plt.title('distance regressed - spearman r = %1.3f, - p = %1.3f' %
corr_distreg)
figure.figure.set_figwidth(10)
figure.figure.set_figheight(10)
sns.despine(trim=True)
####################################
# tsn plot
####################################
# plot tsn with 7-network assignments
extrVal = np.max([abs(np.min(tsn)), abs(np.max(tsn))])
figure = plotting.plot_mod_heatmap(tsn,
rsnidx.flatten().astype(int),
figsize=(6.4, 4.8),
cmap='RdBu_r',
vmin=-extrVal,
vmax=extrVal,
xlabels=list(rsnlabelsabb),
ylabels=list(rsnlabelsabb),
rasterized=True)
# for 17 networks
lhlabels = ('../data/schaefer/HCP/fslr32k/gifti/' +
'Schaefer2018_400Parcels_17Networks_order_lh.label.gii')
rhlabels = ('../data/schaefer/HCP/fslr32k/gifti/' +
'Schaefer2018_400Parcels_17Networks_order_rh.label.gii')
labelinfo = np.loadtxt('../data/schaefer/HCP/fslr32k/gifti/' +
'Schaefer2018_400Parcels_17Networks_order_info.txt',
dtype='str',
delimiter='tab')
rsnlabels17 = []
nonegative = corr.copy()
nonegative[corr < 0] = 0
ci, Q = bct.community_louvain(nonegative, gamma=1.5)
num_ci = len(np.unique(ci))
print('{} clusters detected with a modularity of {:.2f}.'.format(num_ci, Q))
###############################################################################
# We'll take a peek at how the correlation matrix looks when sorted by these
# communities. We can use the :func:`~.plotting.plot_mod_heatmap` function,
# which is a wrapper around :func:`plt.imshow()`, to do this easily:
from netneurotools import plotting
plotting.plot_mod_heatmap(corr, ci, vmin=-1, vmax=1, cmap='viridis')
###############################################################################
# The Louvain algorithm is greedy so different instantiations will return
# different community assignments. We can run the algorithm ~100 times to see
# this discrepancy:
ci = [bct.community_louvain(nonegative, gamma=1.5)[0] for n in range(100)]
fig, ax = plt.subplots(1, 1, figsize=(6.4, 2))
ax.imshow(ci, cmap='Set1')
ax.set(ylabel='Assignments', xlabel='ROIs', xticklabels=[], yticklabels=[])
###############################################################################
# We'll provide these different assignments to our consensus-finding algorithm
# which will generate one final community assignment vector:
# Convert the bootstrap ratios into a node x node matrix of functional weights
# and plot them, sorting by community assignment. This will give us an idea of
# which communities / networks are contributing most.
from netneurotools.plotting import plot_mod_heatmap
bsr_mat = np.zeros((630, 630))
bsr_mat[np.tril_indices_from(bsr_mat, k=-1)] = bootstrap_ratios[:, 0]
bsr_mat = bsr_mat + bsr_mat.T
plot_mod_heatmap(bsr_mat,
comm_ids,
vmin=-4,
vmax=4,
ax=ax1,
cmap='RdBu_r',
cbar=False,
edgecolor='red',
xlabels=comm_labels,
xlabelrotation=45)
ax1.tick_params(top=True,
labeltop=True,
bottom=False,
labelbottom=False,
length=0)
ax1.set_xticklabels(ax1.get_xticklabels(), ha='left')
ax1.set(yticks=[], yticklabels=[])
cbar = fig.colorbar(ax1.collections[0],
ax=ax1,
orientation='horizontal',
fraction=0.1,
def gen_figure(data, demographics, agreement, consensus, assignments, zrand):
"""
Generates figure 3
Parameters
----------
data : list of pandas.DataFrame
demographics : pandas.DataFrame
agreement : (N, N) array_like
consensus : (N,) array_like
zrand : (C, K, M) array_like
Returns
-------
fig : matplotlib.figure.Figure
Plotted figure
"""
# make figure
fig = plt.figure(figsize=(16.5, 15))
gs = gridspec.GridSpec(3, 15, figure=fig)
gs.update(wspace=1.2, hspace=0.5)
ax1, ax2, ax3 = (plt.subplot(gs[0, n:(n + 5)]) for n in [0, 5, 10])
###########################################################################
# make hyperparameter similarity plot
coll = ax1.imshow(zrand[1],
cmap=defaults.similarity_cmap,
alpha=0.7,
aspect='auto',
vmin=50,
vmax=150,
rasterized=True)
# make axis a bit prettier
ax1.set(xticks=[], yticks=[], xticklabels=[], yticklabels=[])
ax1.set_title('parameter stability', pad=20, fontsize=20)
ax1.set_xlabel('scaling, μ', labelpad=7)
ax1.set_ylabel('neighbors, K', labelpad=7)
sns.despine(ax=ax1, left=True, bottom=True)
# add colorbar to axis 1
ax1_cbar = fig.colorbar(coll, ax=ax1, drawedges=False, ticks=[])
ax1_cbar.outline.set(linewidth=0)
ax1_cbar.ax.set_ylabel('local similarity', rotation=270, labelpad=25)
ax1_cbar.ax.tick_params(axis='both', which='both', length=0)
###########################################################################
# make cluster assignment map
cbar_kws = {'ticks': [], 'boundaries': np.arange(-0.5, 4.5)}
# re-order community assignments for plotting purposes
comms_plot, idxs = cluster.reorder_assignments(assignments,
consensus,
seed=SEED)
ax2 = sns.heatmap(comms_plot,
cmap=defaults.four_cluster_cmap,
xticklabels=[],
yticklabels=[],
ax=ax2,
cbar_kws=cbar_kws,
rasterized=True)
hlines = np.where(np.diff(consensus[consensus.argsort()]))[0] + 1
ax2.hlines(hlines, 0, comms_plot.shape[-1], color='white', linewidth=2)
ax2.set_title('patient clusters', pad=20, fontsize=20)
ax2.set_xlabel('cluster partitions', labelpad=7)
ax2.set_ylabel('patients', labelpad=7)
# modify colorbar
cbar = ax2.collections[0].colorbar
cbar.outline.set(linewidth=0)
cbar.ax.set_ylabel('cluster label', rotation=270, labelpad=25)
cbar.ax.tick_params(axis='both', which='both', length=0)
###########################################################################
# make sorted cluster heatmap
ax3 = plotting.plot_mod_heatmap(agreement,
consensus,
ax=ax3,
inds=idxs[0].squeeze(),
rasterized=True,
cmap='viridis',
edgecolor='white')
ax3.set(xticks=[], yticks=[], xticklabels=[], yticklabels=[])
ax3.set_title('patient co-assignment', pad=20, fontsize=20)
ax3.set_xlabel('patients', labelpad=7)
ax3.set_ylabel('patients', labelpad=7)
# modify colorbar
cbar = ax3.collections[0].colorbar
cbar.set_ticks([])
cbar.outline.set(linewidth=0)
cbar.ax.set_ylabel('probability', rotation=270, labelpad=25)
cbar.ax.tick_params(axis='both', which='both', length=0)
# add figure labels
for ax, text in zip([ax1, ax2, ax3], ['a', 'b', 'c']):
ax.text(-0.2, 1.1, text, transform=ax.transAxes, fontdict=fontd)
###########################################################################
# make rainplots
axes = [plt.subplot(gs[1, r:(r + 3)]) for r in range(0, 15, 3)]
titles = [
'cortical thickness', 'subcortical volume', 'dopamine binding',
'protein availability', 'clinical score'
]
xlabels = {
'supramarginal_15_r': 'supramarginal gyrus',
'substantia_nigra_pars_compacta': 'substantia nigra',
'caudate_l': 'left caudate',
'ttau': 'total tau',
'tremor': 'tremor'
}
for dat, ax, title in zip(data, axes, titles):
# z-score data for plotting
zdat = sstats.zscore(dat, ddof=1)
zdf = pd.DataFrame(zdat, index=dat.index, columns=dat.columns)
currdata = pd.merge(demographics[['cluster']], zdf, on='participant')
# find maximally discriminating feature in data
grps = (zdat[consensus == cl] for cl in np.unique(consensus))
idx = sstats.f_oneway(*grps).statistic.argmax()
xlabel = xlabels.get(zdf.columns[idx])
# make rainplot
utils.rainplot(x=dat.columns[idx],
y='cluster',
data=currdata,
viol_kws={'linewidth': 0},
ax=ax,
palette=defaults.three_cluster_cmap)
# make axis goodness
ax.set(xticklabels=[-2.5, 2.5],
yticklabels=[],
ylabel='',
xlabel=title,
xlim=(-3.5, 3.5),
xticks=[-2.5, 2.5])
ax.set_title(xlabel, pad=15)
sns.despine(ax=ax, left=True)
utils.shift_axis(ax, lshift=0, rshift=0.02)
axes[0].text(-0.3, 1.10, 'd', transform=axes[0].transAxes, fontdict=fontd)
###########################################################################
# make lineplot
behavior = load_longitudinal_behavior(demographics.index)
axes_lp = [plt.subplot(gs[2, r:(r + 4)]) for r in [4, 9]]
for n, (lp, ax) in enumerate(zip(['pigd', 'tremor'], axes_lp)):
# get data for given test
pd_test = pd.merge(clean_visits(behavior, lp),
demographics,
on='participant')
# normalize data based on first visit mean/stdev
t = pd_test.query('visit == 0')[lp]
pd_test.loc[:, lp] -= t.mean()
pd_test.loc[:, lp] /= t.std(ddof=1)
# plot it
sns.lineplot(x='visit',
y=lp,
data=pd_test.dropna(subset=[lp]),
hue='cluster',
palette=defaults.three_cluster_cmap,
legend=False,
ci=68,
ax=ax)
ax.set(xticklabels=[], xlabel='time')
if n == 0:
ax.set_ylabel('clinical severity', labelpad=10)
ax.text(-0.3, 1.10, 'e', transform=ax.transAxes, fontdict=fontd)
ax.set(ylim=[-0.5, 2.5],
yticks=[-0.5, 0.5, 1.5, 2.5],
yticklabels=[-0.5, 0.5, 1.5, 2.5])
else:
ax.set(ylim=[-0.75, 0.75],
ylabel='',
yticks=[-0.75, 0, 0.75],
yticklabels=[-0.75, 0, 0.75])
ax.set_title(lp, pad=15)
sns.despine(ax=ax)
###########################################################################
# shift axes to make room for everything
utils.shift_axis(ax1, lshift=0.04)
utils.shift_axis(ax1_cbar.ax, lshift=0.04)
utils.shift_axis(ax2, lshift=0.06)
utils.shift_axis(ax2.collections[0].colorbar.ax, lshift=0.06)
utils.shift_axis(ax3, lshift=0.08)
utils.shift_axis(ax3.collections[0].colorbar.ax, lshift=0.08)
for a, lshift in zip(axes[1:], np.cumsum([0.02] * 4)):
utils.shift_axis(a, lshift=lshift)
utils.shift_axis(axes_lp[0], lshift=-0.05, rshift=0.01, tshift=-0.018)
utils.shift_axis(axes_lp[1], rshift=0.05, lshift=-0.01, tshift=-0.018)
return fig