Multimodal data integration
Prostate cancer is a leading cause of cancer death amongst men in Europe, but also a disease that is prone to over-treatment.
Our goal is to develop predictive computational technology that can exploit and integrate multiple molecular and clinical data to improve our understanding of disease mechanisms, stratify patients and inform clinicians about optimized strategies for therapeutic intervention.
We are part of PrECISE, a project funded by the European Union’s Horizon 2020 research and innovation programme.
PrECISE aims to improve patient risk-stratification and treatment in prostate cancer by developing new computational approaches to exploit next generation molecular data. The expected outcome is the development of a predictive computational technology that can exploit molecular and clinical data to improve the understanding of disease mechanism and to inform clinicians about optimized treatment strategies.
Composite biomarkers and drug discovery
Combining patient-specific diagnostic measurements with well-established disease-specific knowledge and cutting-edge research is challenging due to the different data sources and types.
We combine molecular data, data from public data bases and literature using state-of-the-art machine and deep-learning methods adapted to suit the specific nature of the data. After combining the results in interaction networks, graph theoretic approaches are applied to identify disease signatures that can serve as prognostic markers and propose combinations of drugs and treatment strategies.
The molecular-level understanding of disease onset and progression of prostate-cancer are largely unknown. Specifically, stratification of intermediate prostate tumor states based on current markers is difficult.
We integrate multiple types of molecular data into mathematical disease models to stratify patients and propose actionable clinical statements. We combine mass spectrometry, next-generation-sequencing and histopathology image data to obtain an accurate estimate of disease state and aggressiveness.
Understanding tumor heterogeneity is one of the major challenges in cancer medicine. Commonly multiple subclones coexist within a single tumor impeding application of generalized models for disease prognosis.
The current frontier is to move towards personalized models taking into account tumor diversity and patient-specific traits. For each patient we combine mutation data from multiple biopsies and statistical inference approaches to identify and decipher the clonal composition of a tumor. This enables us to learn patient-specific models and understand complex disease-mechanisms.
Ask the experts
IBM Research scientist
IBM Data scientist