从元数据+值创建xarray数据集的简单方法？

Question

我正在研究单细胞RNA测序数据，最近的数据是10k-100k个样本（cells）x 20k特征（genes）的稀疏值，还包括很多元数据，例如起源的组织（“大脑”与“肝脏”）。元数据是~10-100列，我将其存储为pandas.DataFrame。现在，我正在生成xarray.datasets，方法是对元数据进行dictify并将它们添加为坐标。因为我是在笔记本之间复制片段，所以看起来很笨重而且容易出错。有更简单的方法吗？在

cell_metadata_dict = cell_metadata.to_dict(orient='list')
coords = {k: ('cell', v) for k, v in cell_metadata_dict.items()}
coords.update(dict(gene=counts.columns, cell=counts.index))

ds = xr.Dataset(
    {'counts': (['cell', 'gene'], counts),
    },
    coords=coords)

编辑：

为了显示一些示例数据，下面是cell_metadata.head().to_csv()：

^{pr2}$

和counts.iloc[:5, :20].to_csv()

cell,0610005C13Rik,0610007C21Rik,0610007L01Rik,0610007N19Rik,0610007P08Rik,0610007P14Rik,0610007P22Rik,0610008F07Rik,0610009B14Rik,0610009B22Rik,0610009D07Rik,0610009L18Rik,0610009O20Rik,0610010B08Rik,0610010F05Rik,0610010K14Rik,0610010O12Rik,0610011F06Rik,0610011L14Rik,0610012G03Rik
A1-MAA100140-3_57_F-1-1,308,289,81,0,4,88,52,0,0,104,65,0,1,0,9,8,12,283,12,37
A10-MAA100140-3_57_F-1-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
A11-MAA100140-3_57_F-1-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
A12-MAA100140-3_57_F-1-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
A2-MAA100140-3_57_F-1-1,375,325,70,0,2,72,36,13,0,60,105,0,13,0,0,29,15,264,0,65

回复：pandas.DataFrame.to_xarray()-这是难以置信的慢，而且对我来说，将如此多的数字和分类数据编码为100级多索引似乎很奇怪。每次我尝试使用MultiIndex时，总是会让我说“哦，这就是为什么我不使用多索引”并恢复到拥有单独的元数据和统计数据帧的情况。在

Answer 1

Xarray使用pandas索引/列标签作为默认元数据。当所有变量共享相同的维度时，您可以在单个函数调用中进行转换，但是如果不同的变量具有不同的维度，则需要分别从pandas转换它们，然后将它们放在xarray端。例如：

import pandas as pd
import io
import xarray

# read your data
cell_metadata = pd.read_csv(io.StringIO(u"""\
cell,Uniquely mapped reads number,Number of input reads,EXP_ID,TAXON,WELL_MAPPING,Lysis Plate Batch,dNTP.batch,oligodT.order.no,plate.type,preparation.site,date.prepared,date.sorted,tissue,subtissue,mouse.id,FACS.selection,nozzle.size,FACS.instument,Experiment ID ,Columns sorted,Double check,Plate,Location ,Comments,mouse.age,mouse.number,mouse.sex
A1-MAA100140-3_57_F-1-1,428699,502312,170928_A00111_0068_AH3YKKDMXX,mus,MAA100140,,,,Biorad 96well,Stanford,,170720,Liver,Hepatocytes,3_57_F,,,,,,,,,,3,57,F
A10-MAA100140-3_57_F-1-1,324428,360285,170928_A00111_0068_AH3YKKDMXX,mus,MAA100140,,,,Biorad 96well,Stanford,,170720,Liver,Hepatocytes,3_57_F,,,,,,,,,,3,57,F
A11-MAA100140-3_57_F-1-1,381310,431800,170928_A00111_0068_AH3YKKDMXX,mus,MAA100140,,,,Biorad 96well,Stanford,,170720,Liver,Hepatocytes,3_57_F,,,,,,,,,,3,57,F
A12-MAA100140-3_57_F-1-1,393498,446705,170928_A00111_0068_AH3YKKDMXX,mus,MAA100140,,,,Biorad 96well,Stanford,,170720,Liver,Hepatocytes,3_57_F,,,,,,,,,,3,57,F
A2-MAA100140-3_57_F-1-1,717,918,170928_A00111_0068_AH3YKKDMXX,mus,MAA100140,,,,Biorad 96well,Stanford,,170720,Liver,Hepatocytes,3_57_F,,,,,,,,,,3,57,F"""))
counts = pd.read_csv(io.StringIO(u"""\
cell,0610005C13Rik,0610007C21Rik,0610007L01Rik,0610007N19Rik,0610007P08Rik,0610007P14Rik,0610007P22Rik,0610008F07Rik,0610009B14Rik,0610009B22Rik,0610009D07Rik,0610009L18Rik,0610009O20Rik,0610010B08Rik,0610010F05Rik,0610010K14Rik,0610010O12Rik,0610011F06Rik,0610011L14Rik,0610012G03Rik
A1-MAA100140-3_57_F-1-1,308,289,81,0,4,88,52,0,0,104,65,0,1,0,9,8,12,283,12,37
A10-MAA100140-3_57_F-1-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
A11-MAA100140-3_57_F-1-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
A12-MAA100140-3_57_F-1-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
A2-MAA100140-3_57_F-1-1,375,325,70,0,2,72,36,13,0,60,105,0,13,0,0,29,15,264,0,65"""))

# build the output
xarray_counts = xarray.DataArray(counts.set_index('cell'), dims=['cell', 'gene'])
xarray_counts.coords.update(cell_metadata.set_index('cell').to_xarray())
print(xarray_counts)

{cda>计算结果：

^{pr2}$

如果你想要一个数据集，把DataArray对象放到数据集构造函数中，例如

^{3}$