Periodic Reporting for period 3 - KATY (Knowledge At the Tip of Your fingers: Clinical Knowledge for Humanity)
Reporting period: 2024-01-01 to 2025-06-30
AI-empowered Personalized Medicine promises to find tailored cures for patients. Such therapies are starting to be adopted but AI-empowered Personalized Medicine promises a transformation for the cancer community. However, no matter how precise it is and no matter how many lives it can save, if clinicians do not understand its suggestions and decisions, AI-empowered Personalized Medicine will not be a game changer impacting everyday clinical decisions. Hence, the real challenge is building AI-empowered Personalized Medicine systems that can be accepted by clinicians and clinical researchers. Knowledge at the tips of your fingers (KATY) grasped the above challenge and proposed an AI-empowered Personalized Medicine system that brought medical “AI-empowered knowledge” to the tips of the fingers of clinicians and clinical researchers. AI-empowered knowledge is human interpretable knowledge that clinicians and clinical researchers can understand, trust, and effectively use in their day-to-day activities. KATY is an AI-empowered Personalized Medicine system built around two main components: A Distributed Knowledge Graph and A pool of eXplainable Artificial Intelligence predictors. As a stress test and due to the lack of personalized clinical responses, KATY experimented in a low prevalence and complex cancer: Clear cell renal cell carcinoma (ccRCC).
In the project, we carried out substatial work in order to achieve the goals of KATY. In particular, a very big part of the work carried out has been devoted to work towards the achievement of the objective of making use of the existing infrastructures and initiatives in genomic and health data pooling. Indeed, this is a very big objective whose results can go beyond the duration of the KATY project and may foster subsequent research. In this objective, we aim to make use of existing resources, and infrastructures and stay within existing initiatives. We have also paved the way to achieve the objective of demonstrating the potential and benefits of eXplainable AI technologies applied to Europe’s relevant genomic repositories in personalized medicine. In particular, we have developed eXplainable AI technologies, which can be the only ones that can be accepted by clinicians to support their clinical decisions.After defining an overall holistic neural network (KATY NN) that ingests input data for a specific patient with a specific tumor, along with specific drug treatment, and outputs the response. KATY NN is currently under development, we developed inner modules in different technologies and solving the different subproblems.
We have also built the KG and we have described how this should be queried by the overall platform. The KG is nearly fully deployed, and users may interact with it. Moreover, we implemented the overall architecture of the system along with the definition of the GUI. Then, the system and the related KATY NN have been tested with KATY clinical data.
We have also built the KG and we have described how this should be queried by the overall platform. The KG is nearly fully deployed, and users may interact with it. Moreover, we implemented the overall architecture of the system along with the definition of the GUI. Then, the system and the related KATY NN have been tested with KATY clinical data.
KATY developed a series of predictive AI models leveraging publicly available genomics and multi-omic data for decision-making in personalized medicine to predict disease recurrence from transcriptomic and histological data, response to targeted therapy, and response to immunotherapy leveraging proteogenomics. KATY disseminated all this knowledge with a sophisticated Distributed Knowledge Graph that incorporates the results of artificial intelligence models of use to researchers and clinicians and a simple patient-centric interface aimed to support patients alongside clinicians in guiding therapeutic choices along their path to treatment. Further, KATY provided evidence-based information on optimal treatment selection that can support health policy decision-making, whilst ensuring patient centricity in decision-making. In addition to the wealth of publicly available data collected in the KATY knowledge graph, a patient cohort of over 500 renal cancers subjected to combination targeted and immunotherapy has been generated. The resulting patient cohort has been used to refine the predictive models developed within the project.
The KATY project develops a network and collaborative infrastructure to user in the adoption of artificial intelligence into the EU health infrastructure. The pilot project in Renal cancer developed a proof of concept of the added value of artificial intelligence and served as the nucleus to develop the infrastructure to make this happen.
The KATY project develops a network and collaborative infrastructure to user in the adoption of artificial intelligence into the EU health infrastructure. The pilot project in Renal cancer developed a proof of concept of the added value of artificial intelligence and served as the nucleus to develop the infrastructure to make this happen.