#
# Copyright (c) 2019 10X Genomics, Inc. All rights reserved.
#
"""Utilities designed to help create the web summary used in Spatial."""
from __future__ import annotations
from typing import NamedTuple
import pandas as pd
import cellranger.rna.library as rna_library
import cellranger.websummary.analysis_tab_core as cr_atc
import cellranger.websummary.plotly_tools as pltly
from cellranger.analysis.singlegenome import SingleGenomeAnalysis
from cellranger.spatial.data_utils import IMAGEX_LOWRES, IMAGEY_LOWRES
from cellranger.spatial.image import WebImage
from cellranger.websummary.numeric_converters import round_floats_in_list
from cellranger.websummary.react_components import (
SharedCoordinatePlotCollection,
)
SPATIAL_PIPELINE_NAME = "count"
SPATIAL_COMMAND_NAME = "Space Ranger"
class SpatialReportArgs(NamedTuple):
sample_id: str
sample_desc: str
scalefactors: str
matrix: str
tissue_lowres_image: str
detected_tissue_image: str
qc_resampled_cyta_img: str
qc_regist_target_img: str
tissue_positions: str
analysis: str
target_set_name: str
target_panel_summary: str
feature_ref_path: str
targeting_method: str
reorientation_mode: str
loupe_alignment_file: str
filter_probes: bool
aligner: str
def _make_array_plot_layout(title, encoded_image):
xlo, xhi, ylo, yhi = encoded_image.cropbox
layout = {
"title": {
"text": title,
"yanchor": "top",
},
"autosize": True,
"hovermode": "closest",
"legend": {"y": 0.9, "yanchor": "top"},
"xaxis": {
"type": "linear",
"fixedrange": True,
"showgrid": False,
"zeroline": False,
"showticklabels": False,
"range": [xlo, xhi],
"scaleanchor": "y",
"scaleratio": 1.0,
},
"yaxis": {
"type": "linear",
"fixedrange": True,
"showgrid": False,
"zeroline": False,
"showticklabels": False,
"range": [yhi, ylo],
"domain": [0.0, 0.90],
},
"margin": {
"b": 10,
"t": 10,
},
"images": [
{
"source": encoded_image.base64_encoded_str,
"xref": "x",
"yref": "y",
"x": 0,
"y": 0,
"sizex": encoded_image.width,
"sizey": encoded_image.height,
"sizing": "stretch",
"opacity": 1,
"layer": "below",
},
],
"sliders": [
{
"pad": {"t": 10, "r": 0, "b": 10},
"yanchor": "center",
"xanchor": "left",
"active": 4,
"showactive": True,
"tickwidth": 1,
"currentvalue": {
"xanchor": "center",
"prefix": "Spot Opacity: ",
"font": {"color": "#888", "size": 12},
"visible": False,
},
"steps": [
{"label": "0.0", "method": "restyle", "args": ["opacity", 0.00]},
{"label": "", "method": "restyle", "args": ["opacity", 0.25]},
{"label": "", "method": "restyle", "args": ["opacity", 0.50]},
{"label": "", "method": "restyle", "args": ["opacity", 0.75]},
{"label": "1.0", "method": "restyle", "args": ["opacity", 1.00]},
],
}
],
}
return layout
def make_array_plot(title, encoded_image, data):
"""Generate plot layout for displaying spatial images overlaid with spots."""
assert isinstance(encoded_image, WebImage)
return {
"config": pltly.SPATIAL_PLOT_CONFIG,
"layout": _make_array_plot_layout(title, encoded_image),
"data": data,
}
def umi_on_spatial_plot(sample_data, coords, encoded_image, library_type):
""":param sample_data: Instance of webshim.data.SampleData.
:param encoded_image: Instance of WebImage
:return:
"""
analysis = sample_data.get_analysis(SingleGenomeAnalysis)
if analysis is None:
return None
# make a df with the barcodes as the index
bc_umi_counts = pd.DataFrame(index=analysis.matrix.bcs)
# Add the UMI column by getting the sum UMI for each barcode
matrix = analysis.matrix.select_features_by_type(library_type)
umi_per_bc = matrix.get_counts_per_bc()
# This append operation converts the ints to floats to account for the nan
# after dropping missing values we convert back
org_type = umi_per_bc.dtype
umi_col_name = "UMI"
bc_umi_counts[umi_col_name] = umi_per_bc
cbd = coords.join(bc_umi_counts)
cbd = cbd.dropna()
cbd = cbd.astype({umi_col_name: org_type})
color = cr_atc.get_umi_color(cbd[umi_col_name].values)
title = "Tissue Plot with Spots Colored by UMI Count"
data = [
{
"name": "Spots",
"x": round_floats_in_list(cbd[IMAGEX_LOWRES].values),
"y": round_floats_in_list(cbd[IMAGEY_LOWRES].values),
"type": "scatter",
"mode": "markers",
"marker": {
"opacity": 0.9,
"size": encoded_image.markersize,
"sizemode": "diameter",
"color": color,
"colorscale": "Jet",
"colorbar": {
"title": "UMI counts",
"yanchor": "top",
"y": 0.9,
"len": 0.9,
"xpad": 20,
},
},
"hovertemplate": "(%{x:.3f}, %{y:.3f})
" + "UMI counts: %{text}",
"text": [f"{umis:,d}" for umis in cbd[umi_col_name].values],
}
]
return make_array_plot(title, encoded_image, data)
def tissue_plots_by_clustering_spatial(
sample_data, coords, encoded_image, library_type=rna_library.GENE_EXPRESSION_LIBRARY_TYPE
):
"""Get the data structure that represents the plots for all the tissues.
:param sample_data: Instance of webshim.data.SampleData
:param coords:
:param encoded_image: Instance of WebImage
:return: A SharedCoordinatePlotCollection object
"""
if sample_data is None:
return None
analysis = sample_data.get_analysis(SingleGenomeAnalysis)
return tissue_plots_by_clustering_spatial_helper(analysis, coords, encoded_image, library_type)
def tissue_plots_by_clustering_spatial_helper(
analysis,
coords,
encoded_image,
library_type,
title="Tissue Plot with Spots Colored by Clustering",
):
"""Get the data structure that represents the plots for all the tissues.
Args:
analysis: the loaded analysis h5 object
coords:
encoded_image: Instance of WebImage
library_type: GEX or Antibody
Returns:
SharedCoordinatePlotCollection: object
"""
if analysis is None:
return None
# order clustering by order: graph, kmeans=2, =3 etc
clusterings = cr_atc.sort_analysis_clusterings(analysis.clusterings)
# align clustering label with image by barcode
if library_type == rna_library.GENE_EXPRESSION_LIBRARY_TYPE:
clusters_dict = {
clustering_key: clustering.clusters
for (clustering_key, clustering) in clusterings
if clustering_key[0:8] != "antibody"
}
elif library_type == rna_library.ANTIBODY_LIBRARY_TYPE:
clusters_dict = {
clustering_key: clustering.clusters
for (clustering_key, clustering) in clusterings
if clustering_key[0:8] == "antibody"
}
clusters_dict["barcode"] = analysis.matrix.bcs
clusters_df = pd.DataFrame.from_dict(clusters_dict)
clusters_df.set_index("barcode", inplace=True)
cluster_tissue_df = clusters_df.join(coords[[IMAGEX_LOWRES, IMAGEY_LOWRES]])
# pylint: disable=too-many-function-args
tissue_plot_data = SharedCoordinatePlotCollection(
cluster_tissue_df[IMAGEX_LOWRES].to_list(),
cluster_tissue_df[IMAGEY_LOWRES].to_list(),
pltly.SPATIAL_PLOT_CONFIG,
_make_array_plot_layout(title, encoded_image),
"scatter",
{"size": encoded_image.markersize, "sizemode": "diameter"},
)
for clustering_key, clustering in clusterings:
if clustering_key in clusters_dict:
tissue_plot_data.add_clustering(clustering_key, clustering)
return tissue_plot_data