10X单细胞 && 10X空间转录组数据分析之量化样本间异质性

作者，追风少年i

国庆前的最后一周了，大家好好努力，有句台词说得好，Yesterday is history， Tomorrow is a mystery， But today is a gift That is why it’s called the present (the gift) 。

这一篇来回答一些项目中遇到的问题，以及简单分享一下如何衡量样本间的异质性。

最近很多老师来做分析，其中大多数拿着疾病的样本，然后告诉我做什么分析，要分析什么老师也不知道，只是说做分析，搞得我难以下手，简单的课题设计还是要准备好的。

• 疾病样本：其实单细胞样本的分析本质还是分组找差异，只是这个差异从更高的维度来分析，无论是通讯的差异还是转录状态的差异，前提是有组可分，最理想的状态就是normal和disease，当然很多老师无法取到normal的样本，那么临床信息就显得尤为重要，预后好和预后差之间的差异，才具有临床指导的价值。

• 如果说仅仅有疾病样本，临床信息也没有的前提下，是很难分析出有效的信息，这个时候很多老师就开始探索单细胞数据库，从数据库的队列中寻找normal的数据来进行分析，方法是可取的，但是要注意数据的匹配程度，平台差异等等。

• 样本的注释问题，流程化的注释是不可取的，多次强调过，市面上都说新格元注释做的好，也是浪费了很大的人工，并不是流程化带来的，资料和经验，也是科研工作者必备的素质。

• 样本量的问题，单细胞发展到现在，仅靠一两个样本发文章是不现实的（二区及以上），无论从哪个角度分析，都需要在多个样本中验证研究的价值，所以大家如果决定做单细胞分析，就要想好这一点。

好了，关于课题设计，其实也是一个很大的学问，这样的工作通常是技术支持来做的，接下来分享一个简单的内容，衡量样本间的异质性。

• 指标1：Cluster entropy，为了测量来自不同样本的细胞在细胞类型cluster中的混合程度如何，量化了数据集的归一化相对cluster entropy，按cluster大小加权。 cluster entropy为 1 表示样本在cluster 之间完全混合。

图片.png

参考文章，10X单细胞（10X空间转录组）基因表达的熵值分析

图片.png

• 指标2 ：Similarity scores/alignment，为了测量来自相同与不同批次和/或样品的细胞之间细胞类型内细胞状态的转录变异，测量了每个样品/批次之间的成对对齐（就是整合），其中批次由同一天处理的样品组组成。 这个“相似度分数”检查特定样本/批次中每个细胞的局部邻域，询问其 k 个最近邻居中有多少个（在 PC 或 iNMF（这个大家应该都知道） 空间中）属于第二个样本/批次，然后在所有细胞上取平均值 . 这里选择 k 为cluster内细胞总数的 1%。 结果通过每个样品/批次的预期细胞数进行标准化。

最后汇总一下Seurat包的所有函数

函数	作用
AddModuleScore	Calculate module scores for feature expression programs in single cells
AggregateExpression	Aggregated feature expression by identity class
AnchorSet-class	The AnchorSet Class
AnnotateAnchors	Add info to anchor matrix
Assay-class	The Assay Class
AugmentPlot	Augments ggplot2-based plot with a PNG image.
AverageExpression	Averaged feature expression by identity class
BGTextColor	Determine text color based on background color
BarcodeInflectionsPlot	Plot the Barcode Distribution and Calculated Inflection Points
BlackAndWhite	Create a custom color palette
BuildClusterTree	Phylogenetic Analysis of Identity Classes
CalcPerturbSig	Calculate a perturbation Signature
CalculateBarcodeInflections	Calculate the Barcode Distribution Inflection
CaseMatch	Match the case of character vectors
CellCycleScoring	Score cell cycle phases
CellScatter	Cell-cell scatter plot
CellSelector	Cell Selector
Cells.SCTModel	Get Cell Names
CellsByImage	Get a vector of cell names associated with an image (or set of images)
CollapseEmbeddingOutliers	Move outliers towards center on dimension reduction plot
CollapseSpeciesExpressionMatrix	Slim down a multi-species expression matrix,when only one species is primarily of interenst.
ColorDimSplit	Color dimensional reduction plot by tree split
CombinePlots	Combine ggplot2-based plots into a single plot
CreateSCTAssayObject	Create a SCT Assay object
CustomDistance	Run a custom distance function on an input data matrix
DEenrichRPlot	DE and EnrichR pathway visualization barplot
DietSeurat	Slim down a Seurat object
DimHeatmap	Dimensional reduction heatmap
DimPlot	Dimensional reduction plot
DimReduc-class	The DimReduc Class
DiscretePalette	Discrete colour palettes from the pals package
DoHeatmap	Feature expression heatmap
DotPlot	Dot plot visualization
ElbowPlot	Quickly Pick Relevant Dimensions
ExpMean	Calculate the mean of logged values
ExpSD	Calculate the standard deviation of logged values
ExpVar	Calculate the variance of logged values
FastRowScale	Scale and/or center matrix rowwise
FeaturePlot	Visualize 'features' on a dimensional reduction plot
FeatureScatter	Scatter plot of single cell data
FilterSlideSeq	Filter stray beads from Slide-seq puck
FindAllMarkers	Gene expression markers for all identity classes
FindClusters	Cluster Determination
FindConservedMarkers	Finds markers that are conserved between the groups
FindIntegrationAnchors	Find integration anchors
FindMarkers	Gene expression markers of identity classes
FindMultiModalNeighbors	Construct weighted nearest neighbor graph
FindNeighbors	(Shared) Nearest-neighbor graph construction
FindSpatiallyVariableFeatures	Find spatially variable features
FindSubCluster	Find subclusters under one cluster
FindTransferAnchors	Find transfer anchors
FindVariableFeatures	Find variable features
FoldChange	Fold Change
GetAssay	Get an Assay object from a given Seurat object.
GetImage.SlideSeq	Get Image Data
GetIntegrationData	Get integration data
GetResidual	Calculate pearson residuals of features not in the scale.data
GetTissueCoordinates.SlideSeq	Get Tissue Coordinates
GetTransferPredictions	Get the predicted identity
Graph-class	The Graph Class
GroupCorrelation	Compute the correlation of features broken down by groups with another covariate
GroupCorrelationPlot	Boxplot of correlation of a variable (e.g.number of UMIs) with expression data
HTODemux	Demultiplex samples based on data from cell 'hashing'
HTOHeatmap	Hashtag oligo heatmap
HVFInfo.SCTAssay	Get Variable Feature Information
HoverLocator	Hover Locator
IFeaturePlot	Visualize features in dimensional reduction space interactively
ISpatialDimPlot	Visualize clusters spatially and interactively
ISpatialFeaturePlot	Visualize features spatially and interactively
IntegrateData	Integrate data
IntegrateEmbeddings	Integrate low dimensional embeddings
IntegrationAnchorSet-class	The IntegrationAnchorSet Class
IntegrationData-class	The IntegrationData Class
JackStraw	Determine statistical significance of PCA scores.
JackStrawData-class	The JackStrawData Class
JackStrawPlot	JackStraw Plot
L2CCA	L2-Normalize CCA
L2Dim	L2-normalization
LabelClusters	Label clusters on a ggplot2-based scatter plot
LabelPoints	Add text labels to a ggplot2 plot
LinkedPlots	Visualize spatial and clustering (dimensional reduction) data in a linked, interactive framework
Load10X_Spatial	Load a 10x Genomics Visium Spatial Experiment into a 'Seurat' object
LoadAnnoyIndex	Load the Annoy index file
LoadSTARmap	Load STARmap data
LocalStruct	Calculate the local structure preservation metric
LogNormalize	Normalize raw data
LogVMR	Calculate the variance to mean ratio of logged values
MULTIseqDemux	Demultiplex samples based on classification method from MULTI-seq
MapQuery	Map query cells to a reference
MappingScore	Metric for evaluating mapping success
MetaFeature	Aggregate expression of multiple features into a single feature
MinMax	Apply a ceiling and floor to all values in a matrix
MixingMetric	Calculates a mixing metric
MixscapeHeatmap	Differential expression heatmap for mixscape
MixscapeLDA	Linear discriminant analysis on pooled CRISPR screen data.
ModalityWeights-class	The ModalityWeights Class
NNPlot	Highlight Neighbors in DimPlot
Neighbor-class	The Neighbor Class
NormalizeData	Normalize Data
PCASigGenes	Significant genes from a PCA
PercentageFeatureSet	Calculate the percentage of all counts that belong to a given set of features
PlotClusterTree	Plot clusters as a tree
PlotPerturbScore	Function to plot perturbation score distributions.
PolyDimPlot	Polygon DimPlot
PolyFeaturePlot	Polygon FeaturePlot
PredictAssay	Predict value from nearest neighbors
PrepLDA	Function to prepare data for Linear Discriminant Analysis.
PrepSCTIntegration	Prepare an object list normalized with sctransform for integration.
ProjectDim	Project Dimensional reduction onto full dataset
ProjectUMAP	Project query into UMAP coordinates of reference
Radius.SlideSeq	Get Spot Radius
Read10X	Load in data from 10X
Read10X_Image	Load a 10X Genomics Visium Image
Read10X_h5	Read 10X hdf5 file
ReadMtx	Load in data from remote or local mtx files
ReadSlideSeq	Load Slide-seq spatial data
RegroupIdents	Regroup idents based on meta.data info
RelativeCounts	Normalize raw data to fractions
RenameCells.SCTAssay	Rename Cells in an Object
RidgePlot	Single cell ridge plot
RunCCA	Perform Canonical Correlation Analysis
RunICA	Run Independent Component Analysis on gene expression
RunLDA	Run Linear Discriminant Analysis
RunMarkVario	Run the mark variogram computation on a given position matrix and expression matrix.
RunMixscape	Run Mixscape
RunMoransI	Compute Moran's I value.
RunPCA	Run Principal Component Analysis
RunSPCA	Run Supervised Principal Component Analysis
RunTSNE	Run t-distributed Stochastic Neighbor Embedding
RunUMAP	Run UMAP
SCTAssay-class	The SCTModel Class
SCTResults	Get SCT results from an Assay
SCTransform	Use regularized negative binomial regression to normalize UMI count data
STARmap-class	The STARmap class
SampleUMI	Sample UMI
SaveAnnoyIndex	Save the Annoy index
ScaleData	Scale and center the data.
ScaleFactors	Get image scale factors
ScoreJackStraw	Compute Jackstraw scores significance.
SelectIntegrationFeatures	Select integration features
SetIntegrationData	Set integration data
Seurat-class	The Seurat Class
Seurat-package	Seurat: Tools for Single Cell Genomics
SeuratCommand-class	The SeuratCommand Class
SeuratTheme	Seurat Themes
SlideSeq-class	The SlideSeq class
SpatialImage-class	The SpatialImage Class
SpatialPlot	Visualize spatial clustering and expression data.
SplitObject	Splits object into a list of subsetted objects.
SubsetByBarcodeInflections	Subset a Seurat Object based on the Barcode Distribution Inflection Points
TopCells	Find cells with highest scores for a given dimensional reduction technique
TopFeatures	Find features with highest scores for a given dimensional reduction technique
TopNeighbors	Get nearest neighbors for given cell
TransferAnchorSet-class	The TransferAnchorSet Class
TransferData	Transfer data
UpdateSCTAssays	Update pre-V4 Assays generated with SCTransform in the Seurat to the new SCTAssay class
UpdateSymbolList	Get updated synonyms for gene symbols
VariableFeaturePlot	View variable features
VisiumV1-class	The VisiumV1 class
VizDimLoadings	Visualize Dimensional Reduction genes
VlnPlot	Single cell violin plot
as.CellDataSet	Convert objects to CellDataSet objects
as.Seurat.CellDataSet	Convert objects to 'Seurat' objects
as.SingleCellExperiment	Convert objects to SingleCellExperiment objects
as.sparse.H5Group	Cast to Sparse
cc.genes	Cell cycle genes
cc.genes.updated.2019	Cell cycle genes: 2019 update
contrast-theory	Get the intensity and/or luminance of a color
merge.SCTAssay	Merge SCTAssay objects
subset.AnchorSet	Subset an AnchorSet object

周一了，偷个懒，生活很好，有你更好