Common_data = RNA_data_scaled[np.intersect1d(osmFISH_data_scaled.columns,RNA_data_scaled.columns)]
n_factors = 30
n_pv = 30
n_pv_display = 30
dim_reduction = 'pca'
dim_reduction_target = 'pca'
pv_FISH_RNA = PVComputation(
n_factors = n_factors,
n_pv = n_pv,
dim_reduction = dim_reduction,
dim_reduction_target = dim_reduction_target
)
pv_FISH_RNA.fit(Common_data,osmFISH_data_scaled[Common_data.columns])
fig = plt.figure()
sns.heatmap(pv_FISH_RNA.initial_cosine_similarity_matrix_[:n_pv_display,:n_pv_display], cmap='seismic_r',
center=0, vmax=1., vmin=0)
plt.xlabel('osmFISH',fontsize=18, color='black')
plt.ylabel('Allen_SSp',fontsize=18, color='black')
plt.xticks(np.arange(n_pv_display)+0.5, range(1, n_pv_display+1), fontsize=12)
plt.yticks(np.arange(n_pv_display)+0.5, range(1, n_pv_display+1), fontsize=12, rotation='horizontal')
plt.gca().set_ylim([n_pv_display,0])
plt.show()
plt.figure()
sns.heatmap(pv_FISH_RNA.cosine_similarity_matrix_[:n_pv_display,:n_pv_display], cmap='seismic_r',
center=0, vmax=1., vmin=0)
for i in range(n_pv_display-1):
for i in Common_data.columns:
print(i)
start = tm.time()
from principal_vectors import PVComputation
n_factors = 50
n_pv = 50
dim_reduction = 'pca'
dim_reduction_target = 'pca'
pv_FISH_RNA = PVComputation(n_factors=n_factors,
n_pv=n_pv,
dim_reduction=dim_reduction,
dim_reduction_target=dim_reduction_target)
pv_FISH_RNA.fit(Common_data.drop(i, axis=1),
seqFISH_data_scaled[Common_data.columns].drop(i, axis=1))
S = pv_FISH_RNA.source_components_.T
Effective_n_pv = sum(np.diag(pv_FISH_RNA.cosine_similarity_matrix_) > 0.3)
S = S[:, 0:Effective_n_pv]
Common_data_t = Common_data.drop(i, axis=1).dot(S)
FISH_exp_t = seqFISH_data_scaled[Common_data.columns].drop(i,
axis=1).dot(S)
precise_time.append(tm.time() - start)
start = tm.time()
nbrs = NearestNeighbors(n_neighbors=50, algorithm='auto',
metric='cosine').fit(Common_data_t)
distances, indices = nbrs.kneighbors(FISH_exp_t)
n_pv = i
dim_reduction = 'pca'
dim_reduction_target = 'pca'
pv_FISH_RNA = PVComputation(
n_factors = n_factors,
n_pv = n_pv,
dim_reduction = dim_reduction,
dim_reduction_target = dim_reduction_target
)
source_data = RNA_data_scaled[Variance.index[0:i]]
target_data = Starmap_data_scaled[Variance.index[0:i]]
pv_FISH_RNA.fit(source_data,target_data)
S = pv_FISH_RNA.source_components_.T
Effective_n_pv = sum(np.diag(pv_FISH_RNA.cosine_similarity_matrix_) > 0.3)
S = S[:,0:Effective_n_pv]
RNA_data_t = source_data.dot(S)
FISH_exp_t = target_data.dot(S)
nbrs = NearestNeighbors(n_neighbors=50, algorithm='auto',
metric = 'cosine').fit(RNA_data_t)
distances, indices = nbrs.kneighbors(FISH_exp_t)
for j in range(0,Starmap_data.shape[0]):
weights = 1-(distances[j,:][distances[j,:]<1])/(np.sum(distances[j,:][distances[j,:]<1]))
for i in Common_data.columns:
print(i)
start = tm.time()
from principal_vectors import PVComputation
n_factors = 50
n_pv = 50
dim_reduction = 'pca'
dim_reduction_target = 'pca'
pv_FISH_RNA = PVComputation(n_factors=n_factors,
n_pv=n_pv,
dim_reduction=dim_reduction,
dim_reduction_target=dim_reduction_target)
pv_FISH_RNA.fit(Common_data.drop(i, axis=1),
Starmap_data_scaled[Common_data.columns].drop(i, axis=1))
S = pv_FISH_RNA.source_components_.T
Effective_n_pv = sum(np.diag(pv_FISH_RNA.cosine_similarity_matrix_) > 0.3)
S = S[:, 0:Effective_n_pv]
Common_data_t = Common_data.drop(i, axis=1).dot(S)
FISH_exp_t = Starmap_data_scaled[Common_data.columns].drop(i,
axis=1).dot(S)
precise_time.append(tm.time() - start)
start = tm.time()
nbrs = NearestNeighbors(n_neighbors=50, algorithm='auto',
metric='cosine').fit(Common_data_t)
distances, indices = nbrs.kneighbors(FISH_exp_t)
Common_data = RNA_data_scaled[np.intersect1d(Starmap_data_scaled.columns,
RNA_data_scaled.columns)]
n_factors = 50
n_pv = 50
n_pv_display = 50
dim_reduction = 'pca'
dim_reduction_target = 'pca'
pv_FISH_RNA = PVComputation(n_factors=n_factors,
n_pv=n_pv,
dim_reduction=dim_reduction,
dim_reduction_target=dim_reduction_target)
pv_FISH_RNA.fit(Common_data, Starmap_data_scaled[Common_data.columns])
fig = plt.figure()
sns.heatmap(pv_FISH_RNA.
initial_cosine_similarity_matrix_[:n_pv_display, :n_pv_display],
cmap='seismic_r',
center=0,
vmax=1.,
vmin=0)
plt.xlabel('Starmap', fontsize=18, color='black')
plt.ylabel('Allen_VISp', fontsize=18, color='black')
plt.xticks(np.arange(n_pv_display) + 0.5,
range(1, n_pv_display + 1),
fontsize=12)
plt.yticks(np.arange(n_pv_display) + 0.5,
range(1, n_pv_display + 1),
Common_data = RNA_data_scaled[np.intersect1d(Starmap_data_scaled.columns,
RNA_data_scaled.columns)]
n_factors = 50
n_pv = 50
n_pv_display = 50
dim_reduction = 'pca'
dim_reduction_target = 'pca'
pv_FISH_RNA = PVComputation(n_factors=n_factors,
n_pv=n_pv,
dim_reduction=dim_reduction,
dim_reduction_target=dim_reduction_target)
pv_FISH_RNA.fit(Common_data, Starmap_data_scaled[Common_data.columns])
plt.figure(figsize=(4, 4))
sns.heatmap(pv_FISH_RNA.
initial_cosine_similarity_matrix_[:n_pv_display, :n_pv_display],
cmap='seismic_r',
center=0,
vmax=1.,
vmin=0)
plt.xlabel('Starmap', fontsize=12, color='black')
plt.ylabel('Allen_VISp', fontsize=12, color='black')
plt.xticks(np.arange(4, n_pv_display, 5) + 0.5,
range(5, n_pv_display + 1, 5),
fontsize=8)
plt.yticks(np.arange(4, n_pv_display, 5) + 0.5,
range(5, n_pv_display + 1, 5),