.Series
.X这个部分储存的是矩阵信息，数据结构是numpy array，和seurat对象一样，基因variable *细胞observation的稀疏矩阵。
但是.X的结构ndarray，是一个数组，是没有行名列名消息的。行名和列名消息存储在.obs和.var里。
调取矩阵信息：
import scvelo as scv
scv.DataFrame(adata.X)
TTTGCATGAGAGGC-1 454
TTTGCATGCCTCAC-1 724
Name: n_genes, Length: 2638, dtype: int64
查看有多少cluster/细胞亚群...
adata.obs['clusters'].unique()
.var存储的是基因的信息，数据结构是pandas dataframe。
adata.var
adata.uns['leiden']
# {'params': {'resolution': 1, 'random_state': 0, 'n_iterations': -1}}
adata.uns['leiden_colors']
# array(['#1f77b4', '#ff7f0e', '#279e68', '#d62728', '#aa40fc', '#8c564b', '#e377c2', '#b5bd61'], dtype=object)
这个uns的部分不是针对行/列的，而是针对行和列标注的参数的（暂时这么理解），在上述中pbmc的obs中有louvain观测，那么在uns中就是运行louvain算法的参数。是以哈希形式存储的。
obsm
对观测的多维的注释，m指的就是multi-dim多个维度。它是2-多维的ndarray，长度为n_obs。（obs是一个维度可以都放在一个数据框下）
adata.obsm
# AxisArrays with keys: X_pca, X_umap, X_draw_graph_fa
print(adata.obsm['X_pca'].shape,adata.obsm['X_umap'].shape)
adata.obsm['X_pca']
# (2638, 50) (2638, 2)
# array([[-5.5562196 , -0.257729 , 0.18678935, ..., -0.33962035,
# 1.482267 , 1.8977386 ],
# [-7.209527 , -7.4819927 , -0.1626957 , ..., -1.9776347 ,
# -1.5584233 , -1.4961302 ],
# [-2.694437 , 1.5836601 , 0.6631187 , ..., 0.543645 ,
# -0.54527736, -4.3395023 ],
# ...,
# [-0.78539336, -6.718588 , -1.5988318 , ..., -0.5611978 ,
# -0.10546814, 0.58385324],
# [ 0.2812712 , -5.921852 , -1.1628692 , ..., -1.3820586 ,
# 3.5802112 , 1.2988565 ],
# [-0.09076688, -0.66350466, -0.13485757, ..., 0.37319255,
# 0.75012326, -0.6659836 ]], dtype=float32)
varm
对特征的多维的注释，与obsm相对。
adata.varm
# AxisArrays with keys: PCs
print(adata.varm['PCs'].shape)
adata.varm['PCs']
# (1838, 50)
# array([[-2.6014808565e-02, 3.2541397959e-03, 1.8978352891e-03, ...,
# -5.1610143855e-03, 1.4520395547e-02, -6.6632591188e-04],
# [-8.2782376558e-03, 9.0831620619e-03, -7.8140682308e-04, ...,
# 3.0852310359e-02, -8.9336717501e-03, -2.8796317056e-03],
# [-3.3151865937e-03, 3.2096833456e-03, 2.7985233464e-04, ...,
# 1.0144758970e-02, -5.5128755048e-04, 1.5089971712e-03],
# ...,
# [ 8.3417436108e-03, -1.2465091422e-03, -4.1219405830e-03, ...,
# -1.0164264590e-02, 9.2523321509e-03, 2.7965774760e-02],
# [-1.6406573355e-02, 4.4101417065e-02, -2.1357089281e-05, ...,
# 9.9819628522e-03, -4.5361258090e-03, -1.3653293252e-02],
# [-1.5188260004e-02, 4.0008693933e-02, 5.4121930152e-03, ...,
# -3.6789905280e-03, 2.1117802709e-02, 3.5965379328e-02]])
obsp
针对观测的配对的注释，前两维都是n_obs。比如点与点之间的距离和连通性。
adata.obsp
# PairwiseArrays with keys: connectivities, distances
adata.obsp['connectivities']
# <2638x2638 sparse matrix of type '<class 'numpy.float32'>'
# with 41952 stored elements in Compressed Sparse Row format>
adata.obsp['distances']
# <2638x2638 sparse matrix of type '<class 'numpy.float64'>'
# with 23742 stored elements in Compressed Sparse Row format>
layers
在做速率分析的时候，还可以看到adata中有layers这一部分的信息。
adata = scv.datasets.pancreas()
adata
AnnData object with n_obs × n_vars = 3696 × 27998
obs: 'clusters_coarse', 'clusters', 'S_score', 'G2M_score'
var: 'highly_variable_genes'
uns: 'clusters_coarse_colors',
