-
Notifications
You must be signed in to change notification settings - Fork 1
/
03-check-data.Rmd
211 lines (178 loc) · 9.87 KB
/
03-check-data.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
---
title: "PCAs"
output:
pdf_document: default
html_notebook: default
---
```{r load}
load('pca.rda')
# metadata[harmonized_sample_ontology_intermediate == 'CD34-negative, CD41-positive, CD42-positive megakaryocyte cell', harmonized_sample_ontology_intermediate := 'CD34-negative, CD41-positive\nCD42-positive megakaryocyte cell']
# metadata[harmonized_sample_ontology_intermediate %in% names(which(table(harmonized_sample_ontology_intermediate) < 5)), harmonized_sample_ontology_intermediate := 'other']
```
```{r function}
pca_plots <-
function(matrix,
metadata,
title = '',
color_by = 'harmonized_sample_ontology_term_high_order_fig1',
shape_by = 'harmonized_biomaterial_type',
na.omit = FALSE,
na.impute = 'zero') {
require(ggfortify)
if (na.omit){
matrix <- matrix[, colSums(is.na(matrix)) == 0, drop=FALSE]
} else {
if (na.impute == 'zero') {
# set NA to zero
matrix[is.na(matrix)] <- 0
} else if (na.impute == 'mean') {
for (i in 1:ncol(matrix)) {
matrix[is.na(matrix[, i]), i] <- mean(matrix[, i], na.rm = TRUE)
}
}
}
# filter out cols without variance
matrix <- matrix[, apply(matrix, 2, var, na.rm=TRUE) != 0, drop = FALSE]
# perform PCA
pca <- prcomp(matrix, center = TRUE, scale = TRUE)
# compute explained variance of PCs
var_explained <- 100 * summary(pca)[["importance"]][2, ]
cum_var_explained <- cumsum(var_explained) # Cumulative variance
pcs_to_use <- which(cum_var_explained >= 90)[1] # PCs to use
pc_scores <- pca$x[, 1:pcs_to_use]
sil_scores <- cluster::silhouette(as.integer(as.factor(metadata[, harmonized_sample_ontology_term_high_order_fig1])),
dist(pc_scores))
metadata[, sil_score := sil_scores[, 'sil_width']]
metadata <- metadata[base::order(harmonized_sample_ontology_term_high_order_fig1, -sil_score)]
metadata[, epirr_id_without_version := factor(epirr_id_without_version, levels = unique(epirr_id_without_version))]
return(
list(
pca = pca,
sil = sil_scores,
sil_plot = ggplot(metadata,
aes(x = sil_score, y = harmonized_sample_ontology_term_high_order_fig1,
color = harmonized_sample_ontology_term_high_order_fig1)) +
geom_boxplot() +
stat_summary(fun=mean, geom="point", shape=18, size = 3) +
scale_color_manual(values = sample_hex_colors) +
theme_bw() + labs(title = paste(title, 'Mean/median silhouette:', metadata[, round(mean(sil_score), 3)], metadata[, round(median(sil_score), 3)]), color = 'Sample Ontology'),
dimred = ggplot2::autoplot(
pca,
data = metadata,
colour = color_by,
shape = shape_by
) + labs(title = title, color = 'Sample Ontology', shape = 'Biomaterial Type') + scale_color_manual(values=sample_hex_colors[names(sample_hex_colors) %in% metadata[, harmonized_sample_ontology_term_high_order_fig1]]) + theme_bw(),
scree = qplot(seq.int(var_explained), var_explained) +
geom_line() +
xlab("Principal Component") +
ylab("Percent Variance Explained") +
ggtitle(paste("Variance explained:", title))
)
)
}
```
## PCA on gene_level
```{r pca_tpm, fig.height=7, fig.width=10}
final_plot_pca_tpm <- NULL
final_tpm_matrix <- NULL
lapply(names(transcript_types), function(types){
tpm_dt <- readRDS(paste0('suppa_analysis/gene_expressions_', types, '.rds'))
tpm_dt <- dcast(tpm_dt, gene_id ~ epirr_id_without_version)
tpm_matrix <- as.matrix(tpm_dt[, -1])
rownames(tpm_matrix) <- tpm_dt[[1]]
tpm_matrix <- t(log2(tpm_matrix + 1))
tpm_all <- pca_plots(tpm_matrix[metadata[, epirr_id_without_version], ], metadata, paste('TPM', types))
print(tpm_all$scree)
print(tpm_all$sil_plot)
ggsave(filename = file.path(plot_dir, paste0(types, '_silhoutte_TPM.pdf')), width = 10, height = 7)
labels <- metadata[, harmonized_sample_ontology_term_high_order_fig1]
# labels[duplicated(labels, fromLast = TRUE)] <- '' # which(duplicated(labels, fromLast = TRUE))
# labels[rows_to_label] <- metadata[, harmonized_sample_ontology_intermediate][rows_to_label]
plot_pca_tpm <- tpm_all$dimred + geom_text_repel(aes(label = labels), min.segment.length = 0, max.overlaps = 5, force = 5) + guides(col= guide_legend(ncol = 2), shape = guide_legend(ncol = 2))
if (types == 'TSL') {final_plot_pca_tpm <<- plot_pca_tpm
final_tpm_matrix <<- tpm_matrix}
print(plot_pca_tpm)
ggsave(filename = file.path(plot_dir, paste0(types, '_PCA_TPM.pdf')), width = 10, height = 7)
})
```
## UMAP on TPM
```{r umap_tpm, fig.height=7, fig.width=10}
library(umap)
umap_tpm <- umap::umap(final_tpm_matrix[metadata[, epirr_id_without_version], ])
df <- data.frame(x = umap_tpm$layout[,1],
y = umap_tpm$layout[,2],
harmonized_sample_ontology_term_high_order_fig1 = metadata$harmonized_sample_ontology_term_high_order_fig1,
harmonized_biomaterial_type = metadata$harmonized_biomaterial_type,
label = '')
# rows_to_label <- c(250, 270, 300, 47, 333, 275, 351, 315, 166)
# df[duplicated(df$harmonized_sample_ontology_intermediate, fromLast = TRUE), 'label'] <- ''# which(duplicated(df$harmonized_sample_ontology_intermediate, fromLast = TRUE))#''
# df$label <- metadata[, .I]
df$label <- metadata$harmonized_sample_ontology_term_high_order_fig1#[rows_to_label]
plot_umap_tpm <- ggplot(df, aes(x, y, color = harmonized_sample_ontology_term_high_order_fig1, shape = harmonized_biomaterial_type, label=label)) +
geom_point() +
geom_text_repel(color = 'black', min.segment.length = 0, max.overlaps = 25, force = 5) +
labs(x = "UMAP1", y = "UMAP2", title = 'TPM', color = 'Sample Ontology', shape = 'Biomaterial Type') +
scale_color_manual(values=sample_hex_colors) + theme_bw() + guides(col= guide_legend(ncol = 2), shape = guide_legend(ncol = 2))
print(plot_umap_tpm)
ggsave(filename = file.path(plot_dir, 'TSL12_UMAP_TPM.pdf'), width = 10, height = 7)
```
<!-- ## PCA on Isoform Usage -->
<!-- ```{r, fig.height=7, fig.width=12} -->
<!-- isoform_usages <- t(as.matrix(read.table('suppa_analysis/all_isoform.psi', header=TRUE, sep="\t"))) -->
<!-- pca_plots(isoform_usages, metadata, 'IU w/ outliers') -->
<!-- ``` -->
<!-- ```{r, fig.height=7, fig.width=12} -->
<!-- pca_plots(isoform_usages[metadata_filtered[, epirr_id_without_version],], metadata_filtered, 'IU w/o outliers') -->
<!-- ``` -->
## PCA on AS events (PSI values)
```{r pca_psi, fig.height=7, fig.width=10}
final_plot_pca_psi <- NULL
final_psi_table <- NULL
lapply(names(transcript_types), function(types){
psi_files <- list.files('suppa_analysis/events', pattern = paste0('^event_', types, '_.+\\.psi$'), full.names = TRUE)
psi_table <- do.call(cbind, lapply(psi_files, function(f) t(as.matrix(read.table(f, header=TRUE, sep="\t")))))
psi_table <- psi_table[,colSums(is.na(psi_table))<nrow(psi_table)]
stopifnot(all(startsWith(colnames(psi_table), 'ENSG')))
for(i in 1:ncol(psi_table)){
psi_table[is.na(psi_table[,i]), i] <- mean(psi_table[,i], na.rm = TRUE)
}
psi_all <- pca_plots(psi_table[metadata[, epirr_id_without_version], ], metadata, paste('PSI', types))
print(psi_all$scree)
print(psi_all$sil_plot)
ggsave(filename = file.path(plot_dir, paste0(types, '_silhoutte_PSI.pdf')), width = 10, height = 7)
labels <- metadata[, harmonized_sample_ontology_term_high_order_fig1]
plot_pca_psi <- psi_all$dimred + geom_text_repel(aes(label = labels), min.segment.length = 0, max.overlaps = 6, force = 5) + guides(col= guide_legend(ncol = 2), shape = guide_legend(ncol = 2))
if (types == 'TSL') {final_plot_pca_psi <<- plot_pca_psi
final_psi_table <<- psi_table}
print(plot_pca_psi)
ggsave(filename = file.path(plot_dir, paste0(types, '_PCA_PSI.pdf')), width = 10, height = 7)
})
```
```{r umap_psi, fig.height=7, fig.width=10}
library(umap)
umap_psi <- umap::umap(final_psi_table[metadata[, epirr_id_without_version], ])
df <- data.frame(x = umap_psi$layout[,1],
y = umap_psi$layout[,2],
harmonized_sample_ontology_intermediate = metadata$harmonized_sample_ontology_intermediate,
harmonized_biomaterial_type = metadata$harmonized_biomaterial_type,
label = '')
# rows_to_label <- c(250, 270, 300, 47, 333, 275, 351, 315, 166)
# df[duplicated(df$harmonized_sample_ontology_intermediate, fromLast = TRUE), 'label'] <- ''# which(duplicated(df$harmonized_sample_ontology_intermediate, fromLast = TRUE))#''
# df$label <- metadata[, .I]
df$label <- metadata$harmonized_sample_ontology_intermediate#[rows_to_label]
plot_umap_psi <- ggplot(df, aes(x, y, color = harmonized_sample_ontology_intermediate, shape = harmonized_biomaterial_type, label=label)) +
geom_point() +
geom_text_repel(color = 'black', min.segment.length = 0, max.overlaps = 25, force = 5) +
labs(x = "UMAP1", y = "UMAP2", title = 'PSI', color = 'Sample Ontology', shape = 'Biomaterial Type') +
scale_color_manual(values=sample_hex_colors) + theme_bw() + guides(col= guide_legend(ncol = 2), shape = guide_legend(ncol = 2))
print(plot_umap_psi)
ggsave(filename = file.path(plot_dir, 'TSL12_UMAP_PSI.pdf'), width = 10, height = 7)
```
```{r both_pca, fig.height=7, fig.width=17}
print((final_plot_pca_tpm + final_plot_pca_psi) + plot_layout(guides = "collect") & theme(legend.position = 'bottom')& guides(col= guide_legend(nrow = 4), shape = guide_legend(nrow = 4)))
ggsave(filename = file.path(plot_dir, 'TSL12_PCAs.pdf'), width = 14, height = 8)
```
```{r both_umap, fig.height=7, fig.width=17}
print((plot_umap_tpm + plot_umap_psi) + plot_layout(guides = "collect") & theme(legend.position = 'bottom')& guides(col= guide_legend(nrow = 4), shape = guide_legend(nrow = 4)))
ggsave(filename = file.path(plot_dir, 'TSL12_UMAPs.pdf'), width = 14, height = 8)
```