def plot_scores(self, motifs, name=True, max_len=50):
"""Create motif scores boxplot of different clusters.
Motifs can be specified as either motif or factor names.
The motif scores will be scaled and plotted as z-scores.
Parameters
----------
motifs : iterable or str
List of motif or factor names.
name : bool, optional
Use factor names instead of motif names for plotting.
max_len : int, optional
Truncate the list of factors to this maximum length.
Returns
-------
g : FacetGrid
Returns the seaborn FacetGrid object with the plot.
"""
if self.input.shape[1] != 1:
raise ValueError(
"Can't make a categorical plot with real-valued data")
if type("") == type(motifs):
motifs = [motifs]
plot_motifs = []
for motif in motifs:
if motif in self.motifs:
plot_motifs.append(motif)
else:
for m in self.motifs.values():
if motif in m.factors:
plot_motifs.append(m.id)
data = self.scores[plot_motifs]
data[:] = data.scale(data, axix=0)
if name:
data = data.T
data["factors"] = [
join_max(self.motifs[n].factors, max_len, ",", suffix=",(...)")
for n in plot_motifs
]
data = data.set_index("factors").T
data = pd.melt(self.input.join(data), id_vars=["cluster"])
data.columns = ["cluster", "motif", "z-score"]
g = sns.factorplot(data=data,
y="motif",
x="z-score",
hue="cluster",
kind="box",
aspect=2)
return g
def plot_scores(self, motifs, name=True, max_len=50):
"""Create motif scores boxplot of different clusters.
Motifs can be specified as either motif or factor names.
The motif scores will be scaled and plotted as z-scores.
Parameters
----------
motifs : iterable or str
List of motif or factor names.
name : bool, optional
Use factor names instead of motif names for plotting.
max_len : int, optional
Truncate the list of factors to this maximum length.
Returns
-------
g : FacetGrid
Returns the seaborn FacetGrid object with the plot.
"""
if self.input.shape[1] != 1:
raise ValueError("Can't make a categorical plot with real-valued data")
if type("") == type(motifs):
motifs = [motifs]
plot_motifs = []
for motif in motifs:
if motif in self.motifs:
plot_motifs.append(motif)
else:
for m in self.motifs.values():
if motif in m.factors:
plot_motifs.append(m.id)
data = self.scores[plot_motifs]
data[:] = data.scale(data, axix=0)
if name:
data = data.T
data["factors"] = [join_max(self.motifs[n].factors, max_len, ",", suffix=",(...)") for n in plot_motifs]
data = data.set_index("factors").T
data = pd.melt(self.input.join(data), id_vars=["cluster"])
data.columns = ["cluster", "motif", "z-score"]
g = sns.factorplot(data=data, y="motif", x="z-score", hue="cluster", kind="box", aspect=2)
return g
def plot_heatmap(self,
kind="final",
min_freq=0.01,
threshold=2,
name=True,
indirect=False,
figsize=None,
max_len=50,
aspect=1,
**kwargs):
"""Plot clustered heatmap of predicted motif activity.
Parameters
----------
kind : str, optional
Which data type to use for plotting. Default is 'final', which will
plot the result of the rang aggregation. Other options are 'freq'
for the motif frequencies, or any of the individual activities such
as 'rf.score'.
min_freq : float, optional
Minimum frequency of motif occurrence.
threshold : float, optional
Minimum activity (absolute) of the rank aggregation result.
name : bool, optional
Use factor names instead of motif names for plotting.
indirect : bool, optional
Include indirect factors. Default is False.
max_len : int, optional
Truncate the list of factors to this maximum length.
figsize : tuple, optional
Tuple of figure size (width, height).
aspect : int, optional
Aspect ratio for tweaking the plot.
kwargs : other keyword arguments
All other keyword arguments are passed to sns.clustermap
Returns
-------
cg : ClusterGrid
A seaborn ClusterGrid instance.
"""
filt = np.any(np.abs(self.result) >= threshold, 1)
if hasattr(self, "freq"):
filt = filt & np.any(np.abs(self.freq.T) >= min_freq, 1)
else:
filt = filt & (self.counts.sum() / self.counts.shape[0] > min_freq)
idx = self.result.loc[filt].index
if idx.shape[0] >= 100:
logger.warning("The filtered matrix has more than 100 rows.")
logger.warning(
"It might be worthwhile to increase the threshold for visualization"
)
cmap = "RdBu_r"
if kind == "final":
data = self.result
elif kind == "freq":
if hasattr(self, "freq"):
data = self.freq.T
cmap = "Reds"
else:
raise ValueError(
"frequency plot only works with maelstrom output from clusters"
)
elif kind in self.activity:
data = self.activity[kind]
if kind in ["hypergeom.count", "mwu.score"]:
cmap = "Reds"
else:
raise ValueError("Unknown dtype")
# print(data.head())
# plt.figure(
m = data.loc[idx]
vmax = max(abs(np.percentile(m, 1)), np.percentile(m, 99))
vmin = -vmax
if name:
m["factors"] = [
join_max(
_get_factor_list(self.motifs[n], indirect),
max_len,
",",
suffix=",(...)",
) for n in m.index
]
m = m.set_index("factors")
h, w = m.shape
if figsize is None:
figsize = (3 + m.shape[1] / 4, 1 + m.shape[0] / 3)
fig = plt.figure(figsize=figsize)
npixels = 30
g = GridSpec(2,
1,
height_ratios=(fig.get_figheight() * fig.dpi - npixels,
npixels))
ax1 = fig.add_subplot(g[0, :])
ax2 = fig.add_subplot(g[1, :])
ax2.set_title("Significance (-log10(p-value))")
dm = pdist(m, metric="euclidean")
hc = linkage(dm, method="ward")
leaves = dendrogram(hc, no_plot=True)["leaves"]
cg = sns.heatmap(
m.iloc[leaves],
ax=ax1,
cbar_ax=ax2,
cbar_kws={"orientation": "horizontal"},
cmap=cmap,
linewidths=1,
vmin=vmin,
vmax=vmax,
)
plt.tight_layout()
# cg.ax_col_dendrogram.set_visible(False)
# plt.setp(cg.ax_heatmap.yaxis.get_majorticklabels(), rotation=0)
return cg
def plot_heatmap(self, kind="final", min_freq=0.01, threshold=2, name=True, max_len=50, aspect=1, **kwargs):
"""Plot clustered heatmap of predicted motif activity.
Parameters
----------
kind : str, optional
Which data type to use for plotting. Default is 'final', which will plot the
result of the rang aggregation. Other options are 'freq' for the motif frequencies,
or any of the individual activities such as 'rf.score'.
min_freq : float, optional
Minimum frequency of motif occurrence.
threshold : float, optional
Minimum activity (absolute) of the rank aggregation result.
name : bool, optional
Use factor names instead of motif names for plotting.
max_len : int, optional
Truncate the list of factors to this maximum length.
aspect : int, optional
Aspect ratio for tweaking the plot.
kwargs : other keyword arguments
All other keyword arguments are passed to sns.clustermap
Returns
-------
cg : ClusterGrid
A seaborn ClusterGrid instance.
"""
filt = np.any(np.abs(self.result) >= threshold, 1) & np.any(np.abs(self.freq.T) >= min_freq, 1)
idx = self.result[filt].index
cmap = "RdBu_r"
if kind == "final":
data = self.result
elif kind == "freq":
data = self.freq.T
cmap = "Reds"
elif kind in self.activity:
data = self.activity[dtype]
if kind in ["hypergeom.count", "mwu.score"]:
cmap = "Reds"
else:
raise ValueError("Unknown dtype")
#print(data.head())
#plt.figure(
m = data.loc[idx]
if name:
m["factors"] = [join_max(self.motifs[n].factors, max_len, ",", suffix=",(...)") for n in m.index]
m = m.set_index("factors")
h,w = m.shape
cg = sns.clustermap(m, cmap=cmap, col_cluster=False,
figsize=(2 + w * 0.5 * aspect, 0.5 * h), linewidths=1,
**kwargs)
cg.ax_col_dendrogram.set_visible(False)
plt.setp(cg.ax_heatmap.yaxis.get_majorticklabels(), rotation=0);
return cg
def plot_heatmap(self,
kind="final",
min_freq=0.01,
threshold=2,
name=True,
max_len=50,
aspect=1,
**kwargs):
"""Plot clustered heatmap of predicted motif activity.
Parameters
----------
kind : str, optional
Which data type to use for plotting. Default is 'final', which will plot the
result of the rang aggregation. Other options are 'freq' for the motif frequencies,
or any of the individual activities such as 'rf.score'.
min_freq : float, optional
Minimum frequency of motif occurrence.
threshold : float, optional
Minimum activity (absolute) of the rank aggregation result.
name : bool, optional
Use factor names instead of motif names for plotting.
max_len : int, optional
Truncate the list of factors to this maximum length.
aspect : int, optional
Aspect ratio for tweaking the plot.
kwargs : other keyword arguments
All other keyword arguments are passed to sns.clustermap
Returns
-------
cg : ClusterGrid
A seaborn ClusterGrid instance.
"""
filt = np.any(np.abs(self.result) >= threshold, 1) & np.any(
np.abs(self.freq.T) >= min_freq, 1)
idx = self.result[filt].index
cmap = "RdBu_r"
if kind == "final":
data = self.result
elif kind == "freq":
data = self.freq.T
cmap = "Reds"
elif kind in self.activity:
data = self.activity[dtype]
if kind in ["hypergeom.count", "mwu.score"]:
cmap = "Reds"
else:
raise ValueError("Unknown dtype")
#print(data.head())
#plt.figure(
m = data.loc[idx]
if name:
m["factors"] = [
join_max(self.motifs[n].factors, max_len, ",", suffix=",(...)")
for n in m.index
]
m = m.set_index("factors")
h, w = m.shape
cg = sns.clustermap(m,
cmap=cmap,
col_cluster=False,
figsize=(2 + w * 0.5 * aspect, 0.5 * h),
linewidths=1,
**kwargs)
cg.ax_col_dendrogram.set_visible(False)
plt.setp(cg.ax_heatmap.yaxis.get_majorticklabels(), rotation=0)
return cg