Transforming cancer research
EU-funded PREDECT project focused in developing novel in vitro cancer models and finding better target proteins in order to achieve greater predictability of drug efficacy. WebMicroscope served as an essential research tool by enabling the storage and sharing of the image data and opens now new insight with Deep Learning Image Analysis.
“The evolution of virtual image databases and analysis algorithms will transform cancer research, providing a mechanism for more sensitive cancer diagnostics, and driving discovery.”
In recent years the bottleneck for oncology has been the lack of new cancer drugs in the market. Current in vitro cancer models do not fully reflect the complexity and heterogeneity of a tumor in situ, producing the dilemma of poor predictability of drug efficacy. It is not uncommon, that a promising drug candidate fails in the clinical trials.
The PREDECT project, funded by the Innovative Medicines Initiatives (IMI) of the European Union, aimed to tackle this challenge of drug discovery. The main goal of the project was to compare the pathological and molecular profiles of novel in vitro platforms with those of human tumors.
“The PREDECT project assembled a pan-European archive of cancer models grown in different conditions, where image data is now being analyzed using WebMicroscope image analysis tools”, explains Dr. Emmy Verschuren, the Academic Coordinator of the consortium. Her research group in the Institute for Molecular Medicine Finland (FIMM) is collaborating with Pharma partners to develop complex models for target validation and gene therapy approaches.
WebMicroscope was a fundamental instrument in the PREDECT project by permitting researchers to store and survey stained sample slides as an archive of high quality whole slide images, to share image data with collaborators, and to select publication quality images.
“This image-based infrastructure is indispensable for our daily work. It has given us a new appreciation of the spatial complexity of tumor pathology”, Verschuren says.
She finds the novel WebMicroscope Deep Learning algorithms particularly promising: “They will enable a context-dependent quantitative readout of functional biomarkers across model systems, and within specific tissue segments.“
Technology transfer between competitors
One of the outcomes of the PREDECT project has been technology transfer between actors in the field of cancer research. The 6-year consortium (2011-2016) was an international partnership between more than 20 laboratories, including academic research groups as well as biotech and pharmaceutical companies.
“PREDECT project is a good representative of the new precompetitive model, where competitors join forces and collaborate. Also independent research institutes joined in to achieve the mutual goal”, says researcher Sami Blom from FIMM. His role was to coordinate and manage sample flow through the established PREDECT molecular pathology workflow, through which physical tumor samples are converted into digital virtual microscope images and stored in WebMicroscope.
During the project, pharmaceutical companies applied new cancer models introduced by collaborators. One example was an industrial-academic collaboration that created an ex vivo -model, where an in vivo tumor is removed, cut in slices, and cultivated on a plastic platform. This model makes it possible to study tumor complexity and heterogeneity in the disease´s native microenvironment.
“This kind of technology transfer benefits all parties and helps to achieve our common targets”, Blom adds.
“Web-based sharing of pathology expertise with underserved populations will benefit global public health.”
Boosting the data sharing
Within PREDECT project, thousands of samples representing different types of cancer were collected from pharmaceutical companies, small and mid-sized biotech companies, and research laboratories from several universities. The samples were sent to FIMM premises in Helsinki, where samples were turned into tissue microarrays and stained, most commonly by using immunohistochemistry. Next, stained tissues were digitized and integrated into the WebMicroscope platform, which enables the possibility to share and analyze the image data anywhere in the world.
“WebMicroscope played a central role in the project as it provided an excellent tool not only for data storing but also for analyzing and sharing the data between PREDECT parties. The benefit of the technology is that it is not bound to a specific location or software installation. Another advantage of the system is the efficient handling of large-scale images that enables smooth user-experience in terms of image viewing”, Blom reminds.
He tells that the response from the researchers has been very positive because of the ingeniously easy access to and sharing of the image data, regardless the location. Researchers simply send their samples and receive images in a ready-to-use form. As the whole system is centralized and well maintained, the process of the digitalization and the generation of the virtual microscope images has been smooth.
“Thanks to web-based microscopy, the producing and sharing of the image data was probably the least challenge of the whole PREDECT project. The increasing significance of data sharing in cancer research in general has truly been acknowledged, and platforms, such as WebMicroscope, are needed to support this trend”, Blom points out.