Xenium数据分析 | 下机数据读取

今天我们将下载10x官方人肺癌FFPE样本Xenium5k下机数据，使用python的spatialdata库，演示如何进行Xenium单个样本/多样本数据读取，以及简单绘图功能展示。

1. 示例数据下载：

数据下载地址: https://www.10xgenomics.com/datasets/xenium-human-lung-cancer-post-xenium-technote

文件解压缩后，文件层级展示如下图，一般一个FOV对应的是一个样本，一个样本对应一个文件夹结果，一张芯片上最多可以选8个FOV，若果一张芯片上拼的样本数超过8个，就会有多个样本被并到一个FOV中，后续下机数据分析的时候想要拆分开的话，需要使用Xenium browser手动圈选，拿到个样本的barcodes，然后就可以拆分样本（一般TMA样本都需要手动圈选操作）。这里我们下载的数据就只有一个FOV，也就是只有一个样本，所以下图展示的是这一个样本的数据。

各关键文件说明如下，10x官方给出了很详细的说明:https://www.10xgenomics.com/cn/support/software/xenium-onboard-analysis/latest/analysis/xoa-output-understanding-outputs

File type	File and description
Experiment file	experiment.xenium : Experiment manifest file.
Interactive summary	analysis_summary.html : Summary metrics, graphs, and images to QC your run data in HTML format.
Image files	morphology.ome.tif : The 3D nuclei-stained (DAPI) morphology image in OME-TIFF format.
	morphology_focus/ : A directory containing the multi-focus projection of morphology image(s) in a multi-file OME-TIFF format (2D). The directory will contain the nuclei DAPI stain image, as well as three additional stain images for Xenium outputs generated with the multimodal cell segmentation assay workflow.
Cell summary	cells.csv.gz : Cell summary file.
	cells.parquet : Cell summary file in Parquet format.
Cell segmentation masks and polygons	cells.zarr.zip : Cell summary file in zipped Zarr format, only file that contains the nucleus and cell segmentation masks and boundaries used for transcript assignment.
Cell boundary polygons	cell_boundaries.csv.gz : Cell boundary file.
	cell_boundaries.parquet : Cell boundary file in Parquet format.
Nucleus boundary polygons	nucleus_boundaries.csv.gz : Nucleus boundary file.
	nucleus_boundaries.parquet : Nucleus boundary file in Parquet format.
Transcript data	transcripts.parquet : Transcripts data in Parquet format.
	transcripts.zarr.zip : Transcript data in zipped Zarr format.
Cell-feature matrix	cell_feature_matrix/ : Directory of the cell-feature matrix files in Market Exchange format.
	cell_feature_matrix.h5 : Cell-feature matrix file in HDF5 format.
	cell_feature_matrix.zarr.zip : Cell-feature matrix file in zipped Zarr format.
Metric summary	metrics_summary.csv : Summary of key metrics.
Secondary analysis	analysis/ : Directory of secondary analysis results.
	analysis.zarr.zip : Secondary analysis outputs in zipped Zarr format.
Gene panel	gene_panel.json : Copy of input gene panel file.
Auxiliary data (aux_outputs/)	morphology_fov_locations.json : Field of view (FOV) name and position information (in microns). overview_scan_fov_locations.json : FOV name and position information (in pixels). per_cycle_channel_images/ : Directory of downsampled RNA image files in TIFF format from each cycle and channel. overview_scan.png : Full resolution image of entire slide sample. background_qc_images/ : Directory of autofluorescence images (downsampled, TIFF format) that are subtracted from the raw stain images to produce the morphology_focus/ images if Cell Segmentation Staining protocol used.

analysis_summary.html对下机数据有个整体了解

下面展示的是细胞分割依据，有15.8%的细胞是根据细胞膜染色帮助识别分割细胞（这部分细胞分割最准确），有82.0%的细胞是通过针对细胞内18S核糖体RNA染色方法来标记并分割细胞，有2.2%的细胞通过DAPI识别出细胞核后，向外扩5um认为是细胞边界（这些细胞分割最不准确）

下面展示的是morphology_focus文件夹下的4个ome.tif文件，对应的是4个通道，0000是DAPI, 0001是green, 0002是yellow, 0003是red。

2. 安装依赖库

3. 数据读取

由于Xenium下机数据较大，多个样本按顺序读取耗时较长，这里我们使用多线程并行读取，缩短时间。

data_dir参数是xenium下机数据文件位置；

sample_info参数是样本信息.txt文件，一共三列，第一列是下机数据问价夹名称，第二列是样本名称，第三列是样本分组名称，使用'\t'分隔，有多个样本的，文件中就有多行。

import osimport threadingimport spatialdata as sdfrom spatialdata_io import xenium
# 多线程读取Xenium下机数据读取def xenium_data_load_multithreaded(data_dir, sample_info):    def sd_read_xenium(sample_data, sample_name, sdata_dict):        sdata = xenium(path=sample_data, cells_boundaries=True, n_jobs=6)        sdata_dict[sample_name] = sdata    threads = []    sdata_dict = {}    sample_2_group = {}    with open(sample_info, 'r') as f:        for line in f:            raw_name, sample_name, group_name = line.strip().split('\t')[:3]  # 这里根据自己实际情况修改            sample_2_group[sample_name] = group_name            thread = threading.Thread(target=sd_read_xenium, args=(os.path.join(data_dir, raw_name),sample_name, sdata_dict,))            threads.append(thread)            thread.start()    for thread in threads:        thread.join()        sdata = sd.concatenate(                sdata_dict,                concatenate_tables=True, # 这里是将多样本的单细胞数据合并在一起到table中                obs_names_make_unique=True            )    sdata.tables['table'].obs["sample"] = sdata.tables['table'].obs["region"].str.replace('cell_circles-', '')    sdata.tables['table'].obs["group"] = sdata.tables['table'].obs["sample"].apply(lambda x: sample_2_group[x])    sdata.tables['table'].obs["cell_boundaries"] = sdata.tables['table'].obs["region"].str.replace('cell_circles', 'cell_boundaries')    sdata.set_table_annotates_spatialelement(table_name='table', region=[i for i in sdata.shapes.keys() if i.startswith('cell_boundaries-')], region_key='cell_boundaries')
return sdata

简单绘图展示

fig, ax = plt.subplots(figsize=(10, 10))sdata.pl.render_images("morphology_focus-S1").pl.show(ax=ax, title="Morphology plot", coordinate_systems="global")
ax.grid(False)

基因表达

​​​​​​​​​​​​​​

from spatialdata import bounding_box_query
fig, ax = plt.subplots(figsize=(10, 10))crop0 = lambda x: bounding_box_query(    x,    min_coordinate=[10000, 20000],    max_coordinate=[15000, 25000],    axes=("x", "y"),    target_coordinate_system="global",)crop0(sdata).pl.render_shapes(    "cell_boundaries-S1",     color='EPCAM',     outline_width=0.3,     outline_alpha=0.9,     outline_color='grey').pl.show(ax=ax, title="EPCAM gene expression", coordinate_systems="global")ax.grid(False)ax.axis('off')