Collaborator stories: Adrian Tschan
Advancing image-based drug profiling with Fractal
Adrian Tschan is a computational biologist and data scientist working at Apricot BIO, a spin-off of the University of Zurich. Adrian is the main contributor and maintainer of the APx Fractal task collection which extends Fractal’s ability to process high content, multiplexed 2D image data.
Can you tell us about your scientific journey and the project you are currently working on?
I finished my PhD in Systems Biology at the end of 2023, during which I conducted extensive work with image-based assays, including immunofluorescence, FISH, and live-cell imaging. My research involved developing and utilizing various tools for image processing. Previously, our team relied on TissueMAPS, a platform developed in the lab of Lucas Pelkmans, but as it became outdated and difficult to modify, we transitioned to Fractal.
Since January 2024, I have been working at Apricot BIO, a spin-off of the University of Zurich. Our mission is to develop a diagnostic product based on multiplex immunofluorescence imaging, aimed at improving cancer treatment decisions. Specifically, we focus on molecular phenotyping of drug responses, where we analyze how single-cell biopsies from cancer patients respond to various anti-cancer drugs. By assessing multiplexed biomarker signaling states in response to drug treatments, we generate ranked recommendations for the most effective drugs for individual patients.
Our work is integrated into the OV Precision clinical trial, where we provide our drug ranking results to a molecular tumor board. The molecular tumor board integrates multiple data sources—including DNA and RNA sequencing, imaging mass cytometry, and ex-vivo drug response profiling—to determine the best course of treatment for each patient. This approach enhances personalized medicine, ensuring that patients receive the most effective therapies based on their unique molecular cancer context.
How has using Fractal improved or simplified your image analysis workflow?
Fractal has significantly streamlined our image processing and analysis workflows. Previously, manual data processing was time-consuming, requiring considerable computational expertise. With Fractal, we have automated our entire workflow - from image acquisition to data processing and report generation. This automation allows non-technical team members to run image processing pipelines efficiently, reducing the need for specialized computational knowledge at every step.
Additionally, Fractal's modular structure makes it easy to integrate our own image processing tasks into its framework. This flexibility is crucial for us, as we often need to customize processing steps to fit our specific experimental setups. By leveraging Fractal, we have greatly increased the efficiency of our data analysis and made it both reproducible and scalable for future clinical applications.
In addition, which features of Fractal were most helpful to you?
Several key features of Fractal have been particularly beneficial to our work. One of the most important is its support for OME-Zarr, a standardized image file format that ensures compatibility across different imaging devices. This simplifies data handling and facilitates collaboration with other research teams.
Another major advantage is Fractal’s open-source nature, which allows us to contribute our own custom tasks and benefit from those developed by the broader community. The ability to integrate our specific Python-based tasks seamlessly into the Fractal ecosystem has been invaluable as it allows us to precisely tailor our analysis workflows to our needs.
Importantly, Fractal’s automation capabilities extend beyond image processing. Besides the open-source tasks available in the APx Fractal Task collection, we have also developed proprietary tasks enabling automated report generation. These reports contain detailed information on patient specimens, drug responses, and biomarker measurements, all compiled in a structured format for easy interpretation by the molecular tumor board. This automation has saved us countless hours of manual work and improved the consistency of our data output.
What would you tell other researchers or labs about your experience with Fractal?
I highly recommend Fractal to any researcher working with high-content imaging data. It is a powerful and flexible platform to automate image analysis workflows. The ability to integrate custom tasks makes it adaptable to a wide range of applications, from basic research to clinical diagnostics.
Fractal’s open-source community is another major strength. The collaborative development environment means that users benefit from continuous improvements and shared expertise. For those new to the platform, resources such as the Fractal task template provide a straightforward entry point for creating custom tasks.
Overall, Fractal has shaped the way we handle image-based data at Apricot BIO, enabling us to conduct large-scale, reproducible analyses with minimal manual effort. It has been instrumental in advancing our work on molecular phenotyping of drug responses and will continue to play a crucial role in our journey toward developing clinically relevant companion diagnostics.
If you would like to reach out to Adrian to find more about his work, please contact him at atschan@apricotx.com