OwkinConnect

Unlock AI, break data silos,
and protect privacy with our Federated Learning software

Owkin Connect provides the infrastructure and AI technology that open the possibility for an unprecedented breadth of collaboration in healthcare while protecting patient privacy and data governance. Our distributed architecture and federated learning software allow data scientists to securely connect to decentralised, multi-party datasets and train AI models without having to pool data. Data remains local, only algorithms travel.

Get in touch

Aggregate models & insights, not data:

Owkin Connect technology is centered on three core principles — collaboration, confidentiality, and compliance. Life science companies deploy Owkin Connect in a wide variety of settings for clinical research, drug discovery and development.

Build research
networks

Create AI consortia where you can implement and manage data partnerships with your network of medical research partners.

Expand access to connected partner data

By eliminating the requirement to share data, partners can contribute to training AI without giving up competitive assets or compromising patient privacy.

Meet global compliance requirements

With our federated learning capabilities, life science companies can more easily leverage the insights across their international data assets meeting GDPR, HIPAA, and other regulatory requirements.

In consideration of the GDPR application, federated learning is a good tool to protect personal data and help in data privacy compliance. In a federated learning setting, each site keeps control of its data, which isn’t centralized, preserving data confidentiality and safety. As data isn’t merged, the security responsibility is handled by each site, and in case of a security issue, it will be easier to control the impact on data.

Owkin Connect
Federated Network

Owkin Connect is used in collaborative and competitive settings, when protecting confidential, proprietary or sensitive data and insights extracted from the data through the models are essential.

How Owkin Connect works

Owkin Connect is powered by a powerful python library which makes it possible to completely operate the platform from your python environment. It also comes with extensive documentation.

Federated Learning Task

Owkin Connect’s home page will display the federated learning channels the user has access to. Each channel consists of one or more datasets which can be used for training a federated model in order to provide specific results.

Machine Learning Feature

Registering a new dataset can easily be done through the graphical user interface by a medical researcher, or via the Python library by IT or data science specialists.

Machine Learning Feature

Once a new dataset has been uploaded, Owkin Connect provides cohort metrics and an easy way to visualise all data within the cohort before launching any models.

Federated Learning Task

Users can track the model training development (called the Compute Plan), step by step, with key metrics at each step of the way.

Federated Learning Task

Users can also visualise the algorithm code directly in the user interface for each federated model trained. This makes all computations reproducible.

Federated Learning Task

Once the federated run is complete, users can exame the overall performance of the federated models throughout the evaluation steps.

Machine Learning Feature

Users can also return to the cohort to examine the results in more detail, slide by slide for example, to examine which tiles were most predictive of certain features.

Machine Learning Feature

Visualize predictions in the slide viewer, patient by patient, along associated clinical information and annotations.

Machine Learning Feature

Heatmaps are overlaid on slides to assist the user in finding regions of high concordance with the predicted features. Whole slide images can easily be zoomed into with linear and diverging heatmaps turned on/off at the click of a button

Product Guide

Case Studies

Owkin Connect is currently deployed and operating multiple federated networks
around the world, including:

Unparalleled collaboration between pharma competitors in Drug Discovery

Now more than ever innovation and optimization are crucial to increase the pharmaceutical industry’s productivity and sustainability. The MELLODDY Project, a three-year, multi-million euro and 17 partner large research study funded by the IMI 2 Joint Undertaking, has set out to understand how competitive parties in collaboration can benefit from joint machine learning efforts at scale on pre-clinical datasets.

The endeavor has the potential to accelerate drug discovery, without compromising proprietary information or data privacy. Enabling Federated Learning for the MELLODDY project, Owkin Connect helped to secure the first Federated Learning model for drug discovery at scale. Owkin’s proprietary structure has been audited and has been deployed at 10 pharmaceutical companies.

This project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 831472. This Joint Undertaking receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA.

Learn more

Breast Cancer Treatment Evaluation with Owkin Connect

Alongside clinical, research, and technology partners we established that machine learning models can be trained successfully on histology images, siloed at different clinical centers, to predict treatment responses in breast cancer. The model trained with Owkin Connect can help oncologists to choose the most effective breast cancer treatment for each patient based on a single biopsy. Learn more about the HealthChain project here.
Image caption: In January 2020, Connect enabled the first-ever federated deep learning model trained on distributed histology images stored behind hospital firewalls.

Learn more