Technical Deliverables (Public)

Submitted:

WP1:

D1.1 "Final regulatory network inference"
Editor: Maria Rodriguez Martinez (IBM)
Contributors: Roland Mathis (IBM), Matteo Manica (IBM), Ali Oskooei (IBM), Sara Al-Sayed (TUDA), Dominik Linzer (TUDA), Heinz Koeppl (TUDA), Hua-Sheng Chiu (BCM), Pavel Sumazin (BCM)

This report describes graphical user interfaces developed for genomic and clinical data entry for PrECISE.

D1.2 “Final clone interference”
Editor: Pavel Sumazin (BCM)
Contributors: Dorothea Rutishauser (USZ),Jelena Cuklina (ETH),Peter Wild (UZH),Dorothea Rutishauser (UZH),Matteo Manica (IBM)

Knowledge about tumour clonal evolution can help interpret the function of genetic alterations by pointing out initiating events and events that contribute to the selective advantage of proliferative, metastatic, and drug-resistant tumour subclones. Clonal evolution can be reconstructed from estimates of the relative abundance (frequency) of subclone-specific alterations in tumour biopsies, which, in turn, informs on the cellular composition of each tumour subclone. However, estimating these frequencies is complicated by the high genetic instability that characterizes many cancers. Models for genetic instability suggest that copy number alterations (CNAs) can dramatically alter mutation-frequency estimates and thus affect efforts to reconstruct tumour phylogenies. Our analysis suggests that a detailed accounting of CNAs is required for accurate mutation frequency estimates, and that such accounting is impossible for many cancer types using molecular profiling of one biopsy per tumour. Instead, we propose an optimization algorithm, Chimaera, to account for the effects of CNAs using profiles of multiple biopsies per tumour. Analyses of simulated data and profiles of a prostate cancer, hepatocellular carcinoma, and Wilms’ tumours suggest that Chimaera estimates are consistently more accurate than previously proposed methods, resulting in improved phylogeny reconstructions, and the discovery of recurrent initiating mutations and key tumourigenesis events. We used Chimaera to analyse prostate cancer profiles, identifying mutations associated with initiating tumour subclones as well as tumour subclones that respond or are resistant to therapies.

D1.3 “Clonal classification of tumors” Editor: Dorothea Rutishauser (USZ)
Contributors: Pavel Sumazin (BCM), Jelena Cuklina (ETH), Peter Wild (UZH), Dorothea Rutishauser (UZH), Matteo Manica (IBM)

For this deliverable we used clonality interference to select tumours for the follow up by ultra- deep sequencing (done in WP2), and led efforts to improve clonality interference.

WP2:

D2.1 "Targeted ultra-deep sequencing of cancer-gene loci"
Editor: Maria Rodriguez Martinez (IBM)
Contributors: Pavel Sumazin (BCM), Ulrich Wagner (UZH), Peter Wild (UZH)

We set out to test our ability to infer mutations and their clonality using 10 castrate-resistant prostate cancer (CRPC) tumor biopsies. Conclusions from studying these biopsies informed methodology for selecting additional tumors for profiling by exome sequencing, for predicting mutations and exome sequencing, and allowed us to devise criteria for estimating mutation cellularity (a necessary step for inferring clonality in WP1). Clonality inference is a building block for prognostic-biomarker inference in WP3, and tumor classification in WP4.

D2.2 "Ultra-deep sequencing of prognostic biomarkers"
Editor: Pavel Sumazin (BCM)
Contributors: Pavel Sumazin (BCM), Hyunjae Ryan Kim (BCM), Maria Rodriguez Martinez (IBM), Peter Wild (UZH), Laura De Vargas (UZH)

This report describes an effort to sequence prognostic biomarkers for prostate cancer. We designed a prognostically predictive biomarker panel that targets thirty-six commonly mutated predictive genes and used this panel to profile forty-two Castration Resistant Prostate Cancer (CRPC) patients.

D2.3 “Targeted profiling of prospective cohort”
Editor: Dorothea Rutishauser (UZH)
Contributors: Dorothea Rutishauser (UZH), Matteo Manica (IBM), Jelena Čuklina (ETH), Peter Wild (UZH), Pavel Sumazin (BCM)

In D2.3 we report the status of the targeted profiling efforts of the ProCOC and ZTMA76/80 cohorts. To identify driving mutations and characterize intra-tumor heterogeneity, that is arising due to evolution of several co-localized clones, we have conducted in-depth genomic and proteomic analysis of multiple biopsies of 3 representative patients from ProCOC cohort, reported previously as Prostate Cancer Variance (PCV) study. Using SWATH-MS and targeted extraction of peptide to spectrum-matches, we have characterized the heterogeneity of proteome in prostate cancer samples (Chapter 2).

D2.4 “A complete catalogue of tageted profiles”
Editor: Sumazin, Pavel (BCM)
Contributors: Jelena Cuklina (ETH), Peter Wild (UZH), Dorothea Rutishauser (UZH), Matteo Manica (IBM)

We hereby present the complete catalogue of targeted profiles in WP2 “Identification of sub- clonal genomic alterations”. We used targeted sequencing, OncoScan copy number arrays, and SWATH-MS to profile prostate cancers. In total, 94 tumour samples were profiled using targeted sequencing, 85 samples were profiled for copy number alterations, and 82 samples were profiled for protein expression. Of these, 67 tumour samples were profiled for mutations, copy number alterations, and protein expression. These included samples from with 5 patients using at least 5 regions per tumour and at multiple time points. We used these 67 samples to infer clonality and predict the effects of tumour subclone specific mutations on protein expression. The remaining samples are available for prospective testing of inference methods, as proposed in Tasks 2.1 “Profiling selected PC driver loci” and 2.2 “Ultra-deep profiling of prognostic biomarkers in selected proCOC and MetaProC samples”.

WP3:

D3.1 "Computational pipeline to extract prior network information at the proteomic level"
Editor: Luis Tobalina (UKAACHEN)
Contributors: Julio Saez-Rodriguez (UKAACHEN), Maria Rodriguez Martinez (IBM), Laurence Calzone (CI)

This document gives an overview of the Omnipath database and its accompanying Python module Pypath. Omnipath gathers 54 resources, including 27 high-confidence literature curated signaling resources, providing and easy, unified and convenient entry point to much of the protein interaction knowledge available. Overall, Omnipath and Pypath greatly facilitate the integration and extraction of biological prior knowledge for analysis and model building, and they can be incorporated into wider data processing pipelines. Pypath and Omnipath are available here.

D3.2 "Network reconstruction algorithms for MS data"
Editor: Dominik Linzner (TUDA)
Contributors: Sara Al-Sayed (TUDA), Tabea Treppmann (TUDA), Sikun Yang (TUDA), Dominik Linzner (TUDA), Heinz Koeppl (TUDA), Laurence Calzone (CI)

As modern biological-data measurement technologies have been generating more and more high-dimensional data in recent years, data-driven protein–protein interaction (PPI) network reconstruction has garnered a lot of interest. MS time-course perturbation data is expected to be made available in the course of the PrECISE project. Our goal is to develop network reconstruction algorithms appropriate for this kind of data. This report introduces two approaches that have been the focus of work at TUDA towards network reconstruction from MS data.

D3.3 "First data-driven reconstruction of context-specific network"
Editor: Maria Rodriguez Martinez (IBM)
Contributors: Sara Al-Sayed (TUDA), Dominik Linzner (TUDA), Heinz Koeppl (TUDA), Matteo Manica (IBM), Roland Mathis (IBM), Ali Oskooei (IBM)

This deliverable contains a description of the algorithms implemented for data-driven reconstruction of context–specific networks. The reconstruction of context–specific networks, e.g., prostate cancer–specific, treatment-specific) can help to better understand and describe complex disease progression, and ultimately reveal unknown interactions fundamental to the generation of new treatment hypotheses. In this work, we present a set of algorithms developed by the partners of the PrECISE consortium, and discuss the results of the analysis of various datasets using these devised algorithms.

D3.4 "Identification of systematic alterations of networks for different prognosis and for different clonal composition"
Editors: Matteo Manica (IBM), Pavel Sumazin (BCM)
Contributors: Hyunjae Ryan Kim (BCM), Maria Rodriguez Martinez (IBM), Laura De Vargas (UZH)

We report on our use of mutations targeting mutated genes to predict prognosis in a larger prostate cancer cohort, namely, the TCGA PRAD dataset. Our results suggest that mutations at the loci of our selected genes are predictive of the overall outcome of prostate cancer patients.

WP4:

D4.1 "Interactome of molecular interactions in prostate cancer"
Editor: Maria Rodriguez Martinez (IBM)
Contributors: Roland Mathis (IBM), Matteo Manica (IBM), Pavel Sumazin (BCM), Luis Tobalina (UKAACHEN)

The goal of this deliverable was to integrate patient data of the ProCOC cohort into a molecular map specific to prostate cancer. The integrated molecular map is also called network of molecular interactions or interactome. The interactome built from patient data is based on protein measurements of tumor samples from patients. By combining multiple inference methods a data-driven consensus interactome was constructed. To assess the performance of the inference methods we applied the methods to simulated data on known biological networks. To enrich the data driven interaction network, separate interactomes were extracted from publicly available interaction data bases and from publications. We compared the interactomes derived from data and publications to interaction networks from public databases. By combining the interaction networks derived from data, public databases and publications a consensus interactome was built.

D4.2 "Catalogue of molecular alterations and dysregulated pathways"
Editor: Pavel Sumazin (BCM)
Contributors: Roland Mathis (IBM), Matteo Manica (IBM), Maria Rodriguez Martinez (IBM), Dominik Linzner (TUDA), Laura De Vargas (UZH)

Patient mutations for the ProCOC cohort were projected on a prostate–specific network topology to better explain observed clinical phenotypes and tumor sub-typing

D4.3 "Robust cross-cohort clinical patient classifier"
Editor: Matteo Manica (IBM)
Contributors: Matteo Manica (IBM), Ali Oskooei (IBM), Maria Rodriguez Martinez (IBM)

In this report a method called PIMKL developed by partner IBM is presented. PIMKL is a machine learning approach to analyze different omics data in a disease specific fashion that makes use of an interactome and pathway annotation as prior information. PIMKL was applied on different cohorts made available from partner UZH and ETH.

WP5:

D5.1 "Generic model"
Editor: Pauline Traynard (CI)
Contributors: Luis Tobalina (UKAACHEN), Tiannan Guo (ETH), Laurence Calzone (CI)

Logical models are simple, require in principle no quantitative information, and can be hence applied to large networks combining multiple pathways. This deliverable presents some analyses made on a model, which is derived describing the network dynamics in specific contexts (dependent on initial conditions or perturbations for instance). Some physiological conditions are first simulated in order to validate the model. The validation concerns both the choices made on the topology of the network (players and the interactions), and on those on the logical rules that are associated to each of the variable of the model. Then, some modifications of the wild type are explored, corresponding to mutants. The model still has to be validated on patient or clone profiles, and on data by performing some computations on the activity of pathways in the PC39 patients.

D5.2 "Proteomic data sets in cancer cell lines"
Editor: Jelena Cuklina (ETH)
Contributors: Dominik Linzner (TUDA), Luis Tobalina (UKAACHEN), Julio Saez-Rodriguez (UKAACHEN), Ruedi Aebersold (ETH), Ludovic Gillet (ETH), Ana Filipa Goncalves (UZH), Helena Fischer (UZH), Peter Wild (UZH), Arnau Montagud (CI), Laurence Calzone (CI)

This report describes two proteomic profiles – full cell proteome and phosphoproteome – in castration sensitive and castration resistant cell lines upon drug perturbations. Full cell proteome is already being used by TUDA and IBM in WP3, while the phospho-proteomic data will be analysed in task 5.4.

D5.3 "Patient-specific models"
Editor: Arnau Montagud (CI)
Contributors: Luis Tobalina (UKAACHEN), Jelena Cuklina (ETH), Pavel Sumazin (BCM), Jonas Beal (CI), Laurence Calzone (CI)

This deliverable describes the instantiations of our generic Boolean model of prostate cancer to adapt it for different patients.

D5.4 “List of possible drug targets (intervention points), and of individual and combination of drugs”
Editor: Luis Tobalina (UKAACHEN)
Contributors: Montagud, Arnau (CI), Szalai, Benze (UKAACHEN), Čuklina, Jelena (ETH)

One of the objectives of WP5 is to provide a list of candidate drugs and combinations of drugs for cell lines and for specific patients. To get to this point, we leverage the work described in deliverables D5.1, D5.2 and D5.3 and run a series of computational analysis that aim to provide a list of targets of interest in prostate cancer. This deliverable describes how the prostate cancer Boolean model is mixed with prostate cancer related experimental data to derive cell line and patient specific target candidates. As an alternative to Boolean modeling and for comparison purposes, a data based approach to finding cell line specific targets is also used. The final list of proposed targets constitutes a starting point for the design of validation experiments in D6.5.

WP6:

D6.3 "Generate SWATH proteome profiles from sample punches prepared in D6.2"
Editor: Matteo Manica (IBM)
Contributors: Dorothea Rutishauser (UZH), Laura De Vargas Roditi (UZH), Qing Zhong (UZH), Jelena Cuklina (ETH), Arnau Montagud (CI), Ali Oskooei (IBM), Maria Rodriguez Martinez (IBM)

This report consisted of acquiring mass spectrometric proteomics data from the generated punches from the three cohorts described in D6.2 “Generate multiple punches from each validation sample in D6.1”:

(1) Fresh frozen (FF) prostate tissue from normal and tumor areas of prostate cancer (PC) patients with radical prostatectomy from the prostate cancer outcomes cohort study (ProCOC), also referred as PPP1 cohort.

(2) A set of retrospective formalin-fixed paraffin-embedded (FFPE) PC cohort of radical prostatectomy specimens (ZTMA76)

(3) FFPE samples from castration resistant prostate cancer (CRPC) patient tissue (not acquired yet)

Preliminary results of analysis of PPP1 data indicated, that findings obtained with PC39 project, are valid in this dataset as well.

D6.4 "Generate amplicon sequencing profiles from sample punches prepared in D6.2"
Editor: Matteo Manica (IBM)
Contributors: Pavel Sumazin (BCM), Hyunjae Ryan Kim (BCM), Maria Rodriguez Martinez (IBM), Peter Wild (UZH), Laura De Vargas (UZH)

Here, part of the samples from MetaProc and ZTMA204 have been sequenced in different tumor areas as described in D6.2. The remaining samples were prepared for sequencing at multiple time points to better inform and validate algorithms developed by BCM and IBM in WP1 “Regulatory network and clonality inference in prostate cancer tumors”.

D6.5 “Generate cell line drug sensitivity/resistance validation assays”
Editor: Robert Alfoldi (ABT)
Contributors: Laszlo Hackler (ABT), Laszlo Puskas (ABT), Arnau Montagud (CI), Luis Tobalina (UKAACHEN), Bence Szalai (UKAACHEN), Vigneshwari Subramanian (UKAACHEN), Jelena Čuklina (ETH), Dominik Linzner (TUDA), Matteo Manica (IBM), Ali Oskooei (IBM)

The main task of D6.5 was to generate drug sensitivity cellular assays in order to validate computational models developed in work package 5 (WP5). We have generated cell line drug sensitivity/resistance validation assays based on several prostate cell lines. In our protocol we have optimized an end-point cell viability assay as well as a cell perturbation analysis technique that is capable of measuring androgen dependent and independent drug-induced changes in cell viability, proliferation and adhesion in a real-time fashion.

WP7:

D7.1 "Data input and input interface"
Editor: Pavel Sumazin (BCM)
Contributors: Pavel Sumazin (BCM), Zsolt Torok (ABT)

Prostate specific gene regulatory networks were inferred by integration of RNASeq data measurements from TCGA PRAD and ProCOC cohorts using HIPSTER framework.

D7.2 "Design and integrate pathway visualization"
Editor: Laurence Calzone (CI)
Contributors: Matteo Manica (IBM), Nicolas Sompairac (CI), Laszlo Puskas (ABT)

The visualization of genomic, transcriptomics and proteomics data onto maps describing molecular processes proves to be informative. At a first glance, by mapping these data onto maps of processes known to be altered in cancer, some signalling pathways or processes show a higher expression in specific patients or groups of patients and suggest which deregulations are associated with which (groups) of patients. In this deliverable, we demonstrate how visualization of expression data or of pathway activity can provide a first understanding of differences between patients with different clinical information.

D7.3 "Re-implement methods"
Editor: Robert Alfoldi (ABT), Pavel Sumazin (BCM)
Contributors: Laszlo Puskas (ABT), Bence Szalai (UKAACHEN), Laurence Calzone (CI)

We report on the re-implementation and integration of our interface to accommodate the variety and changes in the design of clinical and wet-lab data, as well as analysis methods and products developed by PrECISE into a secure system with varied storage, visualization, and integration capabilities. The system provides a single-point access to tools developed by our partners, as well as molecular data and pathways, dedicated storage, and data and analysis visualizations. The re-implemented interface allows researchers and clinicians to compare their PC results with a wide-range of mutation, transcriptomic, proteomics and clinical findings as well for better outcome estimations. The provided system is ready to be used within PrECISE and is projected to become available for other researchers and clinicians following the completion of the project.

D7.4 “Integrate methods, including ACSN and Watson”
Editor: Róbert Alfoldi (ABT), Matteo Manica (IBM)
Contributors: Laszlo Puskas (ABT), María Martìnez Rodríguez (IBM), Jelena Čuklina (ETH)

We report on the work performed to integrate IBM Research services developed during the project in the SmartBiobank (SBB). IBM and ABT collaborated to link the different APIs allowing the display of the IBM Cloud services as frames inside a page hosted on the SBB Cellanalyzer. Analogously to the work performed in D7.3 to integrate ACSN we show how the integrations has been performed and how SBB users can have direct access to the IBM Cloud services implemented in the context of PrECISE: COSIFER, INtERAcT, Chimaera and PIMKL. The integration is compatible with Chrome and Safari and can be freely used for academic/research purposes. The impact in clinical studies of the services developed has been exhibited and validated by applying the algorithms on the data produced by partners UZH and ETH. The analysis and results obtained can be found in WP1, WP4 and WP6 deliverables.

WP8:

D8.2 "Data Management Plan (DMP)"
Editor: Peter Wild (UZH)
Contributors: Martina Truskaller (TEC), Sandra Lattacher (TEC), Luis Tobalina (UKAACHEN), Ulrich Wagner (UZH), Qing Zhong (UZH), Zsolt Torok (ABT)

The purpose of the DMP is to provide an analysis of the main elements of the data management policy that will be used by the applications with regard to all the datasets that will be generated by the project. The DMP should ensure that most important aspects regarding data management, like metadata generation, data preservation, and responsibilities, are identified in an early stage of the project. This ensures that data is well-managed during the project and also beyond the end of the project. Data which will be generated in the course of the project include output data of random number generators, PUF output data, measurement data, and source code. As the DMP is an incremental tool, it will be adapted in the course of the project.

Ethics Deliverables

D10.1 "HCT Requirement No 2"
Details on cells/tissues type and authorisation by primary owner of data (including references to ethics approval) is provided.

D10.2 "POPD Requirement No 6"
Detailed information on the informed consent procedures that are implemented in regard to the collection, storage and protection of personal data has been submitted.

D10.3 "HCT Requirement No 3"
Details on cells/tissues type are provided, as well as details on the biobank to access it.

D10.4 "POPD Requirement No 5"
Detailed information is provided on the procedures that are implemented for data collection, storage, protection, retention and destruction and confirmation that they comply with national and EU legislation.

D10.5 "POPD Requirement No 4"
This deliverable includes the amendment summarizing the PrECISE project that has been approved by the Cantonal Ethics Committee of Zurich.

D10.6 "NEC Requirement No 8"
The deliverable confirms that the ethical standards and guidelines of Horizon2020 are rigorously applied, regardless of the country in which the research is carried out.

D10.7 "NEC Requirement No 9"
The deliverable provides details on the material which will be imported to/exported from EU.

D10.8 "POPD Requirement No 7"
The deliverable includes explanation why data is not publicly available.

D10.9 "HCT Requirement No 1"
Details on cells/tissues type and ethics approval is provided.