PhD Visual Analytics for 3DOMICS

Updated: about 1 month ago
Deadline: 20 Jul 2019

There is an urgent need for fast and scalable visual analytics methods to capitalize on the discovery potential of multimodal, high-dimensional data. This position aims to develop methods to support the interactive exploration of such data for hypothesis generation which is the main analysis bottleneck. These methods developed will be inspired by the needs of single cell and spatially resolved imaging "-omics" (3DOMICS) data. Without further advances towards scalable, interpretable pattern discovery methods, much of the potential of large-scale single-cell / 3DOMICS experiments will remain buried in the data.

The rapid advancements in single-cell measurement technologies are currently revolutionizing biology, with clinical impact from immunology to cancer research to neuroscience. 3DOMICS techonologies allow the acquisition of tens to hundreds of genomics and/or proteomics measurements at subcellular resolution. For example, in Imaging Mass Cytometry a single pixel represents measurments of up to 50 different proteins over an area of 1μm squared. This allows for the extraction of cellular structures and their functionality and can show how cells are organized in their natural “tissue habitat”. Such data holds the key to unraveling diverse disease mechanisms, from interaction between immune and cancer cells to how the immune system derails in auto-immune diseases. However, the amount and complexity of the data mandate fast, scalable and interpretable data analytics methods. The aim of this project is to develop visual analytics techniques for integrated exploration of cellular contexts from multi-modal, large scale single cell- and 3DOMICS data.

We have recently demonstrated that hierarchical dimensionality reduction strategies such as Hierarchical Stochastic Neighbor Embedding (HSNE) represent a major step towards effective analysis of single-cell data at scale: enabling an interactive interpretation of data sets, while preserving the details of the data. We aim towards combining massive single-cell and spatially resolved data sets through this visual analytics concept. The sheer data size and complexity of a single experiment is becoming the most important bottleneck towards understanding the patterns in the data. Also, visualization and effective expert interaction with such large datasets is essential for hypothesis generation.

This position is part of a larger project focused in immunology and neuroscience. In collaboration with various research partners including Leiden University Medical Center (LUMC), and with the Allen Brain Institute in Seattle.

View or Apply

Similar Positions