Personalized medicine is based on the exploitation and analysis of large quantities of data from a variety of sources including genomic, epidemiological, and medical imaging. To be able to do this, cross-referencing and aggregating mutually intelligible data is necessary, even when they come from very different sources.

Interoperability is a challenge for biomedical research. This can be addressed through semantics, or the organization of medical knowledge into a finite set of classes, such as scores or classifications.

However, semantics usually is not well adapted to a wide range of knowledge types or different organizations. Furthermore, biomedical data models vary based on the goal of the research and on the community that will use the data.

In Switzerland, the Swiss Personalized Health Network (SPHN) has attempted to address these issues by developing a national infrastructure that has been adopted by all Swiss university hospitals and academic institutions. The SPHN uses a common semantic framework and data organization to stimulate research and innovation for personalized medicine in Switzerland.

"Despite major investments over the past decade, there are still major disparities," said Dr. Christian Lovis, director of the department of radiology and medical informatics at the University of Geneva faculty of medicine and head of the division of medical information sciences at the University Hospitals of Geneva (HUG), in a statement. "This is why we wanted, with our partners and the SPHN, to propose a strategy and common standards that are flexible enough to accommodate all kinds of current and future databases."

A three-pillar strategy

All data generated in the framework were compiled and maintained by the SPHN Data Coordination Center (DCC). The center coordinates the development of the specification structure and semantics of the dataset. The project is grouped around three pillars.

The first pillar relies on every research project providing a list of variables that users need. The variables are validated and published on a public platform. The list is an evolving database and the authors noted that as more use cases arise, new encodings will be added.

The second pillar involves data storage and transport to the resource description framework (RDF). All data are transformed from a relational model representation into a graph representation.

The third and final pillar uses converters to transform the data into purpose-specific data models, serializing the data into common data formats. The data can even be processed by research-enabling software or machine-learning pipelines.

"Our aim has therefore been to unify vocabularies so that they can be communicated in any grammar, rather than creating a new vocabulary from scratch that everybody would have to learn anew," Lovis explained. "In this sense, the Swiss federalism is a huge advantage: it has forced us to imagine a decentralised strategy, which can be applied everywhere. The constraint has therefore created the opportunity to develop a system that works despite local languages, cultures and regulations."

The framework makes it possible to apply specific data models to be adapted to the formats required by a particular project as the very last step of the process. For instance, they could use the U.S. Food and Drug Administration (FDA) format in the case of collaboration with an American team, or any other specific format used by a particular country or research initiative. This leads to a mutual understanding of data and a huge time saving.

The strategy has been implemented stepwise in Switzerland since the middle of 2019 in the framework of the Swiss Personalized Health Network.

"Swiss university hospitals are already following the proposed strategy to share interoperable data for all multicentric research projects funded by the SPHN initiative," said Katrin Crameri, PhD, director of the personalized health informatics group at Swiss Institute of Bioinformatics (SIB) in charge of the SPHN DCC.

Do you have a unique perspective on your research related to bioinformatics or drug development? Contact the editor today to learn more.

Algorithm expands what scientists can uncover from individual cells
A new software tool can help researchers further utilize single-cell RNA sequencing data to learn more about individual cells by incorporating biologically...

Survey: Scientists want to boost use of lab automation
In the era of big data and new analytical technologies, research laboratories are rapidly adopting new lab automation equipment and processes. While most...

New method allows for study of genetic changes in individual DNA molecules
An improved sequencing method, called nanorate sequencing, allows for the study of genetic changes in human cells with unprecedented accuracy. The new...