Meet MOSAIC: the world’s largest spatial and multiomics dataset in oncology

Joseph Lehár (SVP Business Strategy, Owkin), Fabrice André (Director of Research Division, Gustave Roussy) unveil the launch of MOSAIC, a study that aims at building the largest spatial and multiomics dataset in oncology.

‍

‍

What is MOSAIC and what is the goal of this project?

MOSAIC is an ambitious program to collect spatial omics data across thousands of patients in cancer. The goal is to go after seven cancer indications and for each one of them collect a thousand patients' worth of information. This is a very large number of patients. These spatial biology technologies are brand new - they've only just started being used in many centers and the largest studies that I've seen in the literature consist of a few tens of samples. So going to the level of a thousand per indication is a dramatic increase in scale.

The MOSAIC collaboration has several important goals. First, scientifically, we would like to break up complex cancer diagnoses into distinct subgroups with different biology enabling precision medicine. Secondly, we want to use MOSAIC as an opportunity to really understand how to make the most use of spatial omics data, setting standards, and trying out novel analytic algorithms to enable the dissection of biological information from spatial omic data.

Joseph Lehár, SVP Business Strategy, Owkin

‍

What is the rationale to build a large-scale spatial omics dataset?

We now have large portfolios of drugs - we must define which patient is most likely to benefit from which drug and beyond that, we need to define the best theoretical drug for each patient. For this, we need to characterize the biology in populations of patients, but also in each patient - and this is what we call precision medicine.

What has been reported is that the way cancer cells and the immune cells are organized could define the sensitivity of patients to specific drugs. And this is the rationale to build a large-scale spatial biology dataset.

Currently we can only say that the spatial distribution of cells in the tumor microenvironment has some impact but we can’t really establish its clinical utility. If we want to reach the vision of using spatial biology to predict patient outcomes and to develop new drugs we need to build a dataset with data from several thousands of patients, several thousands of samples, and several thousands of proteins and genes.

Fabrice André, Director of Research Division, Gustave Roussy

‍

What are the challenges of building such a large-scale spatial omics dataset?

What will be both exciting and challenging about the MOSAIC data set is multimodal and spatially resolved. What that means is that we're looking at an unprecedented level of complexity of information about every single patient.

Data like that are extremely hard to analyze and the methods that people have come up with to deal with complex data of that kind right now primarily rely on deep learning.

Owkin is very well positioned to lead the MOSAIC effort firstly because we have a lot of experience working with deep learning models of that kind, secondly because we've been looking at histogenomic information for for many years trying to understand how image features relate to molecular profiling features and how that relates to cancer biology.

Joseph Lehár, SVP Business Strategy, Owkin

‍

Why has a project like MOSAIC not been done before, and why is now the right moment to launch it?

When I first heard about MOSAIC, my reaction was: “This is exactly what we wanted to do, but we are not able to do.” We strongly believe that the topic of spatial biology is so complex that the only way to move forward is through a consortium of multiple centers and scaling up the ambition of the project.

At Gustave Roussy, we are very happy to participate to MOSAIC; the aim of the project is to improve the outcome of the patient and this is the mission of Gustave Roussy.

Fabrice André, Director of Research Division, Gustave Roussy

‍

First of all, the spatial omics platforms are now at a stage where you can scalably generate a lot of data from a lot of patients. Secondly, immuno-oncology has gotten to the point where we can really try to understand how the immune system interacts properly with cancers and can be used to control the disease. We are excited to see other institutions join the collaboration and would love to begin working with our academic and industrial partners to make this an impactful effort.

Joseph Lehár, SVP Business Strategy, Owkin

‍

‍

_{Adapted from transcription}