Skip to main content

Computational ONcology TRaining Alliance

Periodic Reporting for period 1 - CONTRA (Computational ONcology TRaining Alliance)

Reporting period: 2018-01-01 to 2019-12-31

Cancer is a major cause of death and suffering, rendering it a huge concern to the general public. Consequently, it has been targeted by the application of the molecular techniques developed during the last 20-25 years, thereby improving diagnosis and treatment. Nevertheless, although current biomarker and treatment concepts often are successful initially, they subsequently frequently fail to achieve durable drug response and long-term survival for cancer patients. We study somatic evolution in cancer, which is the reason for the latter as well as many other of cancer's problematic properties and, consequently, a very promising approach to improve cancer treatment. Fortunately, single-cell genomic sequencing has recently begun to provide the opportunity to unprecedented detailed insights into tumour evolution, and new techniques are emerging for assaying the spatial distribution of the tumour heterogeneity caused by tumour evolution. Analysing and developing methods for these emerging data sets have been under-researched areas that lie at the intersection of medical, evolutionary biology, and computational research. In CONTRA, European researchers with complementary expertise have joined forces in order to collectively facilitate the training of future European computational cancer researchers.

CONTRA’s objectives are:
(1) To develop novel, powerful single-cell models and tools for somatic evolution in cancer, including spatial aspects.
(2) Among the partners and in collaboration with other groups, apply new tools in cutting edge cancer studies.
(3) Facilitate translational training including academia, software industry, and pharmaceutical and biotech industry.
(4) Develop novel methods and tools with the potential for improving clinical cancer care.
Work Package 1
Nico Borgsmüller developed BnpC, a novel computational method for clustering individual cells into clones and infer their genotypes based on noisy mutation profiles. BnpC employs a Dirichlet process mixture model coupled with a Markov chain Monte Carlo sampling scheme, including a modified non-conjugate split-merge move and a novel posterior estimator to predict clones and genotypes.

Mandi Chen’s current work focuses on reconstructing tumor phylogeny from direct library preparation (DLP) single-cell genome sequencing data. An EM-based algorithm has been designed and being developed to infer cell lineage while modeling somatic evolution. This method exploits intratumor heterogeneity by inferring latent copy number states as well as breakpoints among the genomes.

Fausto Fabian Crespo has developed a structured coalescent model for the estimation of functional clones and their corresponding growth rates, from bulk and single-cell sequencing data.

Mohammadreza Mohaghegh Neyshabouri
Mohammadreza has developed a probabilistic model for tumor progression, in which driver genes are clustered into several ordered driver pathways.

Work package 2
Hania Kranas aims to understand the causes of variability of mutation rates along the genome, by searching for the mutual influences of DNA damage, repair, mutations, and 3D chromatin structure.

Shadi Darvish Shafigh's project is entitled TUMOROSCOPE: Inferring a map of tumor subclones from Spatial Transcriptomics and bulk DNA sequencing data.

Reda Keddar's work in this period focused on the identification of mechanisms of resistance to immune checkpoint blockade therapy in the context of both genetic and microenvironment intra-tumour heterogeneity. In particular, a characterisation of the microenvironment determinants of response was performed, leveraging on RNA- and T cell receptor-sequencing data from a multi-regional colorectal cancer cohort.

Monica Valecha is exploring how mutations accumulate in cancer genomes by exploring intra-tumoral heterogeneity at the single-cell and bulk levels. We have started with detecting single nucleotide variants from single-cell whole-genome data and we are currently benchmarking the current methods available for variant detection. At the same time, we are exploring mutational signatures in association with cancer and its progression.

Work package 3
Michael Schneider has been working on extending and developing the concept of mutational signatures in the domain of single-cell data. Specifically, he is working on the development of statistical methods and algorithms that allow us to extend the concept of mutational signatures to single-cell genomics data and to make use of the new perspective that these data offer to better understand the mutational landscape of cancer genomes.

Jose Bonet has worked on the single-cell DNA data clustering (BnpC). The current focus of the project aims at detecting, through Nanopore sequencing, how damage appears when cells are treated with different genotoxins.

Arthur Dondi's work has been focused on producing and analyzing high-grade serous ovarian cancer single-cell full-transcripts data. A pilot experiment was designed to produce and sequence single-cell PacBio long-reads, and a pipeline has been developed to pre-process this unpublished data-type, which includes trimming, de-multiplexing, and mapping. Tools were created either from scratch or based on existing software.

Senbai Kang performs a project called A Mathematical Model to Infer Tumor Evolutionary History from Single-cell DNA Sequencing Data. The project is concerned with understanding intratumour heterogeneity, which is the cornerstone of developing effective cancer treatment and precision medicine. The development of single-cell DNA sequencing technology remarkably increases the resolution of DNA profiles to the single-cell level, facilitating inference of phylogenetic trees with individual tumor cells as leaves.

Work package 4
Paula Martin-Gonzalez has been developing further the concept of habitats extracted from radiological imaging. She is especially working under the hypothesis that regions with different appearance in radiological images due to different tumour microenvironment features are subjected to different selection forces and thus drive the prevalence of different tumour genomic clones in each of them.

Yeman Brhane Hagos
Yeman Brhane Hagos has demonstrated considerable progress during the first year of his Ph.D. focusing on the development of novel deep-learning approaches to deciphering cancer evolution through combined histopathology and genomics. Using these approaches, Yeman hopes to identify novel predictive markers of cancer progression in lung cancers and myeloma for his Ph.D. He has developed a deep-learning method that outperformed state-of-art deep-learning methods for classifying cell type in multiplex immunohistochemical slides, demonstrating improved recall and F1-score.
The methods developed by CONTRA ESR are building on state of the art methodologies and constitute methods that progress beyond the state of art in the application area of cancer genomics. In addition, these 15 well trained and ESRs with well-developed contact networks will leave our training program and participate in translating the academic approach into accessible software products and will also make sure that these approaches have an impact on clinical research or even clinical decision making.