Supplementary
Information:
Algorithms
for
a
Commons
Cell
Atlas
A.
Sina
Booeshaghi
1*
,
Ángel
Galvez-Merchán
2*
&
Lior
Pachter
3,4+
1.
Department
of
Bioengineering,
University
of
California
Berkeley,
Berkeley,
CA,
USA
2.
Cellarity,
Somerville,
MA,
USA
3.
Division
of
Biology
and
Biological
Engineering,
California
Institute
of
Technology,
Pasadena,
CA,
USA
4.
Department
of
Computing
&
Mathematical
Sciences,
California
Institute
of
Technology,
Pasadena,
CA,
USA
*These
authors
contributed
equally
+To
whom
correspondence
should
be
addressed.
Supplementary
Figure
1:
CellAssign
and
mx
were
used
to
assign
cell-types
on
Splatter
simulated
data
across
different
DE
probability
values.
For
each
DE
probability,
we
performed
simulations
with
different
numbers
of
cells
(1000,
2000,
4000
and
8000)
and
different
number
of
cell
types
(2,
4,
6
and
8).
CellAssign
and
mx
were
run
using
1
to
15
marker
genes,
and
the
assignment
accuracy
was
measured.
Marker
genes
were
selected
using
the
same
strategy
as
in
CellAssign’s
original
publication
(Zhang
et
al.
2019)
,
genes
that
were
in
the
top
fifth
percentile
of
log-fold
change
and
the
top
tenth
percentile
of
mean
expression.
As
a
consequence
of
this
marker
gene
selection
method,
less
than
15
markers
could
be
selected
for
low
DE
probabilities.
Supplementary
Figure
2:
The
number
of
cells
assigned
to
each
cell
type
by
CellAssign
(first
column)
and
mx
(second
column).
The
expression
of
the
marker
genes
for
cell-type
3
was
set
to
0
sequentially
in
random
order,
and
the
number
of
cells
assigned
to
each
group
was
measured.
An
extra
cell-type
with
no
marker
genes
(
Other
)
was
included
in
the
assignment
with
the
goal
of
capturing
cells
that
couldn’t
be
assigned
to
any
of
the
defined
cell-types.
Supplementary
Figure
3:
The
same
approach
as
that
described
in
Supplementary
Figure
1
was
used,
but
15
marker
genes
were
chosen
for
all
DE
probabilities
by
selecting
genes
with
the
highest
log-fold
change
among
those
that
were
in
the
top
tenth
percentile
of
mean
expression.
Supplementary
Figure
4:
Summarizing
atlas
content.
Gene
counts
are
averaged
across
cells
within
individual
cell-by-gene
matrices
to
create
“organ”
level
summaries
of
count
matrices.
Averaged
counts
are
ranked
across
genes
and
then
subsequently
averaged
to
create
whole-atlas-level
summaries
of
count
matrices.
Zhang,
Allen
W.,
Ciara
O’Flanagan,
Elizabeth
A.
Chavez,
Jamie
L.
P.
Lim,
Nicholas
Ceglia,
Andrew
McPherson,
Matt
Wiens,
et
al.
2019.
“Probabilistic
Cell-Type
Assignment
of
Single-Cell
RNA-Seq
for
Tumor
Microenvironment
Profiling.”
Nature
Methods
16
(10):
1007–15.