-
Notifications
You must be signed in to change notification settings - Fork 2
/
DECIPHER-seq.Rmd
438 lines (287 loc) · 15.3 KB
/
DECIPHER-seq.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
---
title: "DECIPHER-seq"
output:
md_document:
variant: markdown_github
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, error = TRUE)
```
## Data and code for "DECIPHER-seq uncovers cell-cell interaction networks in the human breast" (Murrow et al.)
Included is the processed data and coded needed to replicate key findings from Murrow et al. Raw gene expression and barcode count matrices will be uploaded to the Gene Expression Omnibus (insert GEO ID here). Any additional information required to reproduce the data in this study is available from the corresponding authors upon request.
```{r renv, results = FALSE}
renv::restore()
```
## Integrated breast dataset with cell type clustering
Here, we've chosen to read in and plot a subset of the full dataset for speed.
```{r loaddata, message=F}
library(Seurat)
library(ggplot2)
source('R/DECIPHER-seq.functions.R')
source('R/DECIPHER-seq.util.R')
breast.data.sub <- readRDS('Data/breast.data.sub.rds')
```
```{r plotUMAP, echo=FALSE}
Seurat::DimPlot(breast.data.sub, pt.size=1, label=F, group.by="Type", cols = cell_type_cols) +
labs(x=NULL, y=NULL, title = NULL) +
theme(axis.text = element_blank(), axis.ticks = element_blank(), axis.line=element_blank())
```
## Define Activity Programs
First, we run consensus iNMF (using the LIGER package) across a range of K values. A heuristic for the maximum value of K for the sweep can be chosen by inspecting PC elbow plots for each cell type. Here, we chose a range of 2 to 40 for all cell types.
```{r run_iNMF, eval = F}
library(rliger)
library(RANN)
breast.data <- readRDS('Data/breast.data.rds')
NMF_results <- iNMF_ksweep(breast.data, k.max = 40)
```
Alternatively, read in the consensus iNMF results from the (entire) processed dataset instead of rerunning the full workflow.
```{r NMF_results}
NMF_results <- readRDS('Data/NMF_results.rds')
```
Build phylogenetic trees for K optimization workflow
```{r phylo_trees}
library(rlist)
library(ape)
library(matrixStats)
phylo_trees <- lapply(NMF_results, build_phylo_tree)
```
Identify "outlier" activity programs representing rare contaminating cells and plot phylo partitions
```{r outlier_id}
program_outlier_score <- lapply(NMF_results, identify_outlier_programs)
```
Choose distance threshold and partition phylogenetic trees
```{r phylo_dist_thresholding, warning = F}
library(patchwork)
suggested_thresholds = suggest_dist_thresh(phylo_trees)
thresh_use = round(max(suggested_thresholds), 1)
p = plot_hists(phylo_trees, thresh.use = thresh_use)
p[[1]]+p[[3]]+p[[2]]+p[[4]]+p[[5]]
```
```{r phylo_partitioning, message = F}
library(geiger)
library(igraph)
suggested_thresholds = suggest_dist_thresh(phylo_trees)
thresh_use = round(max(suggested_thresholds), 1)
phylo_partitions = mapply(partition_phylo_tree, x = phylo_trees, y = program_outlier_score, dist.thresh = thresh_use, outlier.thresh = 5, SIMPLIFY = F)
```
```{r plot_phylos}
par(mfrow=c(1,5), mar=c(2.5,5.5,1,1))
plot_phylo_trees(phylo_trees, phylo_partitions)
```
Plot K metric (number of weighted subtrees identified at each choice of NMF rank)
```{r plot_K_curves, message = F}
library(dplyr)
K_metrics = lapply(phylo_partitions, calculate_K_metric)
k.use = lapply(K_metrics, suggest.k)
print(unlist(k.use))
par(mfrow=c(5,1), mar=c(2.5,5.5,1,1))
for (i in 1:length(K_metrics)){
plot(2:(dim(K_metrics[[i]])[1]+1), K_metrics[[i]]$weighted_n_subtrees,
ylab = "Weighted N subtrees", xlab = "NMF Rank K",
main = names(K_metrics)[i], pch = 16, col = cell_type_cols[i], type = "b")
abline(v = k.use[[i]])
}
```
Select NMF results at optimized K values and filter "outlier" programs representing expression in rare cells.
```{r select_K}
NMF_results_atK <- mapply(FUN = NMF_results_opt_k, NMF_results, k.use, program_outlier_score, SIMPLIFY = F)
```
## Construct DECIPHER-seq network
Calculate the average expression score of each activity program in each sample
```{r H_score}
metadata <- readRDS('Data/breast.data.metadata.rds')
Expression_score = mapply(NMF_results = NMF_results_atK, metadata = list(metadata), FUN = calc.H.score, SIMPLIFY = F)
```
Calculate pairwise Pearson correlations of average sample expression scores for each activity program, perform BCA bootstrap resampling to estimate confidence intervals
```{r bootstrap_resampling, eval = F}
source("R/boot.cor.complete.R")
source("R/cor.m.boot.test.R")
library(wBoot)
library(parallel)
Expression_score_cor = cor.m.boot.test(list.cbind(Expression_score), null.hyp = 0, alternative="two.sided")
Expression_score_cor$sig.cor = Expression_score_cor$cor
Expression_score_cor$sig.cor[which(Expression_score_cor$p> 0.05)] = NA
```
```{r H_score_data, echo = F}
Expression_score_cor <- readRDS('Data/expression_score_correlation.rds')
```
Build weighted network.
```{r network}
Network <- Construct_network(Expression_score_cor)
```
Perform community detection.
```{r community_detection, eval = F}
library(reticulate)
adjacency_matrix <- Network$mat
py_run_string("import leidenalg as la; import igraph as ig; import numpy as np")
py_run_string("G = ig.Graph.Weighted_Adjacency(r.adjacency_matrix.tolist())")
# sweep across a range of resolutions
py_run_string("optimiser = la.Optimiser()")
py_run_string("profile = optimiser.resolution_profile(G, la.CPMVertexPartition,
weights = 'weight', resolution_range=(0.001, 0.4), number_iterations = 0)")
sweep = py$profile
modularity = lapply(sweep, function(x){x$modularity})
# Use "resolution" that gives max modularity
partition_use = sweep[[which.max(unlist(modularity))]]
py_run_string("partition = r.partition_use")
py_run_string("diff = optimiser.optimise_partition(partition)")
# Optimise this partition
while(py$diff!=0){
py_run_string("diff = optimiser.optimise_partition(partition)")
py_run_string("print(diff)")
}
clustering_res = py$partition
modules = clustering_res$membership + 1
names(modules) = colnames(adjacency_matrix)
Network$modules = modules
```
```{r community_detection_data, echo = F}
Network$modules <- readRDS('Data/CPM_results.rds')
Network$network_coords <- readRDS('Data/Network_coords_unfiltered.rds')
```
Filter isolated nodes and modules using weighted topological overlap.
```{r wTO_filtering, eval = F, message = F}
library(wTO)
Network <- Filter_network(Network)
```
```{r filtered_network_data, echo = F}
Network <- readRDS('Data/Network.rds')
```
Plot results (pre- and post-filtering).
```{r Network_plots, echo = F, message=F}
library(DescTools)
edge_weights_fisher = abs(FisherZ(edge.attributes(Network$network)$weight*edge.attributes(Network$network)$sign))
plot(Network$network, layout = Network$network_coords,
vertex.color= c(network_module_cols[c(1:6)], "grey65", network_module_cols[c(7,8)], c("grey50", "grey35", "grey20"))[Network$modules],
vertex.label= NA,
edge.width=0.5*edge_weights_fisher,
vertex.label.cex=0.4, vertex.size = 7, vertex.label.color="black",
vertex.label.family="Helvetica", vertex.frame.color=NA, vertex.label.font=2,
edge.color = c("lightblue", "red")[factor(E(Network$network)$sign>0)],
main = "Unfiltered network")
legend("right",legend=1:length(unique(Network$modules)),
col=c(network_module_cols[c(2,4,3,1,5,6,7,8)], "grey65", "grey50", "grey35", "grey20"),
bty = "n", pch = 19)
edge_weights_fisher = abs(FisherZ(edge.attributes(Network$filtered_network)$weight*edge.attributes(Network$filtered_network)$sign))
plot(Network$filtered_network, layout = Network$filtered_network_coords,
vertex.color=network_module_cols[factor(Network$filtered_modules)],vertex.label = NA,
edge.width=0.5*edge_weights_fisher, vertex.label.cex=0.5, vertex.size = 7, vertex.label.color="black",
vertex.label.family="Helvetica", vertex.frame.color=NA, vertex.label.font=2,
edge.color = c("lightblue", "red")[factor(E(Network$filtered_network)$sign>0)],
main = "Filtered network")
legend("right",legend=1:length(unique(Network$filtered_modules)),
col=c(network_module_cols[c(2,4,3,1,5,6,7,8)]), bty = "n",
pch = 19)
library(gplots)
library(scales)
Heatmap_wrapper(Expression_score_cor, Network)
```
## Downstream analyses of DECIPHER-seq network
### Identify activity programs with similar gene loadings
Infer non cell type specific responses, e.g. in response to a shared cue or common microevironment. Non-specific gene signatures (such as artifacts of tissue processing) would also be expected to display similar gene loadings.
```{r gene_loading_similarities}
gene_correlation_matrix <- Gene_similarity(NMF_results_atK)
gene_similarity_node_pvals <- Permutation_test_gene_cor(Network, gene_correlation_matrix)
edge_weights_fisher = abs(FisherZ(edge.attributes(Network$filtered_network)$weight*edge.attributes(Network$filtered_network)$sign))
plot(Network$filtered_network, layout = Network$filtered_network_coords,
vertex.color=colorRampPalette(c("grey", "red"))(100)[cut(-log10(gene_similarity_node_pvals), breaks = c(0, seq(-log10(0.01), 4, length.out = 100)))],
edge.width=0.25, vertex.label.cex=0.5, vertex.size = 7, vertex.label.color="black",
vertex.label.family="Helvetica", vertex.frame.color=NA, vertex.label.font=2, vertex.label = NA,
edge.color = c(NA, "grey20")[factor(E(Network$filtered_network)$sign>0)],
main = "Gene loading similarity within module")
legend_image <- as.raster(matrix(rev(colorRampPalette(c("grey", "red"))(100)), ncol=1))
text(x=1.9, y = c(-0.4,0.4), labels = c("1e-2", "1e-4"), cex = 0.7)
rasterImage(legend_image, 1.55, -0.45, 1.7,0.45)
```
### Infer high-confidence direct cell-cell interactions.
This simplified model does not consider the effects of signal amplification, cooperation between signaling pathways, or higher-order interactions between more than two cell types, but can be used to identify a subset of “high-confidence” direct cell-cell interactions that meet a series of simple criteria. Only test for cell types where the single-cell dataset reflects the tissue composition (e.g. cell types where dissociation artifacts are minimized and sort gates that don't enrich for one cell type selectively).
```{r direct_interactions}
library(broom)
interaction_term_res = Infer_direct_interactions(Expression_score, Network, metadata, celltypes.test = c("HRpos_Luminal", "Secretory_Luminal", "Basal"), sort.gate = c("Live_singlet", "Epithelial"))
mat = interaction_term_res$adjacency_matrix
mat[is.na(mat)] = 0
network_directed <- graph_from_adjacency_matrix(t(mat), weighted=T, mode="directed", diag=F)
plot.igraph(Network$filtered_network, layout = Network$filtered_network_coords,
vertex.color=network_module_cols[factor(Network$filtered_modules)],vertex.label = NA,
edge.width=0.5*edge_weights_fisher, vertex.label.cex=0.5, vertex.size = 7, vertex.label.color="black",
vertex.label.family="Helvetica", vertex.frame.color=NA, vertex.label.font=2,
edge.color = c(NA, "grey50")[factor(E(Network$filtered_network)$sign>0)])
plot.igraph(network_directed, layout = Network$filtered_network_coords,
vertex.color= alpha("black", 0),
edge.width=2, vertex.label.cex=0.5, vertex.size = 3, vertex.label.color="black",
vertex.label.family="Helvetica", vertex.frame.color=NA, vertex.label.font=2,
edge.color = "black",
vertex.label = NA, edge.arrow.size = 0.3, edge.arrow.width = 2, add = T)
```
### Gene Set Enrichment Analysis
##### Marker gene analysis
Identify marker genes statistically associated with each gene program, using ordinary least squares regression of each gene’s normalized (z-scored) expression against the activity program expression score (e.g. cell "expression") for each program in each cell type, after filtering genes not expressed in that cell type
```{r gene_marker_matrix, eval = F}
marker_gene_list <- Marker_gene_analysis(NMF_results_atK, NMF_results)
```
```{r read_in_gene_marker_matrix, echo = F}
marker_gene_list <- readRDS('Data/W_mat.rds')
```
##### Gene set enrichment
Perform gene set enrichment analysis across activity programs in the DECIPHER-seq network.
```{r gsea, eval = F}
library(fgsea)
library(msigdbr)
# Targeted Gene Sets: luteal phase (Pardo et al.) and involution-related (Stein et al.)
path_list = readRDS('Data/genesets/targeted_pathways.rds')["Pardo"]
Pardo_fgsea_res = fgsea_test(marker_gene_list, Network, path_list)
path_list = readRDS('Data/genesets/targeted_pathways.rds')["Stein"]
Stein_fgsea_res = fgsea_test(marker_gene_list, Network, path_list)
# GO Biological Processes
path_df = msigdbr(species = "Homo sapiens", category="C5")
path_df=subset(path_df, gs_subcat%in%c("GO:BP"))
path_list = path_df %>% split(x = .$gene_symbol, f = .$gs_name)
GO_BP_fgsea_res = fgsea_test(marker_gene_list, Network, path_list)
# Hallmark Gene Sets
path_df = msigdbr(species = "Homo sapiens", category="H")
path_list = path_df %>% split(x = .$gene_symbol, f = .$gs_name)
Hallmark_fgsea_res = fgsea_test(marker_gene_list, Network, path_list)
```
```{r read_in_fgsea_res, echo = F}
Pardo_fgsea_res = readRDS('Data/genesets/Pardo_fgsea_res.rds')
Stein_fgsea_res = readRDS('Data/genesets/Stein_fgsea_res.rds')
GO_BP_fgsea_res = readRDS('Data/genesets/GO_BP_fgsea_res.rds')
Hallmark_fgsea_res = readRDS('Data/genesets/hallmark_fgsea_res.rds')
```
Plot enrichment of selected gene sets.
```{r plot_node_fgsea_res}
# A few example plots
plot_fgsea(Network, Pardo_fgsea_res, "Pardo", fdr_min = 1e-10)
plot_fgsea(Network, Stein_fgsea_res, "Stein", fdr_min = 1e-4)
plot_fgsea(Network, Hallmark_fgsea_res, "HALLMARK_HYPOXIA", fdr_min = 1e-20)
```
##### Enrichment in modules
Identify gene sets enriched across modules in the DECIPHER-seq network.
```{r module_enrichment, eval = F}
# choose gene sets to test
sets_to_test = rbind(Stein_fgsea_res, Pardo_fgsea_res, GO_BP_fgsea_res, Hallmark_fgsea_res)
enrichment_pval = Get_enrichment_pvals(sets_to_test, Network)
```
```{r read_in_module_enrichment, echo = F}
enrichment_pval = readRDS('Data/enrichment_pval.rds')
# list top enriched gene sets within, e.g. module 3
print(sort(unlist(lapply(enrichment_pval, '[[', "Module_3")))[1:5])
```
#### Association with metadata features
Measure the effect size and p-values for association with metadata features. Mann-Whitney test and Wilcoxon effect size (r) for binary features, linear regression and Pearson correlation (r) for continuous features.
```{r calc_effect_size, message = F}
library(rstatix)
# for binary variables, create a metadata column with only two levels
metadata$Parity.Y.N[metadata$Parity.Y.N=="unknown"] = NA
metadata$Parity.Y.N = factor(metadata$Parity.Y.N, levels = c("yes", "no"))
Parity_effect_size = Calculate_metadata_associations(Network, Expression_score, metadata, feature.to.test = "Parity.Y.N", type = "binary")
BMI_effect_size = Calculate_metadata_associations(Network, Expression_score, metadata, feature.to.test = "BMI", type = "continuous")
metadata$HC_use[metadata$HC_use=="progestin"] = NA
metadata$HC_use = factor(metadata$HC_use, levels = c("combined", "none"))
HC_use_effect_size = Calculate_metadata_associations(Network, Expression_score, metadata, feature.to.test = "HC_use", type = "binary")
```
```{r plot_effect_size}
Plot_metadata_association(Network, Parity_effect_size, plot.title = "Parity")
Plot_metadata_association(Network, BMI_effect_size, plot.title = "BMI")
Plot_metadata_association(Network, HC_use_effect_size, plot.title = "Combined HC Use")
```