Project Structure

Structure

KATY spanned 48 months and was split into three phases:

Phase 1 – The focus was on eliciting user requirements, mapping them to the technical specifications and defining the system architecture. A user-centric design process was followed involving health professionals, project stakeholders and the External Advisory Board. The implementation of system components was completed, leading to platform integration. This phase ended with delivery of the system prototype.

Phase 2 – The second phase capitalized on the developed system to perform a pre-study pilot. The deployed services and platforms were verified and evaluated for usability, performance and compliance with the requirements gathered in Phase 1.

Phase 3 – This final phase included a clinical study and the evaluation of the final system in real-life conditions. The results of the clinical study were analysed with regard to drug recommendations, and treatment data were collected.

Work Packages

The consortium activities were organized in nine work packages (WP) with clearly defined objectives and responsibilities for each specific partner.

WP1: Project Management

Work package 1 Provides an organizational framework and support mechanisms to enable a smooth project workflow. Its objectives range from optimal guidance and support of consortium partners, transparency and proper project documentation, effectiveness of project activities, to the timely and qualitative achievement of project results. One major goal is to manage the project as sustainable and resource-saving as possible. WP1 is led by coordinator UNITOV with close support of EURICE and covers the entire project duration.

WP2: Requirements and clinical needs

Work package 2 will provide the requirements and the specifications for the KATY system. It aims at carefully describe the clinical scenarios to clinical and non-clinical partners will define the requirements for privacy and security of data and procedures, and the architecture of the KATY platform. It is led by UEDIN and was completed by December 2021.

WP3: KATY Platform

Work package 3 provides the software architecture of the KATY platform that can manage the Distributed Knowledge Graph and can incorporate eXplainable Machine Learning predictors. It will implement and deploy the data lake with -omics information to feed the Distributed Knowledge Graph and will design a friendly and usable surface for mobile and desktop devices. Further, WP 3 will prepare KATY to host eXplainable AI Processors and test the security of the KATY platform. WP3 is led by DST and will be completed by December 2023.

WP4: Domain knowledge graph backbone construction; connecting-omic knowledge to patient data

Work package 4 will build the data infrastructure necessary for the AI methods used in KATY. Ethical and privacy concerns are evaluated to ensure that laws and norms are respected. In addition, WP3 aims at identifying and collecting publicly available resources to be exploited in the project, it will further collect and format public and consortium omic data, extract features from histological data and unifies public resources using a consistent, semantic knowledge graph representation. Finally, this WP will expose the unified resources as a service to feed the needs of the AI use cases. It is led by coordinator UNITOV and will be completed by December 2023.

WP5: A unified framework for explainable, graph-based AI

Work package 5 will design, develop, evaluate and deploy AI-based algorithms on the KATY platform to firstly enrich the transcriptomic and histological data, and to secondly predict the tumour response to targeted therapies and immunotherapies. More specifically, WPs is developing ML- based, explainable knowledge graph prediction models, comparing federate machine learning to traditional machine learning, and enriching transcriptome data by predicting the deregulation state of the tumour microenvironment and of biological pathways. It will develop 1) AI models predicting tumour molecular perturbations and the response to targeted therapies and immunotherapies from histological data, 2)AI models predicting response to small-molecules targeted therapies from transcriptomic data, and 3) AI models predicting neoantigen burden and to select neoantigens for cancer vaccines. This WP is led by CEA and will be completed by December 2023.

WP6: Large Scale Pilot and Validation

Work package 6 is contributing by actively engaging the clinical community to demonstrate the potential of models trained, using publicly available data to truly impact clinical decision making. It will supervise the execution and the validation of the AI models predicting response to small-molecules targeted therapies as well as immunotherapy in renal cancer. It will determine the potential of the user interfaces developed in KATY to enrich research, clinical and patient awareness around therapeutic choices, and model the cost efficiency of KATY versus current treatment selection pathways to optimize resource allocation and health policy decision making. It will further assess patient preference for KATY proposed treatment options, including in structured assessment all aspects of care provision, to ensure optimal patient centricity in decision making, and validate the predicted therapies obtained by the KATY platform using in vitro and in vivo models. WP6 is led by UG and will only be completed with project end in December 2024.

WP7 and WP9: Ethical and regulatory domain analysis

Work package 7 is dedicated to key ethical and legal questions within the project, taking into account the ethic requirements allocated in WP9. It serves as a legal helpdesk to the consortium, providing the a legal framework for sharing data within the consortium and with third parties, and analyses the socio-economic impact of AI interventions at the healthcare system level. It is led by UNIVIE and covers the whole project duration.

WP8: Communication, dissemination and exploitation

Work package 8 will ensure consistent dissemination and communication of project activities and results, safeguarding optimal visibility and a wide project outreach to relevant stakeholders as well as strategic planning and operational support of project exploitation through dedicated innovation management. It aims at creating visibility, disseminating results to targeted stakeholders and the scientific community, and to foster innovation capacity and the exploitation of results. It is led by EURICE and covers the whole project lifetime.