from our two research teams, this project seeks to advance beta cell replacement therapy for T1D by developing a smarter, AI-driven framework for protecting transplanted cells. If successful, this research could not only improve treatment outcomes but also pave the way for the use of similar AI-based tools in other areas of regenerative medicine.
Background Rationale
Type 1 diabetes (T1D) results from an autoimmune loss of insulin producing beta cells. It is normally treated with insulin injections or insulin infusion with a pump. Alternatively, it is treated by transplanting a whole pancreas or the pancreatic islets of an organ donor, which reconstitute the intrinsic insulin secretion. This is referred to as cell replacement therapy. It is associated with better control of blood glucose and slower progression of diabetic complications than insulin injections. Although effective, cell replacement therapy is limited by low availability of organ donors, the need for immunosuppressive medication and substantial losses of the transplanted beta cells (60-80% loss). The first two limitations are addressed by the use of stem cell derived islets which can be generated limitlessly and they can potentially be engineered to evade the immune system. However, the third limitation concerns both stem cell- and organ donor islets. With substantial and variable losses of the therapeutic cells, the cost of treatment is high and the outcomes difficult to predict. In the case of stem cell derived islets, these losses can induce enrichment for unwanted cells that can arise as a byproduct of their generation in the lab, which is a safety concern.
Single-cell multi-omics (SCMO) is a cutting-edge technology that helps scientists understand the unique characteristics of individual cells by studying various aspects of their biology all at once. The vast amounts of data currently available, or being produced, can be of enormous benefit to T1D research, such as helping to improve beta cell transplantation success, yet need advanced artificial intelligence (AI) approaches to fully unlock their potential.
By combining SCMO data with generative AI for predictive modeling, underutilized in T1D research to date, our project can analyze complex datasets to identify subtle molecular patterns that would otherwise go unnoticed. AI models help predict which cells are most likely to survive and thrive after transplantation, guiding the design of targeted interventions to protect vulnerable beta cells. These predictions will be used to plan a series of laboratory experiments, where we will test different approaches to enhance cell resilience against post-transplant stressors like inflammation and oxygen deprivation.
This iterative "lab-in-the-loop" process will be central to our research, meaning that laboratory and computational approaches work hand in hand, learning from and improving upon each other. Data generated from laboratory screens will be fed back into the AI models, refining their predictive accuracy and enabling them to account for new variables observed in real-world conditions. With each cycle, the models will improve, ultimately enhancing the reliability of predictions for transplant success. This approach not only promises to improve the success rate of islet cell transplants but also helps pave the way for more personalized and cost-effective treatments for T1D. Our project thus represents a major step forward in harnessing cutting-edge technology, integrated with hands-on experimentation, to ultimately create more reliable and accessible therapies for patients living with T1D.
Description of Project
Stem cell islet therapies are becoming a promising alternative for treating patients with Type 1 Diabetes (T1D), a disease with currently no available cure. Significant advancements for effective transplant therapies have been made in recent years, especially in creating functional islet cells from stem cells. However, concerns still persist: like primary islets, SC-islets have been reported to suffer during transplantation, with preferential loss of the on-target beta cells, which could enrich undesired off-target cells. Therefore, one of the most important remaining obstacles in making diabetes cell replacement therapy a clinically routine procedure are the ischemic beta cell losses upon transplantation.
Artificial Intelligence (AI) and Machine Learning (ML) approaches have the potential to revolutionize T1D research, but have been underutilized to date. For example, AI/ML can enable researchers to utilize huge and complex biological and clinical data, such as single-cell omics datasets (containing individual cells' genetic information), in unparalleled scale and functionality, to predict and model factors potentially affecting SC-islet transplantation success, as well as identify potential interventions which could improve long-term transplantation viability.
Targeting this research gap, two research groups from Helmholtz Munich and Technical University of Munich are collaborating on this project: Fabian Theis, a specialist in AI and single-cell genomics, and Heiko Lickert, an expert in beta cell and stem cell biology, along with their respective teams, will lead the proposed efforts. Their complementary, interdisciplinary expertise at the intersection of AI and pancreas biology is crucial for the success of the project. Now is the ideal time for this research because recent advances in AI, stem cell technology, and single-cell genomics have converged, providing unprecedented tools and data to tackle the challenges of T1D treatment with a level of precision and effectiveness previously unattainable.
In this project, we plan to leverage advanced AI/ML techniques alongside essential wet-lab screening experiments, pioneered by our two research groups, to enhance transplanted cell viability in early-stage islet transplantation. By analyzing large-scale datasets, we aim to identify stressors—such as inflammation and lack of oxygen—that compromise islet cell health and design strategies to mitigate beta cell loss. Specifically, we plan to use data-driven AI analysis to identify key markers of ischemic beta cell death, guiding the identification and pre-clinical testing of pharmaceutical or siRNA interventions that strengthen cell resilience during transplantation. These insights will inform preconditioning protocols to improve cell survival and uncover actionable pathways for addressing transplantation challenges.
Through sophisticated AI approaches developed and employed throughout this project, including in-silico perturbation modeling (i.e. testing how cells respond to different treatments or changes virtually) and lab-in-the-loop workflows (i.e. keeping AI and lab-based elements in constant contact), we will optimize beta cell survival and function across varied transplantation conditions. This strategy will enable precise interventions, maximize limited donor material, reduce outcome variability, and enhance safety by minimizing off-target cell proliferation. Ultimately, our approach will provide critical insights to improve transplantation effectiveness and accelerate the clinical adoption of stem cell therapies for T1D, moving closer to a sustainable, long-term solution and potential cure for the disease.
Anticipated Outcome
In this project, we will first gather available data on the consequences of transplantation at the cellular level. As these consequences are manifold, involving several signaling pathways such as regulation of cell death and identity and responses to nutrients and oxygen availability. We will develop an Artificial Intelligence (AI)-assisted tool to identify a cellular signature of the engraftment process that follows transplantation. With this signature, we will identify critical pathways and interventions to modify them.
Then, we will conduct a large-scale experimental screen in the dish, in conditions that simulate those encountered during engraftment. With this, we will narrow down the interventions that come in the form of approved medicines and gene expression modifying agents, to a handful for subsequent testing. We will then test whether these interventions actually improve the survival of stem cell derived islets in vivo by transplanting the therapeutic cells in mice, and if so, whether the use of these interventions improves the cure rate of experimentally induced diabetes in mice using both stem cell and donor islets.
Finally, we will feed back the information gathered during our engraftment simulating conditions in the dish and during the testing in live animals to our AI-assisted tool in order to refine our experimental approaches and to provide a tool for Type 1 diabetes researchers and the wider community.
The work proposed here will enable more precise intervention during cell therapies, reduce therapy costs by enabling the limited donor material to be used by more patients, reduce variability in outcomes, and improve safety by avoiding off-target cell types using an enrichment strategy.
Relevance to T1D
Type 1 Diabetes (T1D) is an autoimmune condition in which the body mistakenly attacks and destroys insulin-producing beta cells in the pancreas. For patients with T1D, islet cell replacement therapy offers a promising path toward restoring insulin production, potentially freeing them from constant blood sugar monitoring and insulin injections. This therapy involves transplanting functional islet cells, either from human donors or developed from stem cells, to replace the lost beta cells. However, current methods face challenges, particularly the loss of transplanted cells in the initial stages after transplantation.
Our project aims to overcome these challenges by using advanced Artificial Intelligence (AI) and Machine Learning (ML) techniques to reduce early cell loss, making cell replacement therapy a more effective and accessible treatment for T1D. The collaborative effort planned in this project draws on the complementary expertise of two research teams: one specializing in AI and predictive modeling, and the other in pancreas biology and stem cell-derived beta cells, combining strengths to tackle T1D from both a technological and biological perspective.
A major goal of this project is to significantly reduce the number of therapeutic cells lost shortly after transplantation. This improvement could allow patients to receive islet cells from a single donor, as opposed to multiple donors, thus alleviating the scarcity of donor material and making the process more efficient. In the case of stem cell-derived islet cells, reducing cell loss would lower the costs associated with producing these cells, ultimately making the treatment more affordable. Additionally, better survival rates for transplanted beta cells would reduce the likelihood of graft failure, helping to ensure the therapy’s effectiveness in the long term.
One key innovation of our project is the development of a highly sophisticated AI tool to predict and mitigate the factors leading to beta cell loss after transplantation. By integrating and analyzing a wealth of biological data, this tool will allow us, and other researchers, to identify the stressors—such as inflammation or lack of oxygen—that newly transplanted cells encounter, and will help in developing targeted strategies to improve their resilience. Our AI tool will be iteratively refined using both predictive modeling and experimental data gathered throughout the project, resulting in a powerful resource for the T1D research community. This tool could also have applications in other research areas, such as beta cell regeneration, immune system modulation, and beyond, making it a valuable resource for the broader scientific community.
AI is still a relatively new addition to the field of diabetes research, but it offers immense potential. By employing ML and deep learning methods, this project opens up possibilities for handling complex datasets that were previously difficult to manage. AI can streamline the development and optimization of stem cell therapies, improving transplantation success rates and advancing the broader field of T1D treatment.
By combining single-cell omics data and AI, our project will accelerate the development of new therapies, improving their safety and effectiveness. These advancements have the potential to benefit not only T1D patients but also individuals affected by other diseases requiring cell replacement or regenerative approaches. In this way, our project marks a significant step forward in applying cutting-edge technology to healthcare, aiming to improve quality of life and treatment outcomes for those with T1D.