NEW YORK (GenomeWeb) – An international consortium called the Personalized Engine for Cancer Integrative Study and Evaluation (PrECISE) intends to develop computational approaches for studying prostate cancer that consortium members hope will yield new insights into cancer's genetic mechanisms and help clinicians suggest more effective treatments for patients.
The list of participants includes IBM Research, Technikon, Technical University of Darmstadt, Aachen University Hospital, ETH Zurich, University of Zurich, Baylor College of Medicine, Curie Institute, and AstridBio Technologies. The project is partially funded by the European Union's Horizon 2020 research and innovation program, which has contributed over €3 million (over $3.3 million) of the €5.6 million (over $6.2 million) project budget. The balance of the funds will come from the Swiss government and IBM, according to Zsolt Torok, co-founder and CEO of AstridBio Technologies.
The PrECISE project, which was initiated by IBM Research in Zurich, officially launched on the first of this year and is expected to run for three years. The partners responded to a call from the European Union asking for applications from researchers interested in developing computational tools for personalized medicine.
"The aim is to develop different algorithms that allow us to understand tumor heterogeneity, understand better why drugs work and don't work, and come up with more effective therapies [and] in particular combination therapies," Julio Saez-Rodriguez, professor of computational biomedicine at the Aachen University hospital and PrECISE's scientific leader, told GenomeWeb.
He explained that the group decided to work with prostate cancer first because it has access to good specimens from a biobank hosted by one of the partners in Zurich. However, the computational tools that the consortium develops under PrECISE could be used to study any cancer type.
The partners will develop computational approaches that integrate genomic, epigenetic, transcriptomic, proteomic, and clinical information. They will combine data from a patient cohort with publicly available omics datasets as well as information from scientific literature. Consortium members will use the computational models that they develop to investigate prostate cancer's molecular mechanisms and to try to predict new targets for therapy. The models will also help researchers stratify patients based on clinically significant and insignificant disease leads, which should help minimize unnecessary surgeries and ultimately lower healthcare costs, according to the consortium.
As part of their efforts, the consortium will use the IBM Watson system to develop a genetic model of prostate cancer using the combined data types that can then be tailored and personalized to each patient's case, María Rodríguez Martínez, a research staff member in IBM Zurich's research laboratory, told GenomeWeb. They will use the model to test the outcome of administering different treatments to patients, she said.
The consortium will also develop web interfaces as well as storage and data management infrastructure that will hold the clinical and experimental data that the project generates. It will also develop a prototype of a clinical decision support tool that will help clinicians use the fruits of the consortium's efforts in clinical care. This portion of the project will be the responsibility of AstridBio Technologies.
Formerly known as Astrid Research, the company initially established branches in Hungary and Canada under separate but related monikers — Astrid Research and AstridBio Technologies respectively. It began using the name AstridBio Technologies across all business branches over three years ago. It has since refocused on supporting clinical users of genomics rather than on research-based projects, and has also begun offering software services rather than a bundled hardware and software solution for next-generation sequencing analysis.
Torok said that the company will use some of its existing infrastructure for the project and develop other parts from scratch. Specifically, AstridBio will use its SmartBiobank data management system to host clinical data collected as part of the project. It will also develop a separate system from scratch to host experimental data including genomic, transcriptomic, and proteomic data, he said.
The company is also developing two interfaces, one of which will offer tools for visualizing patient genomic and clinical data, and a second, separate prototype interface that will provide clinicians with decision support tools that leverage PrECISE data and models. "The system will be able to integrate the different kinds of datasets and advise medical doctors on how to treat the patients," he explained. The first release will be a prototype and not intended for routine clinical use, but further developing the tool for that purpose will be the logical next step, Saez-Rodriguez noted.
Prostate cancer is unique in two ways. For one thing, it occurs most often in older men. According to one set of statistics, most new cases are diagnosed in men who are between the ages of 70 and 74. Also, the disease progresses very slowly with cases spanning 10, 20, or even 30 or more years. The majority of prostate cancer cases can be classified as low risk or indolent tumors, meaning that it's unlikely that patients will have symptoms or die as a result of their cancer. But there are cases that are classified as high risk meaning that they will likely cause symptoms or kill the patient in the near future.
Being able to determine which category a patient's tumor falls into is crucial. Since prostate cancer progresses slowly, patients with these kinds of cancers may die as a result of other diseases or simply old age. For the indolent tumor cases, since their cancers are not-lethal, anti-cancer therapies may not be the best option as these have harmful side effects that outweigh their benefits. However, patients with the more lethal tumors can benefit from therapy if it's administered appropriately and quickly.
The PrECISE project aims to create tools that help clinicians stratify patients more effectively and connect them to treatment as needed. This will include developing computational models that predict a patient's life expectancy with or without prostate cancer, allowing researchers to compare the two outcomes, Torok said. Other models will help researchers predict patients' responses to different therapeutic options and whether treatments improve their outcomes.
The consortium also hopes to identify new biomarkers that could be used to better diagnose prostate cancer cases or could even serve as targets for treatment regiments, he added.
The three-year pilot is organized into nine sub-projects or "work packages" which build on each other. The first three sub-projects will focus on identifying cancer clones and any associated biomarkers by looking at genomic, transcriptomic, and proteomic data. For this portion of the project, the researchers will biopsy different parts of the prostate and try to identify the different clones present in the prostate using single-cell sequencing, Torok said.
They'll use the data collected from these projects to build mathematical models of clones found in the sample as well as the regulatory networks that influence clone behavior as part of other work packages planned for the three-year project. The researchers also plan to build interaction networks out of SWATH mass spectrometry data from patient prostate cancer samples, prostatic cell lines, and public proteomic datasets. This will be a task for Saez-Rodriguez's group along with some other partners in the consortium.
In another sub-project, researchers will build a molecular map that incorporates genomic, transcriptomic, and proteomic data. They'll then use the map to detect genetic alterations in clusters of proteins and identify any relevant pathways and molecular mechanisms that are involved in prostate cancer's progression. The researchers also plan to develop a classifier that will enable them to group patients into clinically meaningful groups.
A fifth sub-project will focus on developing a mathematical model that combines the important molecular elements identified in previous sub-projects with clinical data. Researchers will try to find relationships between patient survival, therapies, and the omics background of the patient. The sixth sub-project will focus on experimentally validating any new biomarkers identified over the course of the consortium's analysis.
Two other sub-projects focus on sharing research results within and outside the consortium. The experimental and clinical data collected as part of the project will be available to non-consortium members for free but they will have to contact the consortium for access, Torok said. They'll also have to assent to the consortium's guidelines that govern use of patients' data to ensure its security and protect their privacy before being granted access. The consortium will also provide training for community members that want to use the resources, and will put mechanisms in place to ensure the long-term success of the project.
Meanwhile, models developed by the consortium will be incorporated into Watson, IBM's Rodríguez Martínez said, adding that she will continue developing and improving them within Watson even after the project wraps.
"I look at it as a feedback loop," she said. "Initially Watson gives us some input, we work on it, develop it, make it richer, and in the end we feed it back into Watson to increase its performance and capabilities."
In addition to this project, Watson is being used by institutions such as the New York Genome Center. There, the system supports a pilot study that will integrate and analyze sequence and other kinds of data from a 200-cancer patient cohort. Previously, Watson was used for a glioblastoma study at NYGC that focused on about 20 recently diagnosed patients.