Healthy and B-cell precursor Acute Lymphoblastic Leukemia (ALL) cells analyzed via CyTOF

This is a dataset of 1,108,853 blood and bone marrow cells collected from 3 pediatric B-cell Precursor Acute Lymphoblastic Leukemia (BCP-ALL) patients and 3 healthy controls. Each ALL sample is made up of a mixture of cancer cells and healthy cells, whereas the healthy samples do not contain and cancer cells. This dataset can be used to evaluate models trained to classify cells as either cancerous or non-cancerous. 

Each BCP-ALL patient has samples collected from 3 timepoints and 2 tissues: diagnosis (bone marrow and blood), day 8 post-treatment initiation with chemotherapy (blood), and day 15 post-treatment initiation (blood). Healthy patients only have samples collected from a single timepoint (the time of donation) and one tissue (bone marrow). These different tissues and timepoints can be used to assess a classifier's ability to generalize to new contexts (i.e. from bone marrow to blood, or from the diagnostic timepoint to a timepoint later in treatment). 

Methods
All cells have been analyzed for the presence of 28 proteins as previously described using mass cytometry (CyTOF), a high-dimensional cytometry platform similar to multicolor flow cytometers commonly used to analyze leukemic tissue specimens in clinical laboratories. CyTOF analysis allows a high-dimensional sample characterization, extending the capabilities beyond those of conventional multicolor flow cytometers, typically employed for the analysis of leukemic tissue specimens in clinical laboratories.

The files are in the flow cytometry standard (.FCS) file format and include information about 28 proteins as read off the mass cytometer (unit: ion counts) and an additional column called ('cell_type') that encodes healthy cells with a value of 0 and cancerous cells as a value of 1. These labels were manually annotated by an expert BCP-ALL cytometrist and verified by a physician-scientist board-certified in pediatric hematology and oncology. The samples were extracted, debarcoded, and filtered for doublets and dead cells according to standard mass cytometry protocols. The ion counts for each protein have NOT been transformed in any way.