import sys
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import loompy
import velocyto as vcy
import logging
import pandas as pd
from sklearn.svm import SVR
from sklearn.linear_model import LinearRegression
from statsmodels.nonparametric.smoothers_lowess import lowess
from scipy.interpolate import interp1d
# In[4]:
vlm = vcy.load_velocyto_hdf5(snakemake.input[0])
# In[ ]:
# In[ ]:
## color condes consistent with main text
from collections import defaultdict
monocytelabels = defaultdict(lambda: 'Grey')
#ice_cream = {}
monocytelabels['4'] = '#9382ae'
monocytelabels['3'] = '#fdceb8'
monocytelabels['12'] = '#df5974'
def main():
if len(sys.argv) != 2:
print('runs the velocyto workflow on a given hdf5 object')
print('usage: velocyto_workflow <path to file>')
sys.exit()
input_path = sys.argv[1]
output_path = input_path[:-5] + '_tsne33_velocity.hdf5'
print('loading data')
vlm = vcy.load_velocyto_hdf5(input_path)
print(len(vlm.ca['CellID']), 'cells')
print(len(vlm.ra['Gene']), 'genes')
print('filtering cells')
vlm.filter_cells(bool_array=vlm.initial_Ucell_size > np.percentile(
vlm.initial_Ucell_size, 0.5))
print('filtering genes')
vlm.score_cv_vs_mean(3000, plot=False, max_expr_avg=35)
vlm.filter_genes(by_cv_vs_mean=True)
print(len(vlm.ca['CellID']), 'cells')
print(len(vlm.ra['Gene']), 'genes')
#print('setting sample names as clusters')
#samplenames = list(map(lambda x: x.split(':')[0], vlm.ca['CellID']))
#vlm.ca['sample_name'] = samplenames
#vlm.set_clusters(vlm.ca["sample_name"])
print('normalizing data matrices')
vlm._normalize_S(relative_size=vlm.S.sum(0),
target_size=vlm.S.sum(0).mean())
vlm._normalize_U(relative_size=vlm.U.sum(0),
target_size=vlm.U.sum(0).mean())
print('running pca')
vlm.perform_PCA()
print('knn smoothing')
vlm.knn_imputation(n_pca_dims=15,
k=500,
balanced=True,
b_sight=3000,
b_maxl=1500,
n_jobs=20)
print('fit gammas')
vlm.fit_gammas()
print('calculate velocity')
vlm.predict_U()
vlm.calculate_velocity()
vlm.calculate_shift(assumption="constant_velocity")
vlm.extrapolate_cell_at_t(delta_t=1.)
print('running tsne')
bh_tsne = TSNE(random_state=33)
vlm.ts = bh_tsne.fit_transform(vlm.pcs[:, :15])
print('projection of velocity onto embeddings')
vlm.estimate_transition_prob(hidim="Sx_sz",
embed="ts",
transform="sqrt",
psc=1,
n_neighbors=3500,
knn_random=True,
sampled_fraction=0.5)
print('calculate embedding shift')
vlm.calculate_embedding_shift(sigma_corr=0.05, expression_scaling=True)
print('calculate grid arrows')
vlm.calculate_grid_arrows(smooth=0.8, steps=(40, 40), n_neighbors=100)
print('saving hdf5')
vlm.to_hdf5(output_path)
import loompy
import glob
import velocyto as vcy
import numpy as np
from sklearn.manifold import TSNE
#print('reading loom file')
#vlm = vcy.VelocytoLoom('/projects/pytrik/sc_adipose/analyze_10x_fluidigm/data/velocyto/all10x.loom')
print('loading data')
vlm = vcy.load_velocyto_hdf5(
'/projects/pytrik/sc_adipose/analyze_10x_fluidigm/data/velocyto/all10x-downsampled-36.hdf5'
)
print(len(vlm.ca['CellID']), 'cells')
print(len(vlm.ra['Gene']), 'genes')
print('filtering cells')
vlm.filter_cells(bool_array=vlm.initial_Ucell_size > np.percentile(
vlm.initial_Ucell_size, 0.5))
print('filtering genes')
vlm.score_cv_vs_mean(3000, plot=False, max_expr_avg=35)
vlm.filter_genes(by_cv_vs_mean=True)
print(len(vlm.ca['CellID']), 'cells')
print(len(vlm.ra['Gene']), 'genes')
#print('setting sample names as clusters')
#samplenames = list(map(lambda x: x.split(':')[0], vlm.ca['CellID']))
#vlm.ca['sample_name'] = samplenames
]
cellid["strain"] = np.where(cellid["sample"].isin(ab1), "ab1", "renca")
cellid["response"] = np.where(cellid["sample"].isin(respondervec), "responder",
"nonresponder")
# In[ ]:
targetmono = pd.read_csv(snakemake.input[3])
monos = targetmono["loom_cellid"]
ind = cellid["cellnames"].isin(monos)
cluster15mono = cellid.index[ind]
# In[ ]:
# extract projections from renca velocyto object and calculate momentum
renca = vcy.load_velocyto_hdf5(snakemake.input[1])
delta_embedding_renca = pd.DataFrame(renca.delta_embedding)
delta_embedding_renca.index = renca.ca["CellID"]
renca_ly6_cells_deltas = delta_embedding_renca.reindex(cluster15mono).dropna()
renca_ly6_cells_deltas.columns = ["deltax", "deltay"]
renca_ly6_cells_deltas["vel"] = np.square(
renca_ly6_cells_deltas["deltax"]) + np.square(
renca_ly6_cells_deltas["deltay"])
ly6crenca = pd.concat([renca_ly6_cells_deltas, cellid], axis=1,
join="inner") # merge with cell metadata
del (renca)
# In[ ]:
# extract projections from ab1 velocyto object and calculate momentum
ab1 = vcy.load_velocyto_hdf5(snakemake.input[0])
import loompy
import glob
import velocyto as vcy
import numpy as np
from sklearn.manifold import TSNE
print('loading data')
vlm = vcy.load_velocyto_hdf5(
'/projects/pytrik/sc_adipose/analyze_10x_fluidigm/data/velocyto/180831.hdf5'
)
print(len(vlm.ca['CellID']), 'cells')
print(len(vlm.ra['Gene']), 'genes')
#print('filtering cells')
#vlm.filter_cells(bool_array=vlm.initial_Ucell_size > np.percentile(vlm.initial_Ucell_size, 0.5))
print('filtering genes')
#vlm.score_cv_vs_mean(3000, plot=False, max_expr_avg=35)
vlm.score_cv_vs_mean(3000, plot=False)
vlm.filter_genes(by_cv_vs_mean=True)
print(len(vlm.ca['CellID']), 'cells')
print(len(vlm.ra['Gene']), 'genes')
print('normalizing data matrices')
vlm._normalize_S(relative_size=vlm.S.sum(0), target_size=vlm.S.sum(0).mean())
vlm._normalize_U(relative_size=vlm.U.sum(0), target_size=vlm.U.sum(0).mean())
print('running pca')
vlm.perform_PCA()
vlm.estimate_transition_prob(hidim="Sx_sz",
embed="ts",
transform="sqrt",
psc=1,
knn_random=True,
sampled_fraction=0.3,
random_seed=42)
vlm.calculate_embedding_shift(sigma_corr=0.05, expression_scaling=True)
vlm.calculate_grid_arrows(smooth=0.5, steps=(40, 40), n_neighbors=50)
vlm.to_hdf5("combined.hdf5")
###
###
vlm = vcy.load_velocyto_hdf5("combined.hdf5")
def despline():
ax1 = plt.gca()
# Hide the right and top spines
ax1.spines['right'].set_visible(False)
ax1.spines['top'].set_visible(False)
# Only show ticks on the left and bottom spines
ax1.yaxis.set_ticks_position('left')
ax1.xaxis.set_ticks_position('bottom')
def minimal_xticks(start, end):
end_ = np.around(end, -int(np.log10(end)) + 1)
xlims = np.linspace(start, end_, 5)