DATA INTEGRATION
Aim is to
• Develop completely new, and refine existing, models and algorithms for the prediction of cancer treatment response and clinical behaviour. These models are likely to have major impact across the spectrum of medical practice.
• We will thereby develop and implement a completely new paradigm in personalised cancer treatment which integrates multi- modality analyses and clinical characteristics in near-real time.
Data Integration is the central element of the MFICM approach. It brings together the themes of genomics (somatic, germline, ctDNA), imaging and clinical data and focuses on innovative approaches to analysis in a near real-time and longitudinal fashion. It uses prospectively collected data from clinical trials to explore these analyses. The overarching goal is to deliver a new paradigm in personalised cancer treatment which integrates multi-modality analyses in real-time that improves treatment selection and ultimately prognosis
Project 1) Using retrospective Breast Cancer Data (Metabric) we want to answer the following questions, by using novel AI and deep learning methods, like predictive clustering, redescription mining, deep learning and reinforcement learning.
• Prognosis: looking at imaging features, can we predict the risk of relapse of a tumour?
• Metastasis: looking at morphology, can we predict the site of the metastasis?
• Investigate HER2 status
• Can we identify visual features for the 11 molecular subtypes of breast cancer?
• Investigate new features: quantifying lymphoyctes or stroma, based on pre-processed data or by using novel learning.
Project 2) Using retrospective Breast Cancer Data (DETECT trial) we want to integrate radiological and clinical response with quantitative ctDNA metrics and radiomics features to improve accuracy.
Project 3) Using retrospective Breast Cancer Data (TRANSNEO trial) we want to visualise intra-tumour heterogeneity using digital pathology analysis and various imaging modalities (DWI or DCE MRI), while also exploring the correlation with genomic clonality
Project 4) Utilizing a pre-trial to one of our sponsored trials, WIRE (renal cancer), we are going to map MRI images and pathological sections to measure intra-tumour heterogeneity based on multi-parametric MRI
Project 5) Using data from ovarian cancer patients (NeoV, OV04) we will integrate imaging and molecular data streams, measure intra- and inter-tumour heterogeneity based on CT features and assess changes during treatment as well as the correlation with chemotherapy response and ctDNA.
Project 6) By utilizing patient derived xenograft models from breast cancer patients we will link genomic profiles with treatment strategies using in silico mechanistic models, to determine the mechanistic explanation to the observed behaviour.
Project 7) We will develop the online PREDICT tool into a novel Molecular-Predict tool by adding prognostic variable to the model.