CONTRA’s specific research and innovation objectives have been:
To develop novel, powerful single cell models and tools for somatic evolution in cancer, including spatial aspects, building on methods for classic evolution as well as probabilistic machine learning in general.
Among the partners and in collaboration with other groups apply new tools in cutting edge cancer studies - with a special focus on clinical and pharmaceutical relevance.
Construct a novel translational training triangle between academia, software industry, and pharmaceutical and biotech industry.
Develop novel methods & tools with potential for improving clinical cancer care.
Regarding the first objective. We have developed and published a number of computaional methods building on machine learning methodology, such as probabilistic graphical model.
For instance, the CACTUS method for analysing the evolutionary history of tumours and the phenotypes of their clones from integrated whole exome sequencing (WES), single cell RNA sequencing (scRNA-seq) and additional type of data that induces a clustering of the single cells [cite]. CACTUS leverages the additional prior information on the clustering to better map the single cells to the clones identified from WES, based on the alternative and total read counts found in scRNA-seq. Another example, is the SIEVE method, an approach combining statistical phylogenetic model of tumor evolution with a probabilistic model of the raw single cell DNA sequencing data (scDRNA-seq) [cite]. The latter method allows us to simultaneously reconstruct the partial order of occurrence of gene mutations, and calls somatic single nucleotide variants.
Both CACTUS and SIEVE are single cell models and tools for somatic evolution in cancer, provide major innovative solutions and can be used to advance knowledge on cancer progression.
Regarding the second and fourth objective. For example, we have developed and applied and approach that takes advantage of biopsies from HGSOC patients to studying the biological association between different radiomic habitats and pathology. The development method extracts radiomic habitats from standard of care CT scans in High Grade Serous Ovarian Cancer (HGSOC). HGSOC patients most often have multiple metastatic lesions at diagnosis, which makes it hard to quantify the intra and inter lesion heterogeneity. However, clustering patch-wise texture features into radiomic habitats, render quantitative studies possible. The results of this studies have been published in European Radiology.
Regarding the third objective. Our innovative training network included partners not only from academia but also from software and biotech industry, such as Ardigen and Gradient, as well as pharma (Merck Healthcare). The training events exposed the ESRs to not only research and academic tutors, but also included interactions and tutoring from these partners. Specifically, a highly programming intensive 2nd workshop: “Software development” was organised by Ardigen, 3rd summer school: “Promoting your career” was co-organized by Gradient, while our partner PI from Merck, Eiike Staub, gave lectures on two training events, (1st workshop: “kick-off and background” and 1st summer school: “Handling single cell data”).
