def plot_y_centroids(yi,
lami,
wi,
vlim=(-4, 4),
fs_xlabel=12,
fs_ylabel=12,
fs_xticks=12,
fs_yticks=12,
rotation=90,
ha="center",
markernames=[],
ylab="subpopulations",
gridlines_color='black',
gridlines_lw=1,
cm=blue2red.cm(9),
population=None,
Zi=None):
J = yi.shape[1]
K = wi.shape[0]
selected_features = np.argwhere(wi > 0)[:, 0]
K_sel = selected_features.shape[0]
wi_sel = wi[selected_features]
selected_features = selected_features[np.argsort(wi_sel)]
wi_sel_sorted = wi_sel[np.argsort(wi_sel)]
y_centers = np.zeros((J, K_sel))
yticks = [0] * K_sel
for j in range(J):
for k in range(K_sel)[::-1]:
k_ = selected_features[k] + 1
y_centers[j, k] = yi[lami == k_, j].mean()
w_ik_perc = (wi_sel_sorted[k] * 100).round(1)
if population is None or Zi is None:
yticks[k] = '{} ({}%)'.format(k_, w_ik_perc)
else:
label = population.label(Zi[:, k_ - 1])
yticks[k] = '{} ({}%)'.format(label, w_ik_perc)
im = plt.imshow(y_centers.T,
aspect='auto',
cmap=cm,
vmin=vlim[0],
vmax=vlim[1])
if markernames == []:
markernames = np.arange(J) + 1
plt.xticks(range(J),
markernames,
rotation=rotation,
fontsize=fs_xticks,
ha=ha)
plt.yticks(range(K_sel), yticks, fontsize=fs_yticks)
plt.xlabel('markers', fontsize=fs_xlabel)
plt.ylabel(ylab, fontsize=fs_ylabel)
gridlines(y_centers.T, color=gridlines_color, lw=gridlines_lw)
colorbar_horizontal(im)
def plot_y(yi,
wi_mean,
lami_est,
fs_lab=10,
fs_cbar=10,
lw=3,
cm=blue2red.cm(6),
vlim=(-3, 3),
fs_xlab=10,
fs_ylab=10,
ylab="cells",
markernames=[],
interpolation=None,
rotation=90,
ha="center"):
J = yi.shape[1]
vmin, vmax = vlim
if type(wi_mean) == int:
K = wi_mean
wi_mean = np.array([(lami_est == k + 1).mean() for k in range(K)])
lami_new, counts = relabel_lam(lami_est, wi_mean)
counts_cumsum = np.cumsum(counts)
yi_sorted = yi[np.argsort(lami_new), :]
im = plt.imshow(yi_sorted,
aspect='auto',
vmin=vmin,
vmax=vmax,
cmap=cm,
interpolation=interpolation)
for c in counts_cumsum[:-1]:
plt.axhline(c, color='yellow', linewidth=lw)
plt.xticks(rotation=rotation, ha=ha)
if len(markernames) == 0:
plt.xticks(np.arange(J), np.arange(J) + 1, fontsize=fs_xlab)
else:
plt.xticks(np.arange(J), markernames, fontsize=fs_xlab)
plt.yticks(fontsize=fs_ylab)
plt.xlabel("markers", fontsize=fs_lab)
plt.ylabel(ylab, fontsize=fs_lab)
ax = plt.gca()
ax_divider = make_axes_locatable(ax)
cax = ax_divider.append_axes("top", size="7%", pad="2%")
cax.xaxis.set_ticks_position("top")
cbar = colorbar(im, cax=cax, orientation="horizontal")
cbar.ax.tick_params(labelsize=fs_cbar)
y = copy.deepcopy(data['y'])
# Get a subsampmle of data
if 0 < subsample < 1:
for i in range(len(y)):
Ni = y[i].shape[0]
idx = np.random.choice(Ni, int(Ni * subsample), replace=False)
y[i] = y[i][idx, :]
# Print size of data
print('N: {}'.format([yi.shape[0] for yi in y]))
# Color map
cm_greys = plt.cm.get_cmap('Greys', 5)
VMIN, VMAX = VLIM = (-4, 4)
cm = blue2red.cm(9)
# Plot yi histograms
# plt.hist(y[0][:, 1], bins=100, density=True); plt.xlim(-15, 15); plt.show()
# plt.hist(y[1][:, 3], bins=100, density=True); plt.xlim(-15, 15); plt.show()
# plt.hist(y[2][:, -1], bins=100, density=True); plt.xlim(-15, 15); plt.show()
# Heatmaps
for i in range(I):
plt.imshow(y[i], aspect='auto', vmin=VMIN, vmax=VMAX, cmap=cm)
plt.colorbar()
plt.savefig('{}/y{}.pdf'.format(img_dir, i + 1))
plt.close()
K = 30
L = [5, 3]
yi_path = f'{path}/img/txt/y{i}_mean.csv'
yi = np.loadtxt(yi_path, delimiter=',')
lami_path = f'{path}/img/txt/lam{i}_best.txt'
lami = np.loadtxt(lami_path, dtype=int)
wi_path = f'{path}/img/txt/W{i}_best.txt'
wi = np.loadtxt(wi_path)
zi_path = f'{path}/img/txt/Z{i}_best.txt'
zi = np.loadtxt(zi_path)
# Plot
plt.figure(figsize=(6, 6))
plot_yz.plot_y_centroids(yi,
lami,
wi,
vlim=(-3, 3),
cm=blue2red.cm(6),
population=population,
Zi=zi,
fs_xlabel=16,
fs_ylabel=16,
fs_xticks=16,
fs_yticks=16)
outpath = f'{path}/img/y{i}_centroid.pdf'
plt.savefig(outpath, bbox_inches="tight")
plt.close()
# Plot Z estimate.
plt.figure(figsize=(6, 6))
plot_yz.plot_Z(Z_mean=zi,
wi_mean=wi,
lami_est=lami,
population = Population()
# Rcounts
R_path = f'{path}/img/txt/Rcounts.txt'
R = np.loadtxt(R_path)
plt.figure(figsize=(5,5))
zinfo.plot_num_selected_features(R, refs=[6, 5], ymax=1.05)
plt.savefig(f'{path}/img/Rcounts.pdf', bbox_inches='tight')
plt.close()
for i in (1, 2):
# Read data
yi_path = f'{path}/img/txt/y{i}_mean.csv'
yi = np.loadtxt(yi_path, delimiter=',')
lami_path = f'{path}/img/txt/lam{i}_best.txt'
lami = np.loadtxt(lami_path, dtype=int)
wi_path = f'{path}/img/txt/W{i}_best.txt'
wi = np.loadtxt(wi_path)
zi_path = f'{path}/img/txt/Z{i}_best.txt'
zi = np.loadtxt(zi_path)
# Plot
plt.figure(figsize=(6,6))
plot_yz.plot_y_centroids(yi, lami, wi, vlim=(-3, 3), cm=blue2red.cm(6),
population=population, Zi=zi,
fs_xlabel=16, fs_ylabel=16,
fs_xticks=16, fs_yticks=16)
outpath = f'{path}/img/y{i}_centroid.pdf'
plt.savefig(outpath, bbox_inches="tight")
plt.close()
def plot_yz(yi,
Z_mean,
wi_mean,
lami_est,
w_thresh=.01,
cm_greys=plt.cm.get_cmap('Greys', 5),
markernames=[],
rotation=90,
ha="center",
cm_y=blue2red.cm(6),
vlim_y=(-3, 3),
fs_w=10,
w_digits=1):
J = yi.shape[1]
vmin_y, vmax_y = vlim_y
# cm_y.set_bad(color='black')
# cm_y.set_under(color='blue')
# cm_y.set_over(color='red')
# gs = gridspec.GridSpec(1, 2, width_ratios=[2, 5])
gs = gridspec.GridSpec(2, 1, height_ratios=[5, 2])
# Plot y
lami_new, counts = relabel_lam(lami_est, wi_mean)
counts_cumsum = np.cumsum(counts)
yi_sorted = yi[np.argsort(lami_new), :]
plt.subplot(gs[0])
im = plt.imshow(yi_sorted,
aspect='auto',
vmin=vmin_y,
vmax=vmax_y,
cmap=cm_y)
for c in counts_cumsum[:-1]:
plt.axhline(c, color='yellow')
plt.xticks(rotation=rotation, ha=ha)
if len(markernames) == 0:
plt.xticks(np.arange(J), np.arange(J) + 1)
else:
plt.xticks(np.arange(J), markernames)
ax = plt.gca()
ax_divider = make_axes_locatable(ax)
cax = ax_divider.append_axes("top", size="7%", pad="2%")
cax.xaxis.set_ticks_position("top")
colorbar(im, cax=cax, orientation="horizontal")
# Plot Z
k_ord = wi_mean.argsort()
z_cols = []
for k in k_ord.tolist():
if wi_mean[k] > w_thresh:
z_cols.append(k)
z_cols = np.array(z_cols)
Z_hat = Z_mean[:, z_cols].T
plt.subplot(gs[1])
im = plt.imshow(Z_hat, aspect='auto', vmin=0, vmax=1, cmap=cm_greys)
ax = plt.gca()
plt.xticks([])
plt.yticks(np.arange(len(z_cols)), z_cols + 1, fontsize=fs_w)
add_gridlines_Z(Z_hat)
plt.colorbar(orientation='horizontal', pad=.05)
# add wi_mean on right side
K = z_cols.shape[0]
ax2 = ax.twinx()
ax2.set_yticks(range(K))
w_perc = wi_mean[z_cols]
w_perc = [str((wp * 100).round(w_digits)) + '%' for wp in w_perc]
plt.yticks((K - 1) / K * np.arange(K) + .5, w_perc[::-1], fontsize=fs_w)
plt.yticks()
ax2.tick_params(length=0)
fig = plt.gcf()
fig.subplots_adjust(hspace=0.2)