Mass cytometry at the Vaccine & Immunotherapy Center
Massachusetts General Hospital, since July 2026
- Current mass cytometry clustering algorithms require manual inputs, the goal of the project was to explore ways to minimize manual inputs. This would open up the use of these mass cytometry tools to more people and help standardize how the tools are used.
- We built a Python pipeline that identifies immune cell populations across vaccine stimulations to use as a tool to investigate how to augment these algorithms to make them more user friendly. It handles the whole preprocessing path in one pass: FCS parsing, arcsinh transformation, and dimensionality reduction with PCA and UMAP.
- We are looking to add biological validation metrics to the existing clustering, so the population assignments are not just based in math, but also in real biology.
- Now extending it to new datasets (Flu & ALS), using X-shift to track changes in population abundance and adapting the Shannon and Simpson diversity notebooks I wrote for the first project.
Working with
Richard Dzeng
Ethan Rabinowitz
Scale
Fifty-plus parameters per cell, past a million events per dataset
Method
Python, NumPy, scikit-learn, FlowSOM, OMIQ, X-shift, UMAP and PCA