Exploring the latent space of integrated heterogeneous data sources in breast cancer

Principal Investigator

Name
David Longo

Degrees
A.L.M.

Institution
Emergent Dynamics

Position Title
Chief Executive Officer

Email
dave@emergent.bio

About this CDAS Project

Study
PLCO (Learn more about this study)

Project ID
PLCO-623

Initial CDAS Request Approval
May 12, 2020

Title
Exploring the latent space of integrated heterogeneous data sources in breast cancer

Summary
Emergent Dynamics intends to explore the latent space of breast cancer data run through variational autoencoders with deconfounding. We intend to elucidate the underlying causes and contributors to various subtypes of breast cancer. The work takes in expression data in the form of RNASeq, clinical data, and CNA and encodes those input sources down to a latent space that can be walked to evaluate deep connections between data points. Finally, the project aims to construct a state of the art data generator for modeling cancer.

Aims

- Encode breast cancer data into a walkable latent space
- Elucidate key contributors in cancer progression
- Establish state of the art data generation models to model cancer

Collaborators

David Longo, Emergent Dynamics
Bara Badwan, Emergent Dynamics
Chris Zoumadakis, Emergent Dynamics
Nancy Parmalee, Emergent Dynamics
John Lazar, Emergent Dynamics
Tom Murray, Emergent Dynamics
Aly Abdelkareem, Emergent Dynamics
Matthias Denecke, Emergent Dynamics
Andrew Brown, Emergent Dynamics