def _create_gpcca(*, backward: bool = False) -> Tuple[AnnData, GPCCA]:
adata = _adata_medium.copy()
sc.tl.paga(adata, groups="clusters")
vk = VelocityKernel(
adata, backward=backward).compute_transition_matrix(softmax_scale=4)
ck = ConnectivityKernel(adata,
backward=backward).compute_transition_matrix()
final_kernel = 0.8 * vk + 0.2 * ck
mc = GPCCA(final_kernel)
mc.compute_partition()
mc.compute_eigendecomposition()
mc.compute_schur(method="krylov")
mc.compute_macrostates(n_states=2)
mc.set_terminal_states_from_macrostates()
mc.compute_absorption_probabilities()
mc.compute_lineage_drivers(cluster_key="clusters", use_raw=False)
assert adata is mc.adata
if backward:
assert str(AbsProbKey.BACKWARD) in adata.obsm
else:
assert str(AbsProbKey.FORWARD) in adata.obsm
np.testing.assert_allclose(mc.absorption_probabilities.X.sum(1),
1.0,
rtol=1e-6)
return adata, mc
def lineage_drivers(
adata: AnnData,
backward: bool = False,
lineages: Optional[Union[Sequence, str]] = None,
method: str = TestMethod.FISCHER.s,
cluster_key: Optional[str] = None,
clusters: Optional[Union[Sequence, str]] = None,
layer: str = "X",
use_raw: bool = False,
confidence_level: float = 0.95,
n_perms: int = 1000,
seed: Optional[int] = None,
return_drivers: bool = True,
**kwargs,
) -> Optional[pd.DataFrame]: # noqa
"""
%(lineage_drivers.full_desc)s
Parameters
----------
%(adata)s
%(backward)s
%(lineage_drivers.parameters)s
Returns
-------
%(lineage_drivers.returns)s
References
----------
%(lineage_drivers.references)s
"""
# create dummy kernel and estimator
pk = DummyKernel(adata, backward=backward)
g = GPCCA(pk, read_from_adata=True, write_to_adata=False)
if g._get(P.ABS_PROBS) is None:
raise RuntimeError(
f"Compute absorption probabilities first as `cellrank.tl.lineages(..., backward={backward})`."
)
# call the underlying function to compute and store the lineage drivers
return g.compute_lineage_drivers(
method=method,
lineages=lineages,
cluster_key=cluster_key,
clusters=clusters,
layer=layer,
use_raw=use_raw,
confidence_level=confidence_level,
n_perms=n_perms,
seed=seed,
return_drivers=return_drivers,
**kwargs,
)
save='{}_schur.png'.format(key))
g.compute_metastable_states(n_states=1, cluster_key=clusters)
g.plot_metastable_states(show=False,
dpi=300,
save='{}_metastable.png'.format(key))
g.plot_metastable_states(discrete=True,
show=False,
dpi=300,
legend_fontsize=0,
save='{}_discrete_metastable.pdf'.format(key),
size=28)
g.set_final_states_from_metastable_states()
if key == "H508_EV" or key == "HT29_EV":
g.compute_absorption_probabilities()
g.compute_lineage_drivers()
samples[key].var.to_csv(key + "_lineages.tsv", sep='\t')
print(samples[key].var.columns)
if key == "HT29_EV":
g.plot_lineage_drivers(
'17', save='{}_17_lineage_drivers.png'.format(key))
g.plot_lineage_drivers(
'1', save='{}_10_lineage_drivers.png'.format(key))
g.plot_lineage_drivers('11',
save='{}_1_lineage_drivers.png'.format(key))
g.plot_lineage_drivers('4',
save='{}_4_lineage_drivers.png'.format(key))
if key == "H508_EV":
g.plot_lineage_drivers(
'10', save='{}_10_lineage_drivers.png'.format(key))