def explore_smushers():
import macosko2015
six_clusters, six_clusters_cells, six_clusters_genes = \
macosko2015.load_big_clusters()
hue = 'cluster_n_celltype'
groupby = six_clusters_cells[hue]
palette = fig_code.cluster_names_to_color
csv = macosko2015.BASE_URL + 'differential_clusters_lowrank.csv'
lowrank = pd.read_csv(csv, index_col=0)
lowrank_big_clusters = lowrank.loc[six_clusters.index,
six_clusters.columns]
algorithms = {'PCA': PCA, 'ICA': FastICA, 'NMF': NMF}
def _smush(n_components):
smushers = {}
smusheds = {}
summaries = {}
datas = {
'big clusters (lowrank)': lowrank_big_clusters,
'big clusters': six_clusters
}
for data_name, data in datas.items():
# print('data_name', data_name)
for algo_name, algorithm in algorithms.items():
# print('\talgo_name', algo_name)
smusher = algorithm(n_components=n_components,
random_state=2017)
if data.min().min() < 0:
nonnegative = data - data.min().min()
else:
nonnegative = data
if 'digits' in data_name:
# groupby = digits.target
index = digits.target
elif 'amacrine' in data_name:
# groupby = amacrine_cells['cluster_n']
index = data.index
else:
# groupby = cell_metadata['cluster_n_celltype']
index = data.index
smushed = pd.DataFrame(smusher.fit_transform(nonnegative),
index=index)
smushed.columns.name = f'{algo_name} components'
median = smushed.groupby(groupby).median()
mean = smushed.groupby(groupby).mean()
prefix = f'{data_name} {algo_name}'
smusheds[prefix] = smushed
smushers[prefix] = smusher
summaries[f'{prefix}: mean'] = mean
summaries[f'{prefix}: median'] = median
return smusheds, smushers, summaries
smusheds_n10, smushers_n10, summaries_n10 = _smush(n_components=10)
sns.set(style='whitegrid', context='notebook')
def plot_smushed(plot_type, algorithm, statistic, lowrank, n_components):
if n_components != 10:
smusheds_nX, smushers_nX, summaries_nX = _smush(n_components)
smusheds, smushers, summaries = \
smusheds_nX, smushers_nX, summaries_nX
else:
smusheds, smushers, summaries = \
smusheds_n10, smushers_n10, summaries_n10
key = 'big clusters'
if lowrank:
key += ' (lowrank)'
key += f' {algorithm}'
if plot_type == 'heatmap':
# statistics = 'mean', # 'median'
key_summary = f'{key}: {statistic}'
summary = summaries[key_summary]
fig, ax = plt.subplots()
sns.heatmap(summary)
ax.set(title=key_summary)
if plot_type == 'pairplot':
smushed = smusheds[key]
smushed_clusters = smushed.join(groupby)
sns.pairplot(smushed_clusters, hue=hue, palette=palette)
fig = plt.gcf()
fig.suptitle(key)
interact(plot_smushed,
plot_type=['heatmap', 'pairplot'],
algorithm=algorithms.keys(),
statistic=['mean', 'median'],
lowrank=False,
n_components=IntSlider(value=10, min=2, max=10))
def explore_phenograph():
"""Interactively shows KNN graphs and community detection"""
big_clusters, big_clusters_cells, big_clusters_genes = \
macosko2015.load_big_clusters()
amacrine, amacrine_cells, amacrine_genes = macosko2015.load_amacrine()
# --- Read the "lowrank" or "smoothed" data from robust PCA --- #
csv = macosko2015.BASE_URL + 'differential_clusters_lowrank.csv'
lowrank = pd.read_csv(csv, index_col=0)
lowrank_big_clusters = lowrank.loc[
big_clusters.index, big_clusters.columns]
lowrank_amacrine = lowrank.loc[amacrine.index, amacrine.columns]
datasets = {'big clusters lowrank': lowrank_big_clusters,
'big clusters': big_clusters,
'amacrine lowrank': lowrank_amacrine,
'amacrine': amacrine}
correls = {k: correlation_to_distance(data.T.rank().corr())
for k, data in datasets.items()}
def plot_phenograph(dataset='big clusters',
primary_metric='euclidean',
lowrank=False,
k=30, min_cluster_size=10):
key = dataset + ' lowrank' if lowrank else dataset
corr = correls[key]
if dataset == 'big clusters':
metadata = big_clusters_cells
palette = 'Set2'
cluster_col = 'cluster_id'
elif dataset == 'amacrine':
metadata = amacrine_cells
palette = 'husl'
cluster_col = 'cluster_id'
community_col = 'community'
communities, graph, Q = phenograph.cluster(
corr, k=k, primary_metric=primary_metric,
min_cluster_size=min_cluster_size)
network = networkx.from_scipy_sparse_matrix(graph)
positions = networkx.spring_layout(network)
nodes_source = ColumnDataSource(get_nodes_specs(
positions, metadata, corr.index, communities,
other_cluster_col=cluster_col,
community_col=community_col, palette=palette))
edges_source = ColumnDataSource(get_edges_specs(network, positions))
# --- First tab: KNN clustering --- #
tab1 = plot_graph(nodes_source, edges_source, legend_col=community_col,
color_col=f'{community_col}_color', tab=True,
title='KNN Clustering')
# --- Second tab: Clusters from paper --- #
tab2 = plot_graph(nodes_source, edges_source,
legend_col='cluster_n_celltype', tab=True,
color_col='other_cluster_color',
title="Clusters from paper")
tabs = Tabs(tabs=[tab1, tab2])
show(tabs)
interact(plot_phenograph, dataset=['big clusters', 'amacrine'],
primary_metric=['euclidean', 'manhattan'],
k=IntSlider(start=3, stop=100, value=30, step=5),
min_cluster_size=IntSlider(start=3, stop=100, value=10, step=5))
import warnings
import fastcluster
from ipywidgets import interact, IntSlider
import macosko2015
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import polo
from scipy.spatial import distance
import seaborn as sns
FIGURE_FOLDER = 'figures'
DATA_FOLDER = 'data'
expression, cell_metadata, gene_metadata = macosko2015.load_big_clusters()
cluster_ids_unique = cell_metadata['cluster_id'].unique()
cluster_n_to_name = {
24: 'Rods',
25: 'Cones',
26: 'Bipolar cells\n(group1)',
27: 'Bipolar cells\n(group2)',
33: 'Bipolar cells\n(group3)',
34: 'Muller glia'
}
cluster_id_to_name = dict(('cluster_{}'.format(str(i).zfill(2)), name)
for i, name in cluster_n_to_name.items())
colors = sns.color_palette(palette='Set2', n_colors=len(cluster_ids_unique))