PIVOT
THE PROJECT
Researchers at BC Cancer developed innovative techniques to sequence genomes of individual cells. They needed custom visualization tools to explore this wealth of data. As a Research Associate, I led a team from early concepts through to deployment of Montage, a web-application for single cell analysis.
THE DESIGN
DASHBOARDS for cancer genomics
The initial Montage design grew out of a need to analyze and integrate genomic measurements from cancer samples. Genomic mutations in a cancer can range from single nucleotide changes to large genomic deletions or duplications (copy number alterations) to rearrangements between chromosomes. Different computational methods detect these changes and their outputs need to be analyzed together. Standard dashboard tools like Tableau or Power BI did not support the necessary genomic charts nor were they performant with the millions of data points in these data sets.
Montage enabled scientists to create customized dashboards of diverse genomic measurements from cancer samples. Selections in one chart filter all other charts to only show alterations in the same affected genomic location. For example, by selecting a genomic deletion, a scientist can quickly identify if any point mutations are present in the remaining chromosome copy.
adapting to the era of single cell genomics
As technologies for sequencing the genomes of individual cells improved, Montage needed to support visualizing hundreds to thousands of cancer genomes. The final design used a heatmap to easily compare copy number alterations across many single cell genomes at once. A mini-map facilitated navigation across an even larger set of genomes and the raw underlying data for a given cell is available on hover.
SEARCH engine POWERED
To achieve millisecond response times for cross-filtering visualizations involving millions of data points, we turned to Elasticsearch. We carefully designed the indices to store all the necessary information regarding which point mutations and copy number alterations overlapped in genomic coordinates. Pre-computing and effectively storing this information resulted in a responsive front-end experience.
MY CONTRIBUTIONS
DEFINE
the problem
I envisioned the approach based on observing and discussing analytical needs with Sohrab Shah's lab.
WRANGLE
the data
Together with engineers in the lab, I designed the Elasticsearch indices to achieve a responsive front-end.
DESIGN
the prototypes
I led the design process from concept, to prototyping, to refining the deployed solution. I had the pleasure of leading a team of students and engineers who built the software and helped shape the final product.
COMMUNICATE
the methods
I wrote successful grant applications for the Canadian Cancer Society and Genome British Columbia which funded this work.