Events / Area of Influence (AoI)
Hierarchy Construction (high-dimensional space)
Neighborhood Graph Embeddings and Clustering
Hierarchy Exploration (two-dimensional space)
Inte
rmed
iate
Lev
els
(arb
itrar
y nu
mbe
r)O
verv
iew
Lev
elDa
ta L
evel
Construction ExplorationConstructionConstruction
1
2
embedding
color: marker a
density
heatmap
color: marker c
color: marker b
2 3
1
1
2
3
4
HSNE 1
HSN
E 2
CellLandmarkAoI
1
2
2 3
1
1 2 3
abc
abc
1 2
1
2
34
1 2 3 4
abc
AoI
Dens
ity
Expr
essio
n
Interactive Visual Analysis of Mass Cytometry Data by Hierarchical Stochastic Neighbor Embedding
T Höllt1,5, V van Unen3, N Pezzotti1, N Li3, M Reinders2, E Eisemann1, F Koning3, A Vilanova1, B Lelieveldt2,4
1Computer Graphics and Visualization Group, 2Pattern Recognition and Bioinformatics Group, TU Delft,3Dept. of Immunohematology and Blood Transfusion, 4Div. of Image Processing, Department of Radiology,
5Computational Biology Center, LUMC,
marker b
marker a
mar
ker c
HSNE(2 levels)
HSNE 1 HSNE 1
HSN
E 2
HSN
E 2
AoI(# Events)
Overview Level Data Level
Concept: Non-Linear Hierarchy (Abstract to Single-Cell)
Hierarchy Construction and Exploration
Results● State-of-the-art precision in a fraction of the time
Advantages
References[1] van Unen, et al. Mass Cytometry of the Human Mucosal Immune System Identifies Tissue- and Disease-Associated Immune Subsets. Immunity 44, 1227–1239 (2016).[2] Pezzotti et al. Hierarchical Stochastic Neighbor Embedding. Comput. Graph. Forum 35, 21–30 (2016).[3] Höllt, et al. Cytosplore: Interactive Immune Cell Phenotyping for Large Single-Cell Datasets. Comput. Graph. Forum 35, 171–180 (2016).[4] Samusik et al. Automated mapping of phenotype space with single-cell data. Nat. Methods 13, 493–496 (2016).
AcknowledgmentsThe research leading to these results has received funding from Leiden University Medical Center, the Netherlands Organization for Scientific Research (ZonMW grant 91112008) and the NWO Applied and Engineering Sciences grants 12720 (VAnPIRe) and 17721 (Genes in Space).
Contact: [email protected] http://www.cytosplore.org
● Millions(!) of cells● Non-linear distances, based on single cell resolution throughout the hierarchy● Full data access, zoom into single-cell resolution● Ineractive, integrated software Cytosplore[3]
IntroductionImmunesystem-wide CyTOF studies[1] produce millions of cells, which most computational tools cannot handle with-out downsampling, potentially leading to data loss.HSNE[2] builds a hierarchy of non-linear similarities, allowing exploration of large scale datasets at full resolution interac-tively and, through that, efficient discor-very of rare cell populations.
HSNE 1
HS
NE
2
CD8+T cells
CD4+T cells
NKT & γδ T cells
Eosinophils
pDCCD4+
CD4-
IgD+IgM+
B cells
Plasma cells
NK cells
IgD-IgM+
B cells
IgD-IgM-
B cells
Macrophages
mDCs
Basophils
Nonclassicalmonocytes
Intermediatemonocytes
Classicalmonocytes
GMP
MPP
MPP
CLP CMP
Branching point between classicaland nonclassical pathways
HSNE - 5 minutes Vortex[4] - 22 hours