1
Supplementary Information
Multi-omic single-cell snapshots reveal mu
ltiple independent trajectories to drug
tolerance in a melanoma cell line
Su
et al
2
Supplementary Figure 1. Transcriptomic analysis
guides the selectio
n of panel markers.
a.
Pathways that are differentially altered from
day 0 to day 3 after BRA
Fi treatment. Each row
represents a certain signaling pathway and each ba
r indicates normalized enrichment score (NES)
calculated from geneset enrichment analysis (GSEA)
of cells harvested at day 3 versus day 0. Each
pathway is color-coded by its functional category as described in Fig. S1b.
b.
Panel of markers per
pathway selected to quantify w
ith single-cell barcode chip (S
CBC) analysis. 20 markers were
selected for SCBC analysis. Markers with simila
r biological function are
organized together and
color-coded by functional category.
3
Supplementary Figure 2. Distribution of a
ll 20 markers across 4 time points.
Each of the 20
plots represents the distributions of a certain marker
level across 4 time points. Y-axis represents natural
log of measured marker level. Markers within th
e same functional category are boxed together. Border
color of each plot corresponds to the functional
category of each marker, as described in Fig. 1a.
4
Supplementary Figure 3. Quantitation of varian
ce of three different
markers during BRAFi
treatment.
Y-axis represent the variance of the marke
r-level distribution. Data are presented as
mean values +/- SEM. Plots for flow cytometry
experiments is the result
of n = 3 biologically
independent samples per group. For SCBC datase
t, N=100 independent bootstrap runs are
performed. Source data are prov
ided as a Source Data file.
5
Supplementary Figure 4. SA ß-galactosidase ass
ay of untreated control and BRAFi-treated
cells.
Left panel: Significant increased percentage
of SA ß-galactosidase positive cells was
observed in BRAFi treated day5 cell
s. Data are presented as mean values +/- SEM. Right panel:
representative image of untreated melanoma cell
and cells treated with BRAFi for 5 days. Scale
bar 50um.
Each experiment is the result of n =
5 biologically independent samples per group.
Source data are provided as a Source Data file.
6
Supplementary Figure 5. Percentage of apoptosis
cells across five days of BRAFi treatment.
Data are presented as mean values +/- SD. Each
experiment is the resu
lt of n = 3 biologically
independent samples per group. Source data
are provided as a Source Data file.
7
Supplementary Figure 6. Comparison of Fl
ow cytometry analysis and SCBC data.
X-axis
represent different time points after drug treatment. Y-axis of the flow cytometry data represent
the log of measured fluorescent intensity. Y-axis of
the SCBC data represent measured level of the
respective markers. Error bars represent 95% c
onfidence interval of the data and the center
represents mean value. For SCBC, N= 156,
185, 162 and 171 single cells are independently
analyzed for day0, day1
day3 and day5 respectively.
Source data are provided as a Source Data
file.
8
Supplementary Figure 7. Visualizatio
n of SCBC data by FLOW-MAP.
Each dot represents an
individual cell. The distance between each pair
of cells represents the overall multi-omic
dissimilarity between
them. Cell pairs that ar
e close enough are linked with
an edge in between.
The colors of the dots in the central panel repr
esent BRAFi exposure time
(0, 1, 3, or 5 days) of
the corresponding cells. Dot colors in the other pa
nels represent the abunda
nce of each marker in
each cell. Markers belonging to
the same functional category, as
described in the bottom of the
figure, were assigned to a certain shape and color.
The dashed-line box in the panels for MITF,
MART1, and Ki67 levels shows a small subpopulation
of day-0 cells that are slow cycling with
less melanocytic phenotype.
9
.
Supplementary Figure 8. Visualization of marker abundance by t-SNE.
Each dot per plot
represents an individual cell. The distance betw
een each pair of dots repr
esents the overall multi-
omic dissimilarity between that pair of cells. The
dot colors in the central
panel represen
t the drug
exposure time of each cell. Dot colors in the other panels represent the abundance of the specified
marker in each cell. Markers that
belong to the same f
unctional category were assigned to a certain
shape and color, as described in the bottom of the figure.
T-SNE visualizations show both the
heterogeneity that exists at ba
seline as well as the progression
across time through two separate
paths.
10
Supplementary Figure 9. Visualizatio
n of marker abundance by PHATE.
Each dot per plot
represents an individual cell. The distance betw
een each pair of dots repr
esents the overall multi-
omic dissimilarity between that pair of cells. The dot colors in the central panel represent the drug
exposure time of each cell. Dot colors in the other panels represent the abundance of the specified
marker in each cell. Markers that
belong to the same functional ca
tegory were assigned to a certain
shape and color, as described in the bottom of
the figure. PHATE visualizations show both the
heterogeneity that exists at baseline as well as
the progression across
time through two separate
paths.
11
Supplementary Figure 10. Two modules from surprisal analysis recapitulates the
experimental data.
Each plot represents an individual ma
rker. Each dot within a single plot
represents a single cell. The x-axis value of ea
ch dot represents the experimentally measured
marker expression within a cell. The y-axis value
of each dot represents the predicted marker level
of the same cell as calculated by surprisal an
alysis of only module1
and module2. The strong
positive correlation between the x- and y-axis values indicate that surprisal analysis of the two
modules recapitulates experimental
ly measured marker levels pe
r cell. Linear regression with
Pearson correlation r and
two-tailed P value, n
=
674. No adjustments were made for multiple
comparisons.
12
Supplementary Figure 11. Two modules from
surprisal analysis recapitulates the
experimental data. a.
Schematic illustration of workflow
to project raw data and surprisal
analysis-predicted data onto the
same 2-dimensional space. Each ce
ll has measured levels of all
20 markers. Similarly, each cell also has predicted
levels of all 20 markers as calculated from
surprisal analysis. The raw and surprisal-predicted
data matrices were comb
ined to make a bigger
matrix with double the original number of rows,
each row representing a cell from raw data or
predicted data. Each column represents a single
marker, with each matrix value representing a
single cell’s abundance of a marker. The combine
d, 20-dimensional dataset was projected onto a
single t-SNE map where cells with similar levels of
all 20 markers will be in nearby coordinates.
b.
Each dot represents an individua
l cell. In the left panel, the x-
axis represents the t-SNE x-value
of the cell projected from raw data, while the y-axis represents the t-SNE x-value of the cell
projected from surprisal analysis-predicted data. The right panel is similar to left panel, but instead
compared t-SNE y-values. The linear, x = y plots i
ndicate that single cells, as projected from raw
data and from surprisal analysis-predicted data,
are in the same location in a reduced dimension;
therefore, the experimentally measured and surpri
sal analysis-predicted e
xpression profiles of all
20 markers are similar.