單細(xì)胞數(shù)據(jù)整合-1:Harmony原理介紹和官網(wǎng)教程

目錄:
1. 原理介紹
2. 操作演示
3. 關(guān)于Harmony操作是否會(huì)對(duì)差異分析產(chǎn)生影響

1. 原理介紹

官網(wǎng):https://github.com/immunogenomics/harmony
(Harmony必須在R版本3.4以上運(yùn)行,支持 Linux, OS X, and Windows 平臺(tái)。)
文章:https://www.biorxiv.org/content/early/2018/11/04/461954
harmony算法與其他整合算法相比的優(yōu)勢(shì):

(1)整合數(shù)據(jù)的同時(shí)對(duì)稀有細(xì)胞的敏感性依然很好;
(2)省內(nèi)存;
(3)適合于更復(fù)雜的單細(xì)胞分析實(shí)驗(yàn)設(shè)計(jì),可以比較來自不同供體,組織和技術(shù)平臺(tái)的細(xì)胞。

基本原理:我們用不同顏色表示不同數(shù)據(jù)集,用形狀表示不同的細(xì)胞類型。首先,Harmony應(yīng)用主成分分析(一文看懂PCA主成分分析)將轉(zhuǎn)錄組表達(dá)譜嵌入到低維空間中,然后應(yīng)用迭代過程去除數(shù)據(jù)集特有的影響。

(A)Harmony概率性地將細(xì)胞分配給cluster,從而使每個(gè)cluster內(nèi)數(shù)據(jù)集的多樣性最大化。
(B)Harmony計(jì)算每個(gè)cluster的所有數(shù)據(jù)集的全局中心,以及特定數(shù)據(jù)集的中心。
(C)在每個(gè)cluster中,Harmony基于中心為每個(gè)數(shù)據(jù)集計(jì)算校正因子。
(D)最后,Harmony使用基于C的特定于細(xì)胞的因子校正每個(gè)細(xì)胞。由于Harmony使用軟聚類,因此可以通過多個(gè)因子的線性組合對(duì)其A中進(jìn)行的軟聚類分配進(jìn)行線性校正,來修正每個(gè)單細(xì)胞。
重復(fù)步驟A到D,直到收斂為止。聚類分配和數(shù)據(jù)集之間的依賴性隨著每一輪的減少而減小。

2. 操作演示

R包安裝

library(devtools)
install_github("immunogenomics/harmony")

安裝過程可能包括從源代碼編譯C++代碼,因此可能需要幾分鐘。

下載稀疏矩陣示例(https://www.dropbox.com/s/t06tptwbyn7arb6/pbmc_stim.RData?dl=1)

library(Seurat)
library(cowplot)
library(harmony)
load('data/pbmc_stim.RData') #加載矩陣數(shù)據(jù)
#在運(yùn)行Harmony之前,創(chuàng)建一個(gè)Seurat對(duì)象并按照標(biāo)準(zhǔn)PCA進(jìn)行分析。
pbmc <- CreateSeuratObject(counts = cbind(stim.sparse, ctrl.sparse), project = "PBMC", min.cells = 5) %>%
    Seurat::NormalizeData(verbose = FALSE) %>%
    FindVariableFeatures(selection.method = "vst", nfeatures = 2000) %>%
    ScaleData(verbose = FALSE) %>%
    RunPCA(pc.genes = pbmc@var.genes, npcs = 20, verbose = FALSE) #R語言中%>%的含義是什么呢,管道函數(shù)啦,就是把左件的值發(fā)送給右件的表達(dá)式,并作為右件表達(dá)式函數(shù)的第一個(gè)參數(shù)。
pbmc@meta.data$stim <- c(rep("STIM", ncol(stim.sparse)), rep("CTRL", ncol(ctrl.sparse)))#賦值條件變量

未經(jīng)校正的PC中的數(shù)據(jù)集之間存在明顯差異:

options(repr.plot.height = 5, repr.plot.width = 12)
p1 <- DimPlot(object = pbmc, reduction = "pca", pt.size = .1, group.by = "stim")
p2 <- VlnPlot(object = pbmc, features = "PC_1", group.by = "stim", pt.size = .1)
plot_grid(p1,p2)

Run Harmony

運(yùn)行Harmony的最簡(jiǎn)單方法是傳遞Seurat對(duì)象并指定要集成的變量。RunHarmony返回Seurat對(duì)象,并使用更正后的Harmony坐標(biāo)(使用Harmony代替PCA)。將plot_convergence設(shè)置為TRUE,這樣我們就可以確保Harmony目標(biāo)函數(shù)在每一輪中都變得更好。

RunHarmony函數(shù)中主要參數(shù):

  • group.by.vars參數(shù)是設(shè)置按哪個(gè)分組來整合
  • max.iter.harmony設(shè)置迭代次數(shù),默認(rèn)是10。運(yùn)行RunHarmony結(jié)果會(huì)提示在迭代多少次后完成了收斂。
  • ??lambda參數(shù),默認(rèn)值是1,決定了Harmony整合的力度。lambda值調(diào)小,整合力度變大,反之。(只有這個(gè)參數(shù)影響整合力度,調(diào)整范圍一般在0.5-2之間)
  • ??theta參數(shù):Diversity clustering penalty parameter. Specify for each variable in group.by.vars. Default theta=2. theta=0 does not encourage any diversity. Larger values of theta result in more diverse clusters. 這個(gè)參數(shù)我常用默認(rèn)值,但在不同文獻(xiàn)中這個(gè)參數(shù)往往不同。
  • ??dims.use參數(shù):Which PCA dimensions to use for Harmony. By default, use all.
  • sigma參數(shù):Width of soft kmeans clusters. Default sigma=0.1. Sigma scales the distance from a cell to cluster centroids. Larger values of sigma result in cells assigned to more clusters. Smaller values of sigma make soft kmeans cluster approach hard clustering.
options(repr.plot.height = 2.5, repr.plot.width = 6)
pbmc <- pbmc %>%
RunHarmony("stim", plot_convergence = TRUE) #Harmony converged after 8 iterations

Harmory運(yùn)行后的結(jié)果儲(chǔ)存在:

pbmc@reductions$harmony

使用Embeddings命令訪問新的Harmony embeddings。

harmony_embeddings <- Embeddings(pbmc, 'harmony')
harmony_embeddings[1:5, 1:5]

讓我們查看確認(rèn)數(shù)據(jù)集在Harmony運(yùn)行之后的前兩個(gè)維度中得到很好的整合。

options(repr.plot.height = 5, repr.plot.width = 12)
p1 <- DimPlot(object = pbmc, reduction = "harmony", pt.size = .1, group.by = "stim")
p2 <- VlnPlot(object = pbmc, features = "harmony_1", group.by = "stim", pt.size = .1)
plot_grid(p1,p2)

Downstream analysis

許多下游分析是在低維嵌入而不是基因表達(dá)上進(jìn)行的。要使用校正后的Harmony embeddings而不是PC,設(shè)置reduction ='harmony'。

pbmc <- pbmc %>%
    RunUMAP(reduction = "harmony", dims = 1:20) %>%
    FindNeighbors(reduction = "harmony", dims = 1:20) %>%
    FindClusters(resolution = 0.5) %>%
    identity()

在UMAP embedding中,我們可以看到更復(fù)雜的結(jié)構(gòu)。由于我們使用harmony embeddings,因此UMAP embeddings混合得很好。

options(repr.plot.height = 4, repr.plot.width = 10)
DimPlot(pbmc, reduction = "umap", group.by = "stim", pt.size = .1, split.by = 'stim')

TSNE分析

pbmc=RunTSNE(pbmc,reduction = "harmony", dims = 1:20)
TSNEPlot(object = pbmc, pt.size = 0.5, label = TRUE,split.by='stim') 

兩樣本合并的TSNE和UMAP圖

DimPlot(pbmc, reduction = "umap",pt.size = .1,  label = TRUE)
TSNEPlot(pbmc, pt.size = .1, label = TRUE) 

隨后就可以尋找差異表達(dá)基因并對(duì)細(xì)胞進(jìn)行注釋。

3. 關(guān)于Harmony操作是否會(huì)對(duì)差異分析產(chǎn)生影響

Harmony輸入的是scRNA@reductions$pca的數(shù)據(jù),得出的結(jié)果儲(chǔ)存在scRNA@reductions$harmony中。

而差異分析使用的是scRNA@assays$RNA@counts數(shù)據(jù),互不影響。

4. 多樣本批次矯正方法匯總

工具 Batch-effect-corrected output 方法
Seurat2 R Normalized canonical components Canonical correlation analysis and dynamic time warping
Seurat3 R Normalized gene expression matrix CCA and mutural nearest neighbors-anchors
Harmony R Normalized feature reduction vectors Iterative clustering in dimensionally reduced space
MNN Correct R Normalized gene expression matrix Mutual nearest neighbor in gene expression space
fastMNN R Normalized principal components MNN in dimensionally reduced space
ComBat R Normalized gene expression matrix Adjusts for known batches using an empirical Bayesian framework
limma R Normalized gene expression matrix Linear model/empirical Bayes model
scGen R Normalized gene expression matrix Variational auto-encoders neural network model and latent space
Scanorama R/P Normalized gene expression matrix Mutual nearest neighbor and panoramic stitching
MND-ResNet P Normalized principal components Residual neural network for calibration
ZINB-WaVE R Normalized gene expression matrix Zero-inflated negative binomial model, extension of RUV model
scMerge R Normalized gene expression matrix Stably expressed genes (scSEGs) and RUVIII model
LIGER R Normalized feature reduction vectors Integrative non-negative matrix factorization (iNMF) and joint clustering + quantile alignment
BBKNN P Connectivity graph and normalized dimension reduction vectors (UMAP) Batch balanced k-nearest neighbors
最后編輯于
?著作權(quán)歸作者所有,轉(zhuǎn)載或內(nèi)容合作請(qǐng)聯(lián)系作者
【社區(qū)內(nèi)容提示】社區(qū)部分內(nèi)容疑似由AI輔助生成,瀏覽時(shí)請(qǐng)結(jié)合常識(shí)與多方信息審慎甄別。
平臺(tái)聲明:文章內(nèi)容(如有圖片或視頻亦包括在內(nèi))由作者上傳并發(fā)布,文章內(nèi)容僅代表作者本人觀點(diǎn),簡(jiǎn)書系信息發(fā)布平臺(tái),僅提供信息存儲(chǔ)服務(wù)。
禁止轉(zhuǎn)載,如需轉(zhuǎn)載請(qǐng)通過簡(jiǎn)信或評(píng)論聯(lián)系作者。

相關(guān)閱讀更多精彩內(nèi)容

友情鏈接更多精彩內(nèi)容