Ulisse workflow 2: Cell-cell cross-talk analysis on single cell data

In this tutorial we will see how to analyse intercelluar cross-talk by using Ulisse. For this analysis we need two inputs: ranked gene sets of the cell cluster and a communication network.

library(Seurat)
library(OmnipathR)
library(igraph)
library(Ulisse)
library(circlize)
library(ComplexHeatmap)
library(scales)

Preparation of the inputs

Firstly we will see how to prepare the inputs by using Ulisse’s functions. The cell gene sets derive from the cell type or cell clusters of a single cell sample (or multiple integrated samples), while the intercellular network has to be provided.Here, we will use Omnipath communication network (Türei et al. 2021), but other resources can be used. Only take care that Ulisse uses undirected biological network and does not considers sub-units.

Ligand-receptor network

With the code below we will download Omnipath intercellualr communications, that are used to build an intercellualr network by using igraph package functions.

intercell_net <- import_intercell_network(interactions_param = list("datasets" = "omnipath"),
                                          transmitter_param = list(
                                            categories = c('ligand')),
                                          receiver_param = list(
                                            categories =c('receptor')), 
                                          entity_types = "protein")
g.intercell <- unique(intercell_net[,c("source_genesymbol", "target_genesymbol")])
g.intercell <- graph_from_edgelist(as.matrix(g.intercell), 
                                   directed = F)

Pre-processing with Seurat pipeline

To explain the functioning of the intercellular cross-talk calculation implemented in Ulisse we will use one of the samples sequenced by Wu et al. (Wu et al. 2021). In this study, the authors analyse 26 single cell of breast cancer samples, that are composed by tumoral and normal cells composing its microenvironment. The data can be downloaded at https://singlecell.broadinstitute.org/single_cell/study/SCP1039 or at GEO website under accession GSE176078. In particular, we will study intercellular cross-talk in sample CID4515. The code below can be used to read the 10x data and pre-process them to obtain dimensionality reduction through Seurat v4 package (Hao et al. 2021). We will study the cell-cell communication among the cell-types provided by the authors, but the pipeline can be applied also to the clusters.

data <- ReadMtx(mtx = "CID4515/count_matrix_sparse.mtx", 
                cells = "CID4515/count_matrix_barcodes.tsv",
                features = "CID4515/count_matrix_genes.tsv" )
data <- data[which(rowSums(sign(data)) >= 5),]
data <- CreateSeuratObject(counts = data, min.cells = 0, min.features = 0)
data <- NormalizeData(data)
data <- ScaleData(data)
data <- RunPCA(data)
data <- RunUMAP(data, dims = 1:10)

metadata.csv is the file with the cell-types identification provided by the authors (that can be downloaded again at https://singlecell.broadinstitute.org/single_cell/study/SCP1039). After uploading it, we can use this information to plot the UMAP by cell type.

meta <- read.csv("CID44991/metadata.csv", sep = ",", stringsAsFactors = F)
meta[1:5,]

X	orig.ident	nCount_RNA	nFeature_RNA	percent.mito	subtype	celltype_subset	celltype_minor	celltype_major
CID44991_AAACGGGTCCGCATAA	CID44991	2648	1153	1.435045	TNBC	Endothelial ACKR1	Endothelial ACKR1	Endothelial
CID44991_AACCATGAGCTACCTA	CID44991	4554	1795	9.244620	TNBC	Endothelial ACKR1	Endothelial ACKR1	Endothelial
CID44991_CAGATCACAAGGTGTG	CID44991	2001	1032	12.043978	TNBC	Endothelial ACKR1	Endothelial ACKR1	Endothelial
CID44991_CATATTCGTGCACTTA	CID44991	3293	1442	9.626480	TNBC	Endothelial ACKR1	Endothelial ACKR1	Endothelial
CID44991_CCTTTCTCACAGTCGC	CID44991	4452	1884	4.784367	TNBC	Endothelial ACKR1	Endothelial ACKR1	Endothelial

data@meta.data <- cbind(data@meta.data, meta[match(rownames(data@meta.data), meta$X),])
data$celltype_major <- factor(data$celltype_major, levels = unique(data$celltype_major))

pal_ct <- viridis::turbo(9)
names(pal_ct) <- sort(unique(data$celltype_major))

DimPlot(data, reduction = "umap", group.by = "celltype_major", label = T,  cols = pal_ct)

Intercellular cross-talk analysis

Ulisse provides two methods to study intercellular cross-talk, one that uses expression data and the other that uses Differentially Expressed Genes (DEG) data. We exemplify the usage of both approaches, to then compare the obtained results.

Expression data

In this case, cell-type marker genes to be sued for cell-type gene sets are obtained from the normalized count matrix. preparing_cl_list() takes as an inputs:

the normalized gene counts matrix (mtx);
the cell type affiliation of each cell in the count matrix (clusters);
the gene of interest (universe);
two threshold (mean_t and cell_t) to filter the data and remove noise.

The function uses mean_t to identify expressed genes in the normalized expression matrix: if the gene in a cell has a value equal or higher of that threshold is set to 1, 0 otherwise. Subsequently, the function uses the binarized expression matrix to calculate for each gene in each cell-type the frequency of being expressed in at least cell_t cells. Only the genes in universe are considered, which are the genes present in the LR network.

We suggest identifying mean_t considering the mean expression of each gene calculated on non-zero values. Instead, cell_t can be set considering the filtering used to remove low expressed genes at the beginning of the Seurat pipeline. The code below can be used to assess the expression threshold.

mean_norm <- data@assays$RNA@data
mean_norm[which(mean_norm == 0)] <- NA
mean_g <- rowMeans(mean_norm, na.rm = T)

summary(mean_g)

#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>  0.1906  0.5091  0.6242  0.7346  0.8295  5.4092

hist(mean_g, breaks = 100)
abline(v=0.65, col = "red")

Considering the summary of mean_g and the histogram above, we decide to use 0.65 as mean_t (the red line in the histogram). We decided this value as it lays past the high frequency values in the distribution and thus allows to filter out the low and the most common values, maintaining only the high ones. As cell threshold we will use the same threshold used for the pre-processing.

universe <- V(g.intercell)$name
ct_list <- preparing_expr_list(mtx = data@assays$RNA@data, 
                               clusters = data$celltype_major, 
                               mean_t = 0.65, 
                               cell_t = 5,
                               universe = universe)
head(ct_list[[1]])

#>   TNFRSF4  TNFRSF14      ENO1    AGTRAP  TNFRSF1B     EPHA2 
#> 0.2377049 0.2049180 0.7704918 0.1885246 0.2459016 0.1147541

Communication analysis

The required inputs are now ready to analyse intercellular cross-talk with gs_cross_talk() function. This function requires also additional parameter:

k corresponds to the number of permutation needed that, together with the original matrix, are used for the calculation of p-value and FDR. In this case we use 49 permutation that, together with the real one, correspond to a total of 50 permuted list for each cell type pair, and thus the minimal p-value that can be observed is 1/50 or 0.02;
shared which should be set to TRUE to enable possible overlapping in gene-lists of a cell-type pair. Different cells may express the same genes, moreover when considering clusters, which may represents different states of the same cell-type. These shared genes may represents intracellular communication, but as both cells in a cell-pair express the same genes, is impossible to discern if these are autocrine or paracrine communication. Thus, it is important to consider also these in intercellular cross-talk;
hash: logical, used to speed-up calculation when lots of gene-sets are used. Here, we have low cell-type numbers, so we suggest to set it to FALSE;
ct_info: logical, if detail of gene-gene interaction in the gene-set should be returned. This might be highly important in cell-cell cross-talk, so we set it to TRUE. By enabling this parameter, the output of the function will be a two-element list, with the fist (ct_info) a table with all the gene communications between the two cell types; and the second (ct_res) a table with the intercellular cross-talk results. Note that the cumulative score of a cell type pair is equal to the sum of all the gene pair scores in ct _info.
mc_cores_perm and mc_cores_ct control the parallelization of the function in intercellular cross-talk calculation and permutation, respectively. Be aware that mc_cores_perm multiplies mc_cores_ct. Parallelization is useful to reduce the computational time for the calculation but improves the amount of memory needed.

LR.adj <- as_adjacency_matrix(g.intercell, sparse = F)

expr_ccc <- gs_cross_talk(gs_list = cl_list,
                          gene_network_adj = LR.adj,
                          k = 49, 
                          shared = T, 
                          ct_info = T, 
                          hash = F,
                          mc_cores_perm = 1, mc_cores_ct = 1)

expr_ccc$communications_info[1:5,]

gs1	gs1_gene	gs2	gs2_gene	score
Endothelial	MIF	CAFs	TNFRSF14	0.0337512
Endothelial	TNFSF12	CAFs	TNFRSF8	0.0021916
Endothelial	BIRC2	CAFs	TNFRSF1B	0.0059174
Endothelial	TNFRSF1A	CAFs	TNFRSF1B	0.0170948
Endothelial	GRN	CAFs	TNFRSF1B	0.0175331

expr_ccc$cc_communications[1:5,]

gs1	gs2	ct_score	ngenes_gs1	ngenes_gs2	nlink	p_value_link	FDR_link	p_adj_BH	weight_gs1	weight_gs2	genes_gs1	genes_gs2
Endothelial	CAFs	125.23126	399	443	1334	0.02	0.1266667	0.0205714	110.13115	119.11230	MIF;TNFSF12;BIRC2;TNFRSF1A;GRN;EFNA1;TP53;NOTCH1;ITGAV;ITGB1;ITGA5;EDN1;FYN;EFNB1;EFNB2;FGF2;MDK;ANGPTL4;STX3;STX4;NID1;LEPR;IL7R;LIFR;IL6ST;IL15RA;PTPN11;IL4R;TYK2;IFNAR2;JAK1;APOD;JAK2;SOCS2;STAT3;TGFB3;TGFB1;IL6;ITGA9;EZR;ICAM1;SLC7A1;SIRPA;LRP5;FZD4;LRP6;MAGED1;LGALS3BP;DLL1;JAG2;DLL4;JAG1;ITGB4;HLA-DRA;HLA-DRB5;HLA-DRB1;HLA-DQA1;HLA-DQB1;HLA-DQA2;HLA-DMB;HLA-DMA;HLA-DPA1;HLA-DPB1;ARF6;CD36;TLR4;MDM2;BTNL9;IL18BP;ABL1;MAPK8;LGALS3;CD74;P2RY6;TEK;ITGA6;DAG1;ITGA1;ITGA2;CD44;ITGA3;ADRB2;LDLR;S100A10;SELP;SELL;SEMA6B;PTPRF;TNC;THBS1;LAMA5;SNCA;LMAN1;ACVR1B;IL1RAP;IL33;CXCL12;INHBB;BMP6;BMP4;CXCL2;CXCL10;CXCL11;PAM;ADA;LAMC1;SPP1;LAMA4;LAMB1;MFGE8;CX3CL1;CALR;PDGFB;TNFSF10;BMP1;SEMA3F;SEMA3G;RARRES1;VEGFC;FLT4;PGF;ADM;ALCAM;GDF15;VCAM1;VWF;LAMB2;PRND;DUSP18;CCL2;CCL14;CCL23;RPSA;LAMA3;PLXNA2;NRP2;NRP1;RYK;PTPRK;FZD6;MCAM;NLK;MET;PTPRG;LGALS9;DHH;DIP2A;TLN1;LTBP3;VCL;CASP8;RIPK1;TNFRSF10B;TNFRSF10D;TNFRSF10A;TNFRSF11B;IL1R1;ROBO1;PLXNA1;ROBO4;APP;ADAM23;PDGFA;PDGFD;SDC3;ACKR1;KDR;SDC2;PLCE1;PRKACA;ACKR3;S1PR1;PTGER4;FGFR1;IL2RG;FLT1;LYVE1;VCAN;FAM3C;SLC6A8;BMPR2;CD47;EPHA2;EPHA4;CXCR4;TGFBR3;COPA;APLP2;NOTCH4;CD81;HLA-C;CTSS;CTSL;CTSF;CD63;LGMN;TNFRSF21;CIRBP;POSTN;NFE2L2;ADAM15;CD55;SOS1;PRKCE;SCN9A;TUBA4A;ITPR1;RAF1;PIK3CA;LPP;TNK2;CASP3;PIK3R1;FLOT1;MAPK14;CDK6;RPS6KA3;PTK2;THY1;SLK;PTPRE;TUBA1A;CDK2;CSK;LAT;PLCG2;PLD2;CD79B;PECAM1;PLCG1;BSG;STK11;PLAUR;CSF2RB;IGF2;FURIN;HRH1;ACVR2A;ACVR1;ACVR2B;BAMBI;S100A4;CMTM8;ARF4;AREG;HBEGF;ANXA1;DCN;ANXA2;S100A9;S100A8;SAA1;PLXND1;PLAU;PCNA;CD99;RARA;LAMP1;NCL;PTPRB;PLXNB2;CCRL2;PILRB;CD200;EPHB4;APLNR;TIE1;FGF12;SPARC;FGF18;RXRA;HMGB1;APOE;CD40;CAPN1;TGS1;MFAP2;YBX1;CHUK;IGF2R;INSR;CALCRL;RAMP3;RAMP2;SCARB1;RSPO3;STAB1;TXLNA;MMRN2;TNFRSF1B;HLA-A;IL15;SERPINF1;NPNT;ANGPTL2;JAM3;LUM;LGALS1;VEGFB;CKAP4;FLRT2;POMC;APLN;IKBKB;NSMAF;HLA-E;HLA-B;CKLF;ACTR2;RARRES2;NRAS;MPZL1;IFNGR1;MAP3K4;KRAS;MAPK3;CSF1;LY96;RHBDL2;TFRC;PIP5K1A;SMAP1;PIP4K2A;PIP5K1C;ACVRL1;MAPK1;TGFBR2;UNC5B;MMP2;PLAT;IL13RA1;STAT6;PLXDC2;TNFRSF12A;PLK3;MCL1;RGS16;PTGS2;ODC1;EIF2AK2;CASP10;CASP6;PLK2;NR3C1;CSNK2B;HSPA1B;TAP1;HSP90AB1;ABCB1;HIPK2;NOS3;RRM2B;CTSD;BIRC3;CASP1;ATM;HSPA8;PTEN;SLC38A2;PRKAB1;LATS2;BDKRB2;VRK1;ULK2;PPM1D;BCL2L1;DNMT1;MMP14;ETV5;KMT2E;PTHLH;ZFP91;VANGL1;AVPR2;MCFD2;CD46;GPRC5B;TNFRSF14;LTBR;GPIHBP1;CDH11;CDH5;PTAFR;PRMT5;PTPN1;PRKD2;IFNAR1;ARPC5;TAP2;PDIA3;SCARF1;GAS6;FSTL1;CD109;LSR;ITGA10;GMFB;ICAM2;ICAM3	TNFRSF14;TNFRSF8;TNFRSF1B;EPHA2;MFAP2;PLA2G2A;ECE1;EPHB2;SDC3;COL16A1;TXLNA;YBX1;PTPRF;PLK3;JAK1;LEPR;NEGR1;TGFBR3;F3;VCAM1;CSF1;WNT2B;NRAS;NGF;VANGL1;NOTCH2;ECM1;MCL1;CTSS;PIP5K1A;S100A9;S100A4;IL6R;ADAM15;MUC1;COPA;MPZL1;ANGPTL1;RGS16;LAMC1;ARPC5;PRG4;PTGS2;PLA2G4A;CD55;CD46;CD34;PLXNA2;WNT9A;NID1;LGALS8;ODC1;SDC1;EIF2AK2;SOS1;PRKCE;MCFD2;ACTR2;IL1R1;IL1RL1;CXCR4;ACVR2A;ACVR1;DPP4;NFE2L2;ITGAV;CASP8;BMPR2;NRP2;TUBA4A;NCL;ACKR3;COL6A3;RAMP1;CHL1;ITPR1;HRH1;RAF1;KAT2B;TGFBR2;ITGA9;RPSA;CCR1;LAMB2;DAG1;SEMA3B;PRKCD;WNT5A;LRIG1;CNTN3;ALCAM;CD47;BOC;FSTL1;ITGB5;PLXNA1;PODXL2;ACKR4;RARRES1;TNFSF10;PIK3CA;BCL6;LPP;IL1RAP;APOD;TNK2;FGFRL1;SPON2;SLIT2;LGI2;PDGFRA;CXCL8;CXCL6;CXCL1;CXCL3;CXCL2;CXCL9;CXCL10;CXCL11;CXCL13;SPP1;CASP6;FGF2;IL15;EDNRA;PDGFC;VEGFC;CASP3;TLR3;SEMA5A;LIFR;PTGER4;FGF10;ITGA1;ITGA2;FST;IL6ST;PLK2;PIK3R1;THBS4;VCAN;EDIL3;RGMB;EFNA5;HSPA4;CXCL14;HBEGF;FGF1;NR3C1;PDGFRB;CD74;SPARC;SLIT3;RIPK1;HLA-A;FLOT1;HLA-C;CSNK2B;HSPA1B;NOTCH4;HLA-DRA;HLA-DRB5;HLA-DRB1;HLA-DQA1;HLA-DQB1;TAP2;TAP1;HLA-DMA;HLA-DPA1;HLA-DPB1;MAPK14;PI16;TREM1;PTK7;VEGFA;HSP90AB1;SMAP1;CD109;FYN;LAMA4;RSPO3;PTPRK;LAMA2;MAP3K5;IFNGR1;EZR;IGF2R;MAP3K4;THBS2;DLL1;PDGFA;ITGB8;IL6;INHBA;EGFR;CD36;SEMA3C;HGF;SEMA3D;CDK6;PILRB;KMT2E;LAMB1;WNT2;FAM3C;PLXNA4;PTN;HIPK2;RARRES2;CD99;MXRA5;RPS6KA3;AR;EFNB1;IL13RA1;APLN;GPC4;CD99L2;PLXNA3;ANGPT2;BMP1;TNFRSF10B;SCARA5;FGFR1;PLAT;IKBKB;TGS1;NSMAF;LY96;IL7;SDC2;RRM2B;ANGPT1;TNFRSF11B;PTK2;JAK2;IL33;IL11RA;CCL19;TLN1;GNAQ;NTRK2;CTSL;TGFBR1;TNC;ANGPTL2;ABL1;RXRA;NOTCH1;CTSD;IGF2;CD81;ADM;SPON1;SAA1;LGR4;CD59;CD44;MDK;STX3;VEGFB;CAPN1;CD248;CTSF;CLCF1;GAL;P2RY6;BIRC3;BIRC2;CASP1;ATM;THY1;HSPA8;CHEK1;CDON;APLP2;JAM3;IL15RA;PLXDC2;PIP4K2A;BAMBI;ITGB1;NRP1;FZD8;CXCL12;MAPK8;DKK1;UNC5B;PLAU;VCL;BMPR1A;PTEN;FAS;CHUK;SLK;ADRA2A;PTPRE;TNFRSF1A;LAG3;CLEC2D;KLRD1;LRP6;KRAS;PTHLH;SLC38A2;VDR;TUBA1A;ACVRL1;ACVR1B;ITGA5;CD63;GDF11;STAT6;LRP1;MDM2;LUM;DCN;SOCS2;IGF1;CMKLR1;PTPN11;PRKAB1;SCARB1;ULK1;LATS2;SLC7A1;HMGB1;EDNRB;EFNB2;TNFSF13B;LAMP1;ANG;MMP14;PRMT5;ARF6;BMP4;GMFB;LGALS3;PGF;TGFB3;FLRT2;LGMN;BDKRB2;VRK1;HSP90AA1;GREM1;THBS1;PDIA3;FGF7;ANXA2;SEMA7A;CSK;MFGE8;SEMA4B;FURIN;RGMA;IGF1R;TNFRSF12A;VASN;IL4R;MAPK3;STX4;MMP2;CX3CL1;CDH11;CKLF;PLCG2;SERPINF1;PLD2;TNFSF12;CD68;TP53;ULK2;LGALS9;UNC119;CCL2;CCL7;CCL11;CCL8;CCL5;CCL4;ERBB2;RARA;STAT3;CCR10;RAMP2;GRN;MAP3K14;PPM1D;MRC2;PRKCA;LGALS3BP;C1QTNF1;COLEC12;LAMA1;LMAN1;SIRPA;PCNA;JAG1;SRC;PLCG1;PI3;MMP9;PTPN1;AURKA;BSG;STK11;PIP5K1C;CD70;TNFSF14;INSR;ANGPTL4;DNMT1;ICAM1;TYK2;LDLR;CALR;PRKACA;ADGRE5;NOTCH3;GDF15;COMP;FXYD1;MIA;AXL;TGFB1;PLAUR;APOE;PRKD2;CLEC11A;IL11;MAPK1;MIF;CHEK2;LIF;LGALS1;EP300;FBLN1;APP;IFNAR2;IFNAR1;ITGB2;DIP2A
Endothelial	PVL	77.68139	368	331	1056	0.02	0.2866667	0.0205714	101.29508	78.27642	MIF;EDN1;ITGB1;STX3;STX4;EFNB2;NOTCH1;NID1;FYN;GFRA1;TP53;LEPR;IL7R;LIFR;IL6ST;IL15RA;PTPN11;IL4R;TYK2;IFNAR2;TGFB3;TGFB1;IL6;ITGA9;EZR;ICAM1;SLC7A1;SIRPA;MAGED1;LGALS3BP;DLL1;MDK;JAG2;DLL4;JAG1;ITGB4;HLA-DRA;HLA-DRB5;HLA-DRB1;HLA-DQA1;HLA-DQB1;HLA-DQA2;HLA-DMB;HLA-DMA;HLA-DPA1;HLA-DPB1;ARF6;MDM2;JAK1;BTNL9;JAK2;IL18BP;STAT3;EPHA2;EPHA4;ABL1;MAPK8;LGALS3;CD74;P2RY6;SEMA3G;TEK;ITGA6;ITGAV;DAG1;ITGA1;ITGA2;CD44;ITGA3;ADRB2;LDLR;PTPRF;LMAN1;ACVR1B;IL1RAP;INHBB;BMP6;BMP4;VCAM1;SPP1;VCAN;CXCL12;VWF;THBS1;LAMC1;FGF2;LAMA4;LAMB1;TNC;MFGE8;CX3CL1;CALR;PDGFB;ADM;BMP1;SEMA3F;RARRES1;VEGFC;FLT4;PGF;GDF15;LAMB2;PRND;DUSP18;RPSA;LAMA3;LAMA5;MET;NRP1;LGALS9;DIP2A;TLN1;LTBP3;VCL;SELL;NRP2;CASP8;RIPK1;TNFRSF10B;TNFRSF10D;TNFRSF10A;TNFRSF11B;IL1R1;IL33;ITGA5;ACKR1;TLR4;S1PR1;PTGER4;PLXNA2;PLXNA1;FAM3C;SLC6A8;KDR;SELP;TGFBR3;FGFR1;IL17RA;PDGFA;PDGFD;COPA;ROBO1;ROBO4;TNFSF10;TNFRSF1A;APLP2;NOTCH4;CD81;CTSS;CTSL;CTSF;CD63;LGMN;POSTN;NFE2L2;FLT1;SDC3;ADAM15;CD55;SOS1;PRKCE;SCN9A;TUBA4A;ITPR1;RAF1;PIK3CA;LPP;TNK2;CASP3;PIK3R1;FLOT1;MAPK14;CD36;CDK6;RPS6KA3;PTK2;MCAM;THY1;SLK;PTPRE;TUBA1A;CDK2;CSK;LAT;PLCG2;PLD2;CD79B;PECAM1;PLCG1;BSG;STK11;PRKACA;PLAUR;CSF2RB;APP;IGF2;CD47;FURIN;HRH1;ACVR2A;ACVR1;ACVR2B;BAMBI;S100A9;S100A8;SAA1;SDC2;PLAU;PCNA;CD99;RARA;LAMP1;NCL;PTPRB;PLXNB2;CCRL2;PILRB;CD200;EPHB4;TIE1;BMPR2;FGF12;SPARC;FGF18;ANXA2;RXRA;HMGB1;ANGPTL4;APOE;POMC;PTHLH;CAPN1;TGS1;MFAP2;YBX1;CHUK;HLA-C;CALCRL;ACKR3;RAMP3;RAMP2;SCARB1;STAB1;HBEGF;APLN;MMRN2;TNFRSF1B;HLA-A;IL15;NPNT;ANGPTL2;JAM3;LUM;LGALS1;ALCAM;VEGFB;CXCR4;FLRT2;CXCL2;CXCL10;CXCL11;ANXA1;CCL23;IKBKB;NSMAF;GRN;RSPO3;CKLF;ACTR2;DCN;S100A4;INSR;NRAS;MPZL1;IFNGR1;MAP3K4;KRAS;MAPK3;LY96;TFRC;CD40;CCL2;PIP5K1A;SMAP1;PIP4K2A;PIP5K1C;PTPRK;TGFBR2;ACVRL1;LRP5;MMP2;PLAT;EFEMP2;SOCS2;TNFSF12;IL2RG;IL13RA1;STAT6;SNCA;TXLNA;LRP6;PLXDC2;TNFRSF12A;PLK3;MCL1;RGS16;PTGS2;ODC1;EIF2AK2;CASP10;CASP6;PLK2;NR3C1;CSNK2B;HSPA1B;TAP1;HSP90AB1;ABCB1;HIPK2;NOS3;RRM2B;CTSD;BIRC3;CASP1;ATM;HSPA8;PTEN;SLC38A2;PRKAB1;LATS2;BDKRB2;VRK1;ULK2;PPM1D;BCL2L1;DNMT1;MMP14;ETV5;AREG;EFNB1;KMT2E;LGALS8;VANGL1;AVPR2;MCFD2;CD46;BCAM;GPIHBP1;CDH11;CDH5;ARPC5;INPP5D;KITLG;NPTN;PTPN1;TAP2;PDIA3;SCARF1;ADAM23;PAM;GAS6;FSTL1;CD109;LSR;ITGA10;GMFB;TNFRSF14;EFNA1	TNFRSF14;ECE1;COL16A1;TXLNA;RHBDL2;YBX1;PTPRF;ARTN;PLK3;JAK1;TGFBR3;F3;VCAM1;CSF1;NRAS;NGF;VANGL1;NOTCH2;ECM1;MCL1;CTSS;PIP5K1A;S100A4;IL6R;EFNA1;MUC1;COPA;FCGR2A;MPZL1;ANGPTL1;RGS16;LAMC1;ARPC5;CD55;CD46;LAMB3;NID1;LGALS8;ODC1;EIF2AK2;SOS1;MCFD2;ACTR2;IL1R1;ACVR2A;ACVR1;NFE2L2;ITGA4;ITGAV;CALCRL;BMPR2;NRP2;TUBA4A;NCL;COL6A3;RAMP1;ITPR1;RAF1;TGFBR2;RPSA;LAMB2;DAG1;LRIG1;ALCAM;CD47;FSTL1;ITGB5;PLXNA1;PODXL2;RARRES1;TNFSF10;PIK3CA;BCL6;LPP;IL1RAP;SPON2;CXCL3;CXCL2;CXCL9;CXCL10;SPP1;NPNT;CASP6;EDNRA;CASP3;SEMA5A;LIFR;ITGA1;IL6ST;PLK2;PIK3R1;VCAN;EDIL3;RGMB;HSPA4;HBEGF;FGF1;NR3C1;IL17B;PDGFRB;CD74;SPARC;SLIT3;RIPK1;HLA-A;FLOT1;HLA-C;CSNK2B;HSPA1B;HLA-DRA;HLA-DQA1;HLA-DQB1;TAP2;TAP1;HLA-DPA1;HLA-DPB1;MAPK14;PTK7;VEGFA;HSP90AB1;SMAP1;CD109;FYN;LAMA4;PTPRK;MAP3K5;IFNGR1;ESR1;EZR;IGF2R;MAP3K4;THBS2;PDGFA;IL6;INHBA;CD36;HGF;CDK6;PILRB;KMT2E;LAMB1;FAM3C;PTN;HIPK2;RARRES2;CD99;MXRA5;RPS6KA3;AR;EFNB1;IL13RA1;CD99L2;ANGPT2;BMP1;FGFR1;PLAT;IKBKB;TGS1;LY96;SDC2;RRM2B;ANGPT1;TNFRSF11B;PTK2;GPR20;VLDLR;JAK2;CCL19;TLN1;GNAQ;NTRK2;CTSL;TGFBR1;TNC;ANGPTL2;ABL1;RXRA;NOTCH1;CTSD;CD81;ADM;SAA1;CD59;CD44;MDK;APLNR;CAPN1;CD248;CTSF;TRPC6;BIRC3;BIRC2;CASP1;ATM;MCAM;THY1;HSPA8;APLP2;JAM3;IL15RA;ITGA8;PIP4K2A;BAMBI;ITGB1;NRP1;CXCL12;MAPK8;UNC5B;PLAU;VCL;BMPR1A;PTEN;FAS;CHUK;SLK;ADRA2A;PTPRE;TNFRSF1A;LRP6;KRAS;SLC38A2;TUBA1A;ACVRL1;ACVR1B;ITGA5;ITGA7;CD63;GDF11;STAT6;LRP1;AVPR1A;MDM2;LUM;DCN;SOCS2;PTPN11;ULK1;LATS2;HMGB1;EDNRB;TNFSF13B;LAMP1;MMP14;PRMT5;ARF6;LGALS3;PGF;TGFB3;LGMN;VRK1;HSP90AA1;THBS1;PDIA3;SEMA6D;FGF7;ANXA2;CSK;LINGO1;MFGE8;SEMA4B;FURIN;RGMA;IGF1R;TNFRSF12A;VASN;IL4R;MAPK3;STX4;MMP2;CDH11;CKLF;SERPINF1;TNFSF12;TP53;ULK2;LGALS9;CCL2;CCL8;ERBB2;RARA;STAT3;CCR10;GRN;ITGA3;PPM1D;MRC2;BIRC5;LGALS3BP;C1QTNF1;LAMA3;LMAN1;BCL2;PCNA;JAG1;BCL2L1;MMP9;PTPN1;LAMA5;BSG;STK11;INSR;ANGPTL4;DNMT1;ICAM1;LDLR;EPOR;CALR;PRKACA;ADGRE5;NOTCH3;LGI4;FXYD1;LSR;AXL;TGFB1;PLAUR;BCAM;APOE;CLEC11A;MAPK1;MIF;DUSP18;LGALS1;EP300;FBLN1;APP;IFNAR2;IFNAR1;DIP2A;EPHA3;SEMA5B
Endothelial	B-cells	41.40780	364	323	1024	0.02	0.2940000	0.0205714	101.59016	41.86842	MIF;BIRC2;TNFRSF1A;GRN;ITGAV;ITGB1;ITGA5;EDN1;TP53;TYK2;STX3;STX4;FYN;NOTCH1;NID1;LEPR;IL7R;LIFR;IL6ST;IL15RA;PTPN11;IL4R;IFNAR2;SPP1;PDGFB;HLA-C;HLA-DRA;HLA-DRB1;DLL1;MDK;JAG2;DLL4;JAG1;HLA-DRB5;HLA-DQA1;HLA-DQB1;HLA-DQA2;HLA-DMB;HLA-DMA;HLA-DPA1;HLA-DPB1;ARF6;CD36;TLR4;MDM2;EPHA2;EPHA4;FLT4;EFNB2;CD74;P2RY6;INPP5D;CD34;VCAN;ITGA6;DAG1;ITGA1;ITGA2;CD44;ITGA3;ITGB4;ADRB2;LDLR;LGALS9;LGALS1;IL10RB;RAMP3;RAMP2;SNCA;LMAN1;ACVR1B;CXCL12;TGFB1;FGF12;LAMC1;LAMB2;LAMA4;LAMB1;THBS1;LAMA3;LAMA5;DUSP18;VCAM1;VWF;FGF2;TNC;MFGE8;CX3CL1;CALR;TNFSF10;BMP6;BMP1;BMP4;SEMA3F;SEMA3G;RARRES1;VEGFC;PGF;LRP5;LRP6;TBXA2R;TGFB3;GDF15;PRND;CCL2;CCL14;CCL23;CYTL1;NRP1;DIP2A;NRP2;CASP8;RIPK1;TNFRSF10B;TNFRSF10D;TNFRSF10A;TNFRSF11B;IL1R1;IL33;PECAM1;ACKR1;S1PR1;ITGA9;PTGER4;IL2RG;JAK1;PIK3R1;IL7;IL6;JAK2;STAT3;KDR;TEK;ARPC5;POMC;ACTR2;ADM;VEGFB;HSPA8;PTHLH;COPA;FGFR1;APLP2;RGS2;TM4SF1;C6orf15;CDK2;PLCG1;NOTCH4;CD81;TNFRSF14;TNFRSF1B;LTBR;CSF1;SEMA4C;NFE2L2;TNFRSF21;CHUK;TBK1;HMOX1;CD40;BAG6;CTSS;CTSL;CTSF;CD63;LGMN;FLT1;SIRPA;SDC3;PTPRF;ADAM15;CD55;PLXNA2;SOS1;PRKCE;SCN9A;TUBA4A;ITPR1;RAF1;PIK3CA;LPP;TNK2;CASP3;FLOT1;MAPK14;CDK6;RPS6KA3;PTK2;ABL1;MCAM;THY1;SLK;PTPRE;TUBA1A;CSK;LAT;PLCG2;PLD2;CD79B;BSG;STK11;PRKACA;PLAUR;CSF2RB;APP;LGALS3;ICAM1;IGF2;HRH1;PCNA;CD99;BTN3A2;RARA;LAMP1;EFNA1;EFNB1;PILRB;IL15;CXCL10;CXCL11;TIE1;SPARC;FGF18;RXRA;HMGB1;TLN1;CAPN1;MFAP2;YBX1;FURIN;SCARB1;STAB1;HBEGF;NCL;SDC2;PTPRB;TXLNA;VCL;HLA-A;INHBB;NPNT;ANGPTL2;JAM3;LUM;LGALS3BP;LYN;STAT6;INSR;BCR;IFNAR1;FAM3C;S100A8;MYL9;ACVR2A;BMPR2;ACVR2B;S100A4;MET;IGF2R;NRAS;MPZL1;IFNGR1;MAP3K4;KRAS;MAPK3;LY96;RHBDL2;EPHB4;TFRC;PLXNB2;PIP5K1A;SMAP1;PIP4K2A;PIP5K1C;MAPK1;PTPRK;ANXA2;CD47;MMP2;PLAT;ROBO4;PDGFA;DCN;SOCS2;TNFSF12;IL13RA1;ICAM2;ICAM3;PLXDC2;TNFRSF12A;S100A9;PLK3;MCL1;RGS16;PTGS2;ODC1;EIF2AK2;CASP10;CASP6;PLK2;NR3C1;CSNK2B;HSPA1B;TAP1;HSP90AB1;ABCB1;HIPK2;NOS3;RRM2B;CTSD;BIRC3;CASP1;ATM;PTEN;SLC38A2;PRKAB1;LATS2;BDKRB2;VRK1;ULK2;PPM1D;BCL2L1;DNMT1;CCRL2;SLC7A1;KMT2E;ZFP91;VANGL1;MCFD2;AREG;BCAM;GPRC5B;GPIHBP1;CDH11;CDH5;EZR;PTAFR;PRMT5;PTPN1;PRKD2;APOE;KITLG;NPTN;TAP2;PDIA3;SCARF1;MICA;ULBP2;TGFBR3;CXCR4;TGFBR2;FSTL1;CD109;ACVRL1;PLAU;LSR;HLA-B;HLA-F;GMFB;ACKR3;RPSA;MAPK8;CSF3	TNFRSF14;TNFRSF1B;PLA2G2A;ECE1;RPS6KA1;PTAFR;TXLNA;LCK;YBX1;PTPRF;PLK3;JAK1;S1PR1;CD53;NRAS;NOTCH2;MCL1;CTSS;PIP5K1A;S100A9;S100A4;EFNA4;FCRL1;COPA;SLAMF1;FCGR2B;MPZL1;SELL;RGS16;LAMC1;ARPC5;PTPRC;IL24;CD55;CR2;CD46;LAMB3;LGALS8;ODC1;POMC;EIF2AK2;SOS1;PRKCE;MCFD2;ACTR2;CXCR4;SCN3A;ITGA6;NFE2L2;ITGA4;ITGAV;CASP10;CASP8;BMPR2;NRP2;WNT10A;TUBA4A;NCL;ITPR1;GHRL;RAF1;KAT2B;TGFBR2;RPSA;CCR1;CCR2;DAG1;PRKCD;ALCAM;CD47;FSTL1;RARRES1;TNFSF10;PIK3CA;BCL6;LPP;IL1RAP;TNK2;SPON2;CD38;TLR1;TLR6;CXCL3;CXCL10;SPP1;IL15;CASP3;IL7R;PTGER4;IL6ST;PLK2;PIK3R1;RGMB;HSPA4;HBEGF;NR3C1;ADRB2;CD74;SPARC;RIPK1;HLA-A;FLOT1;DDR1;HLA-C;LTA;TNF;LTB;NCR3;CSNK2B;HSPA1B;NOTCH4;HLA-DRA;HLA-DRB5;HLA-DRB1;HLA-DQA1;HLA-DQB1;HLA-DQA2;HLA-DOB;TAP2;TAP1;HLA-DMB;HLA-DMA;HLA-DOA;HLA-DPA1;HLA-DPB1;MAPK14;VEGFA;HSP90AB1;SMAP1;FYN;LAMA4;PTPRK;MAP3K5;IFNGR1;EZR;IGF2R;MAP3K4;CCR6;LFNG;IL6;ABCB1;CDK6;AZGP1;PILRB;PILRA;KMT2E;PIK3CG;LAMB1;FAM3C;HIPK2;EPHB6;CDK5;CD99;RPS6KA3;IL2RG;CXCR3;IL13RA1;PLXNA3;ANGPT2;TNFRSF10B;TNFRSF10A;PTK2B;FGFR1;IKBKB;TGS1;NSMAF;LY96;IL7;RRM2B;TNFRSF11B;PTK2;JAK2;TLN1;CTSL;SYK;TGFBR1;NOTCH1;CTSD;CD81;SPON1;SAA1;CD59;CD44;MDK;PTPRJ;STX3;MS4A1;VEGFB;CAPN1;CLCF1;BIRC3;BIRC2;CASP1;ATM;IL10RA;HSPA8;APLP2;JAM3;IL15RA;IL2RA;PIP4K2A;BAMBI;ITGB1;MAPK8;VCL;PTEN;FAS;CHUK;SLK;PTPRE;PTPN6;CLEC2D;CD69;KRAS;SLC38A2;TUBA1A;ITGB7;CD63;GDF11;STAT6;MDM2;LUM;DCN;IGF1;PTPN11;PRKAB1;ULK1;FGF9;SLC7A1;HMGB1;EFNB2;TNFSF13B;LAMP1;ANG;PRMT5;RIPK3;ARF6;GMFB;LGALS3;LGMN;VRK1;HSP90AA1;THBS1;PDIA3;ANXA2;SEMA7A;CSK;IL16;MFGE8;SEMA4B;FURIN;IGF1R;TNFRSF12A;IL4R;CD19;LAT;MAPK3;ITGAL;STX4;CKLF;PLCG2;SERPINF1;TNFSF12;CD68;TP53;LGALS9;UNC119;CCL5;CCL4;RARA;STAT3;GRN;MAP3K14;PPM1D;CD79B;PECAM1;BIRC5;LGALS3BP;LMAN1;BCL2;PCNA;HCK;PI3;MMP9;PTPN1;LAMA5;BSG;STK11;CD70;FCER2;ANGPTL4;DNMT1;ICAM1;TYK2;LDLR;EPOR;CALR;PRKACA;ADGRE5;JAK3;LSR;HCST;MIA;TGFB1;CD79A;PLAUR;APOE;PRKD2;DBP;CD37;SIGLEC10;LILRB1;MAPK1;MIF;DUSP18;CSF2RB;LGALS1;EP300;APP;IFNAR2;IFNAR1;ITGB2;DIP2A
Endothelial	Plasmablasts	40.09062	162	123	346	0.02	0.8230303	0.0205714	52.92623	44.75000	MIF;NOTCH1;TP53;LEPR;IL7R;LIFR;IL6ST;IL15RA;PTPN11;IL4R;TYK2;IFNAR2;HLA-C;HLA-DRA;HLA-DRB1;HLA-DRB5;HLA-DQA1;HLA-DQB1;HLA-DQA2;HLA-DMB;HLA-DMA;HLA-DPA1;HLA-DPB1;MDM2;JAK1;BTNL9;IL6;JAK2;IL18BP;STAT3;CD74;P2RY6;CD34;VCAN;ADRB2;LDLR;FYN;LGALS9;LGALS1;JAG1;FGF2;TNC;MDK;THBS1;LAMA5;ANGPTL4;LMAN1;ACVR1B;CXCL12;TGFB1;LAMC1;LAMB2;LAMA4;LAMB1;LAMA3;DUSP18;VCAM1;SPP1;VWF;TNFSF10;TNFRSF1A;LRP5;LRP6;PRND;NRP1;NRP2;CASP8;RIPK1;TNFRSF10B;TNFRSF10D;TNFRSF10A;TNFRSF11B;PECAM1;S1PR1;ITGAV;ITGA9;PTGER4;CD44;ITGB1;ITGA5;KDR;TEK;TLR4;ARPC5;POMC;ACTR2;ADM;VEGFB;HSPA8;PTHLH;COPA;APLP2;NOTCH4;CD81;CTSS;CTSL;CTSF;CD63;LGMN;CALR;ICAM1;PCNA;RARA;LAMP1;PILRB;IL15;IL7;RXRA;HMGB1;PTK2;CAPN1;ITGB4;CD99;CD36;SCARB1;STAB1;HBEGF;HLA-A;NID1;NPNT;VEGFC;FLT4;TLN1;ANGPTL2;JAM3;LUM;TGFB3;LGALS3BP;FAM3C;S100A4;IGF2R;INSR;LY96;PIP5K1A;SMAP1;PIP4K2A;PIP5K1C;PTPRK;MCAM;ANXA2;PLAT;ROBO4;LGALS3;CASP3;TNFSF12;TNFRSF1B;EFNB2;CSK;MCFD2;GPRC5B;TAP2;TAP1;PDIA3;SCARF1;ITGA3;CD55;SDC2;LSR;APP;TNFRSF14;CXCR4;ACKR3	TNFRSF14;YBX1;PLK3;JAK1;CD53;NRAS;MCL1;CTSS;S100A4;IL6R;COPA;MPZL1;SELL;ARPC5;PTPRC;CD55;CD46;ODC1;SDC1;EIF2AK2;SOS1;MCFD2;ACTR2;CXCR4;ITGA6;NFE2L2;ITGA4;CASP10;CASP8;WNT10A;TUBA4A;NCL;RPSA;ALCAM;CD47;RARRES1;TNFSF10;CD38;SPP1;CASP3;IL6ST;PIK3R1;HSPA4;NR3C1;ADRB2;CD74;HLA-A;FLOT1;HLA-C;CSNK2B;HSPA1B;HLA-DRA;HLA-DRB5;HLA-DRB1;HLA-DQA1;HLA-DQB1;HLA-DQA2;TAP1;HLA-DMA;HLA-DPA1;HLA-DPB1;HSP90AB1;EZR;CDK6;KMT2E;PIK3CG;FAM3C;HIPK2;CD99;IL2RG;TGS1;LY96;TLN1;SYK;CTSD;CD81;SAA1;CD59;CD44;BIRC3;CASP1;ATM;IL10RA;HSPA8;APLP2;IL15RA;ITGB1;CLEC2D;KRAS;SLC38A2;CD63;MDM2;IGF1;HMGB1;ARF6;LGALS3;HSP90AA1;PDIA3;ANXA2;IL16;MFGE8;TNFRSF12A;CKLF;PLCG2;STAT3;GRN;CD79B;PECAM1;LMAN1;BCL2L1;PTPN1;BSG;CD70;CALR;ADGRE5;CD79A;APOE;PRKD2;CD37;MIF;LGALS1;IFNAR2;IFNAR1
Endothelial	T-cells	52.56231	375	374	1146	0.02	0.2360000	0.0205714	105.78689	49.82816	MIF;TNFSF12;BIRC2;TNFRSF1A;GRN;ITGAV;ITGB1;ITGA5;EDN1;TP53;FYN;STX3;STX4;NOTCH1;APOE;LEPR;IL7R;LIFR;IL6ST;IL15RA;PTPN11;IL4R;TYK2;IFNAR2;TGFB3;TGFB1;ITGA9;EZR;ICAM1;SPP1;PDGFB;SLC7A1;SIRPA;HLA-C;HLA-DRA;HLA-DRB1;LGALS3BP;CD58;CD59;DLL1;MDK;JAG2;DLL4;JAG1;TNFRSF14;HLA-DRB5;HLA-DQA1;HLA-DQB1;HLA-DQA2;HLA-DMB;HLA-DMA;HLA-DPA1;HLA-DPB1;ARF6;CD36;TLR4;MDM2;JAK1;BTNL9;IL6;JAK2;IL18BP;STAT3;EPHA2;EPHA4;CD74;P2RY6;INPP5D;CD34;VCAN;TNFRSF10B;TNFRSF4;ADRB2;LDLR;LGALS9;LGALS1;ITGA6;DAG1;ITGA2;CD44;ITGA3;ITGB4;FGF2;TNC;THBS1;LAMA5;ANGPTL4;RAMP3;RAMP2;LMAN1;ACVR1B;HLA-F;HLA-A;HLA-E;HLA-B;LAT;CXCL12;CXCL2;CXCL10;CXCL11;PAM;ADA;LAMC1;LAMB2;LAMA4;LAMB1;LAMA3;DUSP18;VCAM1;VWF;NID1;MFGE8;CX3CL1;CALR;TNFSF10;BMP6;BMP1;BMP4;PIK3R1;LRP5;LRP6;EFNA1;EFNB1;EFNB2;ITGA1;ADM;FAM3C;GDF15;PRND;CCL2;CCL14;CCL23;S100A4;MET;NRP1;CD200;NRP2;CASP8;RIPK1;TNFRSF10D;TNFRSF10A;TNFRSF11B;IL1R1;IL33;PECAM1;SDC3;ACKR1;KDR;SDC2;PLCE1;PRKACA;S1PR1;PTGER4;IL2RG;IL7;F2R;F2RL3;TEK;ARPC5;POMC;ACTR2;VEGFB;HSPA8;PTHLH;PDGFA;PDGFD;COPA;FGFR1;APLP2;NOTCH4;CD81;TNFRSF1B;LTBR;CSF1;SEMA4C;NFE2L2;FLT4;TNFRSF21;CHUK;TBK1;HMOX1;CD40;BAG6;CTSS;CTSL;CTSF;CD63;LGMN;FLT1;PGF;PTPRF;ADAM15;CD55;PLXNA2;SOS1;PRKCE;SCN9A;TUBA4A;ITPR1;RAF1;PIK3CA;LPP;TNK2;CASP3;FLOT1;MAPK14;CDK6;RPS6KA3;PTK2;ABL1;MCAM;THY1;SLK;PTPRE;TUBA1A;CDK2;CSK;PLCG2;PLD2;CD79B;PLCG1;BSG;STK11;PLAUR;CSF2RB;APP;IGF2;PCNA;CD99;RARA;LAMP1;SEMA6B;CCRL2;PILRB;EPHB4;IL15;SEMA3F;TIE1;FGF12;SPARC;FGF18;ANXA2;RXRA;HMGB1;CAPN1;TGS1;MFAP2;YBX1;FURIN;SCARB1;STAB1;HBEGF;NCL;PTPRB;TXLNA;ALCAM;TLN1;VCL;CD320;NPNT;VEGFC;ANGPTL2;JAM3;LUM;SELP;F8;IKBKB;NSMAF;SNCA;LGALS3;LYN;STAT6;INSR;BCR;IFNAR1;CLEC2B;S100A8;MYL9;ROBO3;ACVR2A;BMPR2;ACVR2B;PSMB8;PSMB9;IFNGR1;IFNGR2;IGF2R;NRAS;MPZL1;MAP3K4;KRAS;MAPK3;LY96;TFRC;PIP5K1A;SMAP1;PIP4K2A;PIP5K1C;MAPK1;PTPRK;CD47;MMP2;PLAT;ROBO4;IL13RA1;ICAM2;ICAM3;TNFRSF12A;PLK3;MCL1;RGS16;PTGS2;ODC1;EIF2AK2;CASP10;CASP6;PLK2;NR3C1;CSNK2B;HSPA1B;TAP1;HSP90AB1;ABCB1;HIPK2;NOS3;RRM2B;CTSD;BIRC3;CASP1;ATM;PTEN;SLC38A2;PRKAB1;LATS2;BDKRB2;VRK1;ULK2;PPM1D;BCL2L1;DNMT1;ETV5;AREG;KMT2E;ZFP91;VANGL1;MCFD2;CD46;GPRC5B;PTAFR;PRMT5;PTPN1;PRKD2;KITLG;SOCS2;NPTN;TAP2;PDIA3;SCARF1;MICA;ULBP2;TGFBR3;CXCR4;TGFBR2;FSTL1;CD109;ACVRL1;DCN;PLAU;LSR;GMFB;ACKR3;OSMR;RPSA;MAPK8;S100A9	TNFRSF14;TNFRSF25;TNFRSF1B;PLA2G2A;ECE1;RPS6KA1;FGR;TXLNA;LCK;YBX1;PLK3;LRP8;JAK1;TGFBR3;VCAM1;S1PR1;CSF1;CD53;NRAS;VANGL1;CD2;NOTCH2;CD160;MCL1;CTSS;PIP5K1A;S100A9;S100A8;S100A4;S100A1;IL6R;ADAM15;EFNA4;COPA;SLAMF1;CD247;MPZL1;SELL;FASLG;TNFSF4;RGS16;ARPC5;PTPRC;CD55;CD46;LAMB3;LGALS8;RRM2;ODC1;SDC1;POMC;EIF2AK2;SOS1;MCFD2;ACTR2;CD8A;CD8B;ZAP70;CXCR4;DPP4;ITGA6;NFE2L2;ITGA4;ITGAV;CASP10;CASP8;BMPR2;CD28;CTLA4;ICOS;WNT10A;TUBA4A;EPHA4;NCL;COL6A3;RAMP1;PDCD1;ITPR1;RAF1;KAT2B;TGFBR2;CX3CR1;RPSA;CCR1;CCR5;DAG1;PRKCD;LRIG1;ALCAM;CD47;CD200R1;RARRES1;TNFSF10;PIK3CA;BCL6;LPP;IL1RAP;APOD;TNK2;SPON2;CD38;TLR1;TXK;CXCL8;CXCL1;CXCL2;CXCL10;CXCL13;SPP1;CASP6;IL15;CASP3;IL7R;PTGER4;ITGA1;ITGA2;GZMA;IL6ST;PLK2;PIK3R1;HSPA4;HBEGF;NR3C1;ADRB2;PDGFRB;CD74;SPARC;HAVCR2;ITK;RIPK1;HLA-A;FLOT1;HLA-C;LTA;TNF;LTB;NCR3;CSNK2B;HSPA1B;HLA-DRA;HLA-DRB5;HLA-DRB1;HLA-DQA1;HLA-DQB1;HLA-DQA2;TAP2;TAP1;HLA-DMB;HLA-DMA;HLA-DPA1;HLA-DPB1;MAPK14;VEGFA;HSP90AB1;SMAP1;CD109;FYN;MAP3K5;IFNGR1;EZR;IGF2R;MAP3K4;CCR6;LFNG;ABCB1;CDK6;AZGP1;PILRB;KMT2E;PIK3CG;FAM3C;PLXNA4;HIPK2;EPHB6;RARRES2;CDK5;CD99;RPS6KA3;EFNB1;IL2RG;CXCR3;CD40LG;PLXNA3;ANGPT2;TNFRSF10B;TNFRSF10A;PTK2B;FGFR1;PLAT;IKBKB;TGS1;NSMAF;LY96;RRM2B;TNFRSF11B;JAK2;TLN1;GNAQ;NTRK2;CTSL;SYK;TGFBR1;ABL1;RXRA;NOTCH1;CTSD;CD81;SPON1;SAA1;CD59;CD44;MDK;PTPRJ;STX3;MS4A1;CD6;CD5;VEGFB;CAPN1;CTSF;BIRC3;BIRC2;CASP1;ATM;IL10RA;JAML;CD3E;CD3D;CD3G;HSPA8;CHEK1;APLP2;IL15RA;IL2RA;PRKCQ;PIP4K2A;ITGB1;MAPK8;CDK1;PLAU;VCL;PTEN;FAS;CHUK;SLK;PTPRE;VWF;TNFRSF1A;LAG3;CD4;PTPN6;KLRB1;CLEC2D;CD69;KLRF1;KLRD1;KLRC2;KLRC1;KRAS;NELL2;SLC38A2;VDR;TUBA1A;ITGB7;ITGA5;CD63;GDF11;CDK2;STAT6;IFNG;MDM2;LUM;DCN;SOCS2;IGF1;PTPN11;PRKAB1;FLT1;SLC7A1;HMGB1;TNFSF13B;LAMP1;PRMT5;RIPK3;ARF6;GMFB;LGALS3;LGMN;VRK1;HSP90AA1;THBS1;PDIA3;ANXA2;CSK;IL16;MFGE8;FURIN;TNFRSF12A;IL4R;LAT;SPN;MAPK3;ITGAL;STX4;ITGAM;ITGAX;CKLF;PLCG2;TNFSF12;TP53;LGALS9;UNC119;CCL5;CCL4;ERBB2;RARA;STAT3;GRN;MAP3K14;PPM1D;CD79B;PECAM1;PRKCA;BIRC5;LGALS3BP;LMAN1;BCL2;CD226;PCNA;JAG1;BCL2L1;PLCG1;PI3;MMP9;PTPN1;AURKA;BSG;STK11;CD70;TNFSF14;DNMT1;ICAM1;TYK2;LDLR;EPOR;CALR;PRKACA;ADGRE5;JAK3;LSR;HCST;TGFB1;CD79A;PLAUR;APOE;PRKD2;DBP;CD37;LILRB1;KIR3DL1;KIR3DL2;MAPK1;MIF;LIF;OSM;DUSP18;IL2RB;LGALS1;EP300;APP;IFNAR2;IFNAR1;ITGB2;DIP2A

Intercellular cross-talk visualization

Intercellular cross-talk results can be filtered to maintain only the significant ones. The results can be visualized by using Ulisse package functions. The code below can be used to plot the results as a network. filtering argument is used to control which intercellular cross-talks have to be visualized in the network. If set to TRUE, then p_val, FDR and ct_val are used to identify the significant communications and plot these with a solid line, whereas all the others will be plotted with a dashed line. Here, we enabled this visualization, by providing p-value and FDR filtering. ct_val = NULL allows to consider all intercellular cross-talk without filtering for this value. community can be logical (if the community should be calculated by using igraph::fastgreedy.community()), or a community object as calculated with igraph package. vertex is used to color the vertices, and can be set to "name" to color them according to the names of the cell-types. edge_col_by and edge_width parameters control the value used to color the edges and if their width should be proportional to that value. edge_adj_col is used to control the transparency of the edges: as in pathway cross-talk there can be lots of links it can e useful to control the transparency. plot_network_CT() function returns the igraph cross-talk network with the communities under “comm_id” vertex attribute (if calculated). plot_network_CT() function uses ggraph package functions, which are ggplot2-based. If file_out is set to NULL the function returns also the ggplot2 network object.


expr_ccc_net <- plot_network_CT(ct = expr_ccc$ct_res, 
                                filtering = T, p_val = 0.05, FDR = 0.2, ct_val = NULL, 
                                community = F, 
                                vertex = "name", vertex_size = 5, vertex_label = TRUE, vertex_pal = pal_ct,
                                edge_col_by = "ct_score", edge_pal=NULL, 
                                edge_width = T, edge_adj_col = 0.7) 
expr_ccc_net

#> IGRAPH be33f36 UN-- 9 36 -- 
#> + attr: name (v/c), label (v/c), ct_score (e/n), ngenes_gs1 (e/n),
#> | ngenes_gs2 (e/n), nlink (e/n), p_value_link (e/n), FDR_link (e/n),
#> | p_adj_BH (e/n), weight_gs1 (e/n), weight_gs2 (e/n), genes_gs1 (e/c),
#> | genes_gs2 (e/c), filt (e/c)
#> + edges from be33f36 (vertex names):
#>  [1] Endothelial--CAFs              Endothelial--PVL              
#>  [3] Endothelial--B-cells           Endothelial--Plasmablasts     
#>  [5] Endothelial--T-cells           Endothelial--Myeloid          
#>  [7] Endothelial--Cancer Epithelial Endothelial--Normal Epithelial
#>  [9] CAFs       --PVL               CAFs       --B-cells          
#> + ... omitted several edges

ct_heatmap() function can be used to plot intercellular cross-talk as an heatmap by selecting the variable that have to be used to color the heatmap (color_by, here we used the cross-talk score), color scale (color = c("lightyellow", "red3")), and the color of cells in the heatmap of the cell types pairs without communications (no_ct_color).

Similarly to plot_network_CT(), ct_heatmap() function can select significant interactions, which are represented with a star. This is achieved by enabling filtering = TRUE, that will further require p-value, FDR and/or cross-talk score thresholds.

ct_heatmap(ct = clCC$ct_res, 
           color_by = "ct_score", color = c("lightyellow", "red3"), no_ct_color = "whitesmoke", 
           filtering = TRUE, p_val = 0.05, FDR = 0.2, ct_val = NULL,label_size = 8)

DEG data

We will now analyse intercellular cross-talk by using DE results.

Here we will use Seurat function for Differential Expression Analysis with default parameters to calculate DEG for each cell type against the others. Then, we will use Benjamini-Hochberg method to obtain adjusted p-value to be used to filter DEGs to maintain only the significant ones. We will select only the up-regulated genes.

Idents(data) <- data$celltype_major

DEGs_data <- FindAllMarkers(data, assay = "RNA") #this may take a while
DEGs_data$p_val_BH <- p.adjust(DEGs_data$p_val, method = "BH")

DEGs_data[1:5,]

	avg_log2FC	pct.1	pct.2	cluster	gene
ACKR1	4.585262	0.615	0.011	Endothelial	ACKR1
VWF	3.505398	0.861	0.012	Endothelial	VWF
PLVAP	3.213029	0.820	0.013	Endothelial	PLVAP
RAMP2	3.016198	0.877	0.011	Endothelial	RAMP2
ECSCR.1	2.816407	0.861	0.006	Endothelial	ECSCR.1

It is not important which package and algorithm is used for DEG calculation, as Ulisse function is structured to handle any input type.

DEGs data can now be used to build the cell-type lists by using preparing_DEG_list() function. This function does not take as an input a table with the results of DEG calculation, but the specific columns (genes, their log2 Fold Change, p-value and the cluster), so that it is not important which package and algorithm is used for the differential analysis. preparing_DEG_list() uses these inputs to calculate for each gene in clusters the associated weights, that are the product between log2FC and the -log10 of p_val. Then, the function builds a vector for each cluster, each one composed by the weights, named after the respective genes. Then, these vectors are filtered to maintain only the gene that are present in universe and returned as a list

DEGs_data <- DEGs_data[DEGs_data$p_val_BH <= 0.05 & DEGs_data$avg_log2FC >= 0.5,]


DEG_list <- preparing_DEG_list(cluster = DEGs_data$cluster, 
                               p_val = DEGs_data$p_val_BH, 
                               log2FC = DEGs_data$avg_log2FC, 
                               gene = DEGs_data$gene, 
                               universe = rownames(adj.m))

head(DEG_list[[1]])

#>     ACKR1       VWF     RAMP2     RAMP3   CLEC14A    ADGRL4 
#> 1397.4371 1068.3300  919.2382  839.4853  801.0583  800.9338

Communication Analysis

The subsequent part of the analysis is identical to the one used for expression data: gs_cross_talk() is used to analyse intercellular cross-talk with the same inputs and parameters. The calculation will results again in a two-object list with the detail of the gene-gene interaction and the cross talk results.


DEG_ccc <- gs_cross_talk(gs_list = DEG_list,
                         gene_network_adj = adj.m,
                         k = 49, shared = T, 
                         ct_info = T, hash = F,
                         mc_cores_perm = 1, mc_cores_ct = 1)

DEG_ccc$ct_info[1:5,]

gs1	gs1_gene	gs2	gs2_gene	score
Endothelial	KDR	CAFs	DCN	369360.44
Endothelial	ITGB1	CAFs	LUM	62407.07
Endothelial	ITGB1	CAFs	COL6A3	44302.73
Endothelial	NOTCH4	CAFs	THBS2	127874.51
Endothelial	ITGB1	CAFs	THBS2	30321.90

DEG_ccc$ct_res[1:5,]

gs1	gs2	ct_score	ngenes_gs1	ngenes_gs2	nlink	p_value_link	FDR_link	p_adj_BH	weight_gs1	weight_gs2	genes_gs1	genes_gs2
Endothelial	CAFs	3706408.2	59	75	148	0.02	0.0200000	0.0566667	10179.2059	18710.881	KDR;ITGB1;NOTCH4;ITGA6;CD36;INSR;ITGA5;ITGAV;CD200;PTPRB;NRP1;PDGFB;TIE1;BMPR2;FZD4;SPARC;MMRN2;FGFR1;VWF;ROBO4;ROBO1;THBS1;CD81;TGFBR3;CDH5;IL6;SELP;ACVRL1;TGFBR2;ACKR1;TEK;ANGPTL2;CXCL12;ANGPTL4;PGF;CD40;VEGFC;SEMA3G;NID1;LAMB1;LAMA4;VCAM1;LAMC1;LAMA5;IL6ST;LIFR;STAT3;RAMP2;RAMP3;CALCRL;CD99;ADM;TNFRSF10B;GRN;PLXNA2;PLXND1;NRP2;APP;FLT1	DCN;LUM;COL6A3;THBS2;IGF1;CXCL12;MXRA5;PTN;PDGFRA;ANGPTL2;FST;WNT2;PDGFRB;FGFR1;CD248;FBLN1;FGF7;SPARC;GREM1;SCARA5;NID1;VCAM1;SLIT3;MMP2;CD99;LRP1;INHBA;ANGPTL4;COMP;F3;PLA2G2A;BMP1;VCAN;LGALS3BP;TGFB3;LAMB1;CCL2;LGALS1;CLEC11A;SPON2;ITGA5;CDH11;IGF2;NRP1;TNFSF13B;MDK;NRP2;ITGB1;ITGA1;IL6;IL6ST;ADM;ITGAV;THBS1;PTHLH;LAMC1;LAMA4;TNC;CD81;CXCL11;RAMP1;TNFSF10;TNFRSF1A;SDC2;CXCL2;STAT3;CXCL9;CLCF1;CXCL10;SEMA3C;CXCL6;SPON1;CXCL1;CXCL3;VEGFA
Endothelial	PVL	1055689.1	51	52	107	0.02	0.0200000	0.0566667	6838.0833	3553.581	NRP1;PDGFB;CXCL12;LAMB1;LAMC1;TEK;TIE1;ROBO4;ROBO1;ACKR1;FLT1;NRP2;FGFR1;ITGAV;PAM;EDN1;ITGB1;DLL4;DLL1;JAG1;JAG2;CCL23;ITGA5;ACVRL1;TGFBR2;TGFBR3;CDH5;ITGA6;MMRN2;IL6ST;NOTCH4;VWF;VEGFC;NID1;LAMA4;VCAM1;ANGPTL2;LAMA5;THBS1;CD81;KDR;ANGPTL4;APP;PTK2;STAB1;CD36;LIFR;IL6;STAT3;PTPRB;GRN	PDGFRB;AVPR1A;ITGA1;ANGPT2;SLIT3;CCL8;PGF;SPARC;EDIL3;FXYD1;EDNRA;NID1;NOTCH3;ADRA2A;ITGA7;EDNRB;ANGPT1;VCAM1;CCL2;TGFB3;LGALS1;ANGPTL4;LAMA4;CD248;ANGPTL2;IL6;JAG1;COL6A3;ITGB1;ANGPTL1;CD99;LAMC1;CD36;PTK2;SDC2;LGALS3BP;LAMB2;APOE;FGF7;TLN1;CD59;LAMB1;THBS1;IL6ST;STAT3;ITGAV;CXCL12;CLEC11A;TNC;ECE1;TNFRSF1A;BCAM
Endothelial	B-cells	134594.1	11	9	14	0.94	0.8782609	0.9987500	1791.6487	1769.084	TNFRSF1A;CD40;CXCL12;CD81;CD63;CD34;EFNB2;TNFSF10;VCAM1;VWF;THBS1	LTB;CXCR4;HLA-DRA;SELL;FCRL1;CCR6;EZR;HLA-DRB1;ITGA4
Endothelial	Plasmablasts	112975.8	1	1	1	0.92	0.9817647	0.9987500	881.9485	128.098	PECAM1	CD38
Endothelial	T-cells	2248603.9	40	23	56	0.02	0.1311111	0.0566667	5940.8172	4960.810	F2R;CLEC2B;CD59;ACKR1;PECAM1;KDR;CSF2RB;EPHA4;MCAM;PLXNA2;APP;ITGB1;ADAM15;SLK;TUBA1A;STAT3;FLOT1;PTK2;PTPRE;CD36;CD55;CXCL12;VWF;VCAM1;THBS1;TNFRSF1A;CD40;GRN;CD34;CD81;LIFR;IL6ST;ACVRL1;TGFBR2;FSTL1;TGFBR3;ITGAV;ICAM2;ITGA5;TEK	GZMA;KLRB1;CD2;CCL5;FYN;KLRF1;CXCL13;CXCR4;ITGA4;LTB;TNFRSF1B;ADGRE5;ITGB7;CD38;SELL;CD99;JAK1;TGFB1;ITGB2;TLN1;STAT3;DIP2A;PIK3R1

Intercellular cross-talk visualization

As done before, we can build intercellular cross-talk network by filtering the results and using Ulisse function and plot it as a network

DEG_CCC_net <- plot_network_CT(ct = DEGcl_ct$ct_res, 
                               filtering = T, p_val = 0.05, FDR = 0.2, ct_val = NULL,  
                               community = F, 
                               vertex = "name", vertex_size = 5, vertex_label = TRUE, vertex_pal = pal_ct,
                               edge_col_by = "ct_score", edge_pal=NULL, 
                               edge_width = T, edge_adj_col = 0.7) 
DEG_CCC_net

#> IGRAPH 22bb420 UN-- 9 34 -- 
#> + attr: name (v/c), label (v/c), ct_score (e/n), ngenes_gs1 (e/n),
#> | ngenes_gs2 (e/n), nlink (e/n), p_value_link (e/n), FDR_link (e/n),
#> | p_adj_BH (e/n), weight_gs1 (e/n), weight_gs2 (e/n), genes_gs1 (e/c),
#> | genes_gs2 (e/c), filt (e/c)
#> + edges from 22bb420 (vertex names):
#>  [1] Endothelial--CAFs              Endothelial--PVL              
#>  [3] Endothelial--B-cells           Endothelial--Plasmablasts     
#>  [5] Endothelial--T-cells           Endothelial--Myeloid          
#>  [7] Endothelial--Cancer Epithelial Endothelial--Normal Epithelial
#>  [9] CAFs       --PVL               CAFs       --B-cells          
#> + ... omitted several edges

or as an heatmap

ct_heatmap(ct = DEGcl_ct$ct_res, 
           color_by = "ct_score", color = c("lightyellow", "red3"), no_ct_color = "whitesmoke", 
           filtering = T, p_val = 0.05, FDR = 0.2, ct_val = NULL,  label_size = 8)

Comparison between expression and DEG intercellular cross-talk results

Ulisse provides a function to build and visualize a union network out of multiple analyses or samples. comparing_results_network() function takes as an input a list of results obtained with gs_cross_talk(). The function does not align the results, so there should be correspondence among the gene sets names of the different samples, especially when using clusters. Here we will compare the significant results obtained with expression and DEG intercellular cross-talk analyses on the same sample, thus the cell-types corresponds. comparing_results_network() needs as an input a named list with the cross-talk results that should be already filtered to maintain only the significant intercellular cross-talk, if needed. This function will create a union network where the vertices are all the cell types present in the results in res_list, the links are the all the communication present between them in the results. The vertices can be colored by considering the number of results in which are present (vertex_number = TRUE) and/or by a discrete variable (in this case vertex = "name", otherwise a two list, where the first element is the name of the variable, the second the named vector with, see function help for detail). If both parameters are enabled, as in this case, the vertex one is used to color voronoi tesselation (see ggrpah package for further detail). When vertex_number = TRUE, a color palette can be provided, composed by minimum and maximum value colors, or the function will use as default “blue” to “red”. Edges can be colored by the number of results that share them (edge_color_by = "number") or, as in this case, by which results contain them (edge_color_by = "which"). Edges width can be proportional to the number of results in which they are present (edge_width = TRUE). comparing_results_network() will return the union network and the plot id file_out argument is not provided.

res_list <- list(expr = expr_ccc$cc_communications,
                 DEG = DEG_ccc$cc_communications)
res_list <- lapply(res_list, function(x) x <- x[x$p_value_link <= 0.05 & x$FDR_link <= 0.2,])

comp.out <- comparing_results_network(res_list = res_list, 
                                      vertex_number = TRUE, 
                                      vertex = "name", vertex_pal = pal_ct,
                                      voronoi_radius = 0.3, voronoi_alpha = 0.3, 
                                      edge_width = TRUE, edge_color_by = "which",
                                      edge_adj_col = 0.5, vertex_number_adj = 0.8)

comp.out

References

Hao, Yuhan, Stephanie Hao, Erica Andersen-Nissen, William M. Mauck, Shiwei Zheng, Andrew Butler, Maddie J. Lee, et al. 2021. “Integrated analysis of multimodal single-cell data.” Cell 184 (13): 3573–3587.e29. https://doi.org/10.1016/j.cell.2021.04.048.

Türei, Dénes, Alberto Valdeolivas, Lejla Gul, Nicolàs Palacio-Escat, Michal Klein, Olga Ivanova, Márton Ölbei, et al. 2021. “Integrated Intra-and Intercellular Signaling Knowledge for Multicellular Omics Analysis.” Molecular Systems Biology 17 (3): e9923.

Wu, Sunny Z, Ghamdan Al-Eryani, Daniel Lee Roden, Simon Junankar, Kate Harvey, Alma Andersson, Aatish Thennavan, et al. 2021. “A Single-Cell and Spatially Resolved Atlas of Human Breast Cancers.” Nature Genetics 53 (9): 1334–47.

2024-03-29

Preparation of the inputs

Ligand-receptor network

Pre-processing with Seurat pipeline

Intercellular cross-talk analysis

Expression data

Communication analysis

Intercellular cross-talk visualization

DEG data

Communication Analysis

Intercellular cross-talk visualization

Comparison between expression and DEG intercellular cross-talk results

References