-
Notifications
You must be signed in to change notification settings - Fork 1
/
ms_proteomics_pipeline.Rmd
executable file
·1963 lines (1669 loc) · 116 KB
/
ms_proteomics_pipeline.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
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
---
title: "AZ-SciLife ExoDiab MS Proteomics analysis"
author: "Klev Diamanti, PhD"
date: "`r format(Sys.time(), '%A, %d %B %Y, %H:%M')`"
output:
html_document:
dev: 'svg'
---
```{r setup, echo=FALSE, warning=FALSE, message=FALSE, results='hide'}
# set output directory
output.directory <- "../results/"
# set knitr options
knitr::opts_chunk$set(echo = TRUE, fig.path = output.directory)
# load libraries
suppressWarnings(suppressPackageStartupMessages(library(readxl)))
suppressWarnings(suppressPackageStartupMessages(library(knitr)))
suppressWarnings(suppressPackageStartupMessages(library(tidyverse)))
suppressWarnings(suppressPackageStartupMessages(library(tidyselect)))
suppressWarnings(suppressPackageStartupMessages(library(reshape2)))
suppressWarnings(suppressPackageStartupMessages(library(cowplot)))
suppressWarnings(suppressPackageStartupMessages(library(kableExtra)))
suppressWarnings(suppressPackageStartupMessages(library(gridExtra)))
suppressWarnings(suppressPackageStartupMessages(library(limma)))
## Other libraries that will be loaded but not attached
# data.table
# plyr
# dendextend
# magrittr
# qlcMatrix
# quantable
# ProteoMM
# DEqMS
# NormalyzerDE
# variancePartition
# pcaMethods
# corrplot
# BiocParallel
# ComplexHeatmap
# circlize
# Rtsne
# fgsea
# mgsub
# xlsx
# umap
# imputeLCMD
# impute
# ggdendro
# set java options to output files
options(java.parameters = "-Xmx8000m")
```
```{r setup.init, echo=FALSE, warning=FALSE, message=FALSE, results='hide'}
## INPUT FILES
proteomics.file <- "../data/proteomics_random.xlsx"
covars.file <- "../covariates/covars_random.xlsx"
gene.id.file <- "../data/gsea_files/all.gene.ids.txt"
## TISSUES
tissues <- c("Fat", "Liver", "Muscle", "Pancreas", "Serum")
tissue.colors <- setNames(object = c('#814374ff', '#51a39dff', '#b7695cff', '#cdbb79ff', '#06425cff'), nm = tissues)
## MAIN PHENOTYPE
covars.pheno <- "pheno_3c"
## FILTERING QC
min.samples.threshold <- 0.8 # fraction of samples a protein was identified in
samples.to.remove.manually <- c("rand3408") # hardcoded removal of samples
## SIGNIFICANCE
# thresholds for pi-values
# thresh_pi <- 0.2572 # p = 0.2
# thresh_pi <- 0.4292 # p = 0.1
# thresh_pi <- 0.6274 # p = 0.05
# thresh_pi <- 1.1733 # p = 0.01
# thresh_pi <- 2.1281 # p = 0.001
significance.signs <- data.frame(sign = c("***", "**", "*", "#", "##", "NS"),
p.q = c(0.001, 0.01, 0.05, 0.1, 0.2, 1),
pi = c(2.1281, 1.1733, 0.6274, 0.4292, 0.2572, 0), stringsAsFactors = F)
significance.threshold.p.q.relaxed <- 0.1
significance.threshold.p.q.strict <- 0.05
significance.threshold.p.q.strictest <- 0.001
significance.threshold.pi <- significance.signs %>%
filter(p.q <= significance.threshold.p.q.relaxed) %>%
slice(n()) %>%
pull(pi)
## CONTRASTS TO CONSIDER IN THE DIFFERENTIAL ANALYSIS (DE)
within.tissues.contrasts <- setNames(object = c("t2dVSctrl", "t2dVSpd", "pdVSctrl", "t2dVSctrl_pd", "pd_t2dVSctrl"),
nm = c("T2D v CTRL", "T2D v PD", "PD v CTRL", "T2D v CTRL+PD", "PD+T2D v CTRL"))
## FILES WITH GO TERMS AND PATHWAYS
biological.terms.files <- data.frame(type = c("GO", "GO", "GO", "Pathway"),
subtype = c("bp", "cc", "mf", ""),
file = c("../data/gsea_files/gmt_files/c5.go.bp.v7.2.symbols.kd.gmt", "../data/gsea_files/gmt_files/c5.go.cc.v7.2.symbols.kd.gmt", "../data/gsea_files/gmt_files/c5.go.mf.v7.2.symbols.kd.gmt", "../data/gsea_files/gmt_files/c2.cp.v7.2.symbols.kd.gmt"),
stringsAsFactors = F) %>%
mutate(combo = paste(type, subtype, sep = ""))
GOslim.file <- "../data/gsea_files/obo_files/goslim_analysis-ready.tsv"
GOall.file <- "../data/gsea_files/gmt_files/c2cpANDc5go.v7.2.symbols.kd.gmt"
## PARAMETERS FOR GO AND PATHWAY ENRICHMENT ANALYSIS
min.num.of.proteins.in.term <- 5
max.num.of.proteins.term <- 300
num.of.permutations <- 1e+06
## SEED
seed <- 1
set.seed(seed = seed)
# OS-SPECIFIC ARGUMENTS AND THREADS
switch(Sys.info()[['sysname']],
Windows = {
num.of.cores.total <- parallel::detectCores() - 1
bash.cmd.prefix <- "wsl " },
Linux = {
num.of.cores.total <- parallel::detectCores() - 1
bash.cmd.prefix <- "" },
Darwin = {
num.of.cores.total <- parallel::detectCores() - 1
bash.cmd.prefix <- "" })
if (num.of.cores.total <= length(tissues)) {
num.of.cores.tissues <- num.of.cores.total
} else {
num.of.cores.tissues <- 5
}
## WRITE TO EXCEL FUNCTION
write.to.excel <- function(df_, file_, sheet_, append_) {
df_ <- as.data.frame(df_, check.names = F, fix.empty.names = F)
rownames(df_) <- NULL
xlsx::write.xlsx2(x = df_, file = file_, sheetName = sheet_, col.names = TRUE, row.names = FALSE, append = append_)
}
# HELPER FUNCTION TO COMBINE ALL DATASETS FOR ALL TISSUES IN ONE MATRIX
merge.full.joins <- function(list.of.df_, by_) {
merge.tmp <- list.of.df_[[1]]
for (i in 2:length(list.of.df_)) {
merge.tmp <- full_join(merge.tmp, list.of.df_[[i]], by = by_)
}
return(merge.tmp)
}
merge.unique.proteins.dia <- function(dia.all_) {
tmp.names <- names(dia.all_)
dia.all_ <- lapply(seq_along(dia.all_), function(i) {
tmp.dia.all_ <- dia.all_[[i]]
colnames(tmp.dia.all_) <- c("AZpID", paste(names(dia.all_)[[i]], tail(colnames(tmp.dia.all_), -1), sep = "."))
return(tmp.dia.all_)
})
names(dia.all_) <- tmp.names
return(merge.full.joins(list.of.df_ = dia.all_, by_ = 'AZpID'))
}
merge.unique.samples.dia <- function(dia.all_) {
merge.unique.samples.dia.help <- function(dia_, tissue_) {
dia.t_ <- data.table::transpose(l = dia_ %>% column_to_rownames("AZpID")) %>%
as.data.frame()
colnames(dia.t_) <- dia_$AZpID
dia.t_ <- dia.t_ %>%
mutate(id = tail(colnames(dia_), -1)) %>%
select(id, everything())
colnames(dia.t_) <- c(head(colnames(dia.t_), 1), paste(tail(colnames(dia.t_), -1), ".", tissue_, sep = ""))
return(dia.t_)
}
dia.t.all_ <- lapply(seq_along(dia.all_), function(i) merge.unique.samples.dia.help(dia_ = dia.all_[[i]], tissue_ = names(dia.all_)[i]))
return(merge.full.joins(list.of.df_ = dia.t.all_, by_ = 'id'))
}
```
# Analyzing mass spectrometry (MS) proteomics for 5 tissues
- Visceral Adipose Tissue (VAT) - herein Fat
- Liver
- Skeletal Muscle - herein Muscle
- Pancreatic Islets - herein Pancreas
- Blood Serum - herein Serum
# Reading the data
## Reading clinical variables
Reading the random clinical variables. Available anthropometric data include _age_, _sex_, _body mass index (bmi)_, _blood group_, _% of purity of islets in sample_ and _cold ischemic time (cit)_. Additional clinical variables related to type 2 diabetes (T2D) include _hemoglobin A1c (hba1c)_, _glucose-stimulated insulin secretion (GSIS)_ in pancreatic islets, and the discrete T2D phenotype _pheno_3c_ of _control (CTRL)_, _pre-diabetes (PD)_ and _T2D_. We introduce 2 additional columns _pehno_ctrl_pd_ and _pheno_pd_t2d_ that merge the group of PD with CTRL and T2D, respectively.
The table below shows distribution of samples across discrete phenotypes in columns _CTRL_, _PD_, _T2D_ and _NA_, where _NA_ shows samples with non available phenotype information. Columns starting with _mean ± sd_ contain the average and the standard deviation for the corresponding row variable in each phenotypic group. The columns _P (CTRL v PD)_, _P (CTRL v T2D)_, _P (PD v T2D)_, _P (CTRL v PD+T2D)_ and _P (PD+T2D v CTRL)_ show the statistical significance measured by a Mann-Whitney U test for the groups in parentheses, while the column _P (CTRL v PD v T2D)_ shows statistical significance across all 3 phenotypic groups in _pheno_3c_ measured by a Kruskal-Wallis test. The columns _P (bmi)_, _P (hba1c)_ and _P (gsis)_ represent associations measured by a linear regression. The row for _sex_ shows the distribution of sexes across phenotypic groups.
```{r read.covars, echo=FALSE, warning=FALSE, message=FALSE, results='asis'}
# read covars file
covars <- read_excel(path = covars.file, sheet = 1) %>%
mutate_all(as.character) %>%
map_df(~parse_guess(.))
# assign predefined colors to phenotypes
phenotypes.colors <- data.frame(names = c(sort(unique(c(covars$pheno_3c, covars$pheno_ctrl_pd, covars$pheno_pd_t2d))), "CTRL_PD__T2D", "PD_T2D__CTRL"),
colors = c('#dcd0c0ff', '#bbbac6ff', '#e2e2e2ff', '#90afc5ff', '#b7ceceff', '#ffffffff', '#ffffffff'), stringsAsFactors = F) %>%
mutate(class = c("pheno_3c", "pheno_ctrl_pd", "pheno_3c", "pheno_pd_t2d", "pheno_3c", "pheno_ctrl_pd", "pheno_pd_t2d")) %>%
bind_rows(data.frame(names = c("CTRL", "T2D"), colors = c('#dcd0c0ff', '#b7ceceff'), class = c("pheno_pd_t2d", "pheno_ctrl_pd"), stringsAsFactors = F))
# mark variables to use as response for various analyses
numeric.response <- c("bmi", "hba1c", "gsis")
discrete.response <- c("pheno_3c", "pheno_ctrl_pd", "pheno_pd_t2d")
exclude.cols <- c("id", "blood.group")
# mark variables account for in the various analyses
numeric.covars <- sapply(covars, class) %>%
as.data.frame() %>%
rename("class" = ".") %>%
rownames_to_column("column") %>%
filter(!(column %in% exclude.cols) & class == "numeric") %>%
pull(column)
discrete.covars <- sapply(covars, class) %>%
as.data.frame() %>%
rename("class" = ".") %>%
rownames_to_column("column") %>%
filter(!(column %in% c(exclude.cols, discrete.response)) & class == "character") %>%
pull(column)
# linear regressions using the numeric responses and all other covars as predictors
numeric.response.p <- bind_rows(lapply(numeric.response, function(resp.var){
tmp.resp <- pull(covars, resp.var)
select(covars, all_of(c(numeric.covars, discrete.covars, discrete.response))) %>%
map(function(tmp.pred) {
if (identical(tmp.pred, tmp.resp)) {
NA
} else {
modelobject <- lm(tmp.resp ~ tmp.pred)
f <- summary(modelobject)$fstatistic
p <- pf(f[1],f[2],f[3],lower.tail=F)
attributes(p) <- NULL
formatC(p, format = "e", digits = 1)
}})})) %>%
mutate(Numeric.Response = numeric.response) %>%
column_to_rownames("Numeric.Response") %>%
t() %>%
as.data.frame() %>%
rownames_to_column("Variable")
colnames(numeric.response.p) <- c(colnames(numeric.response.p)[1], paste("P (", tail(colnames(numeric.response.p), -1), ")", sep = ""))
# Mann_whitney U test for pheno_3c discrete responses
discrete.response.qualitative.3c.num <- select(covars, pheno_3c, all_of(numeric.covars)) %>%
gather(key = Variable, value = value, -pheno_3c) %>%
group_by(pheno_3c, Variable) %>%
summarise(value = list(value)) %>%
spread(pheno_3c, value) %>%
rename("mNA" = "<NA>") %>%
group_by(Variable) %>%
mutate(num_CTRL = length(unlist(CTRL)),
num_PD = length(unlist(PD)),
num_T2D = length(unlist(T2D)),
num_mNA = length(unlist(mNA)),
mean_CTRL = mean(x = unlist(CTRL), na.rm = T),
sd_CTRL = sd(x = unlist(CTRL), na.rm = T),
mean_PD = mean(x = unlist(PD), na.rm = T),
sd_PD = sd(x = unlist(PD), na.rm = T),
mean_T2D = mean(x = unlist(T2D), na.rm = T),
sd_T2D = sd(x = unlist(T2D), na.rm = T),
mean_mNA = mean(x = unlist(mNA), na.rm = T),
sd_mNA = sd(x = unlist(mNA), na.rm = T)) %>%
mutate_at(.vars = vars(num_CTRL:sd_mNA), .funs = list(~(ifelse(is.infinite(.), NA, ifelse(is.nan(.), NA, .))))) %>%
mutate(mean_sd_CTRL = paste(round(x = mean_CTRL, digits = 1), "?", round(x = sd_CTRL, digits = 1)),
mean_sd_PD = paste(round(x = mean_PD, digits = 1), "?", round(x = sd_PD, digits = 1)),
mean_sd_T2D = paste(round(x = mean_T2D, digits = 1), "?", round(x = sd_T2D, digits = 1)),
mean_sd_mNA = ifelse(is.na(mean_mNA) & is.na(sd_mNA), NA, paste(round(x = mean_mNA, digits = 1), "?", round(x = sd_mNA, digits = 1)))) %>%
mutate(p_CTRL_PD_T2D = formatC(x = kruskal.test(x = c(unlist(CTRL), unlist(PD), unlist(T2D)),
g = as.factor(c(rep("CTRL", length(unlist(CTRL))), rep("PD", length(unlist(PD))), rep("T2D", length(unlist(T2D))))))$p.value,
format = "e", digits = 1),
p_CTRL_PD = formatC(x = pairwise.wilcox.test(x = c(unlist(CTRL), unlist(PD), unlist(T2D)),
g = c(rep("CTRL", length(unlist(CTRL))), rep("PD", length(unlist(PD))), rep("T2D", length(unlist(T2D)))),
p.adjust.method = "none", paired = FALSE, na.action = na.omit)$p.value[1,1],
format = "e", digits = 1),
p_CTRL_T2D = formatC(x = pairwise.wilcox.test(x = c(unlist(CTRL), unlist(PD), unlist(T2D)),
g = c(rep("CTRL", length(unlist(CTRL))), rep("PD", length(unlist(PD))), rep("T2D", length(unlist(T2D)))),
p.adjust.method = "none", paired = FALSE, na.action = na.omit)$p.value[2,1],
format = "e", digits = 1),
p_PD_T2D = formatC(x = pairwise.wilcox.test(x = c(unlist(CTRL), unlist(PD), unlist(T2D)),
g = c(rep("CTRL", length(unlist(CTRL))), rep("PD", length(unlist(PD))), rep("T2D", length(unlist(T2D)))),
p.adjust.method = "none", paired = FALSE, na.action = na.omit)$p.value[2,2],
format = "e", digits = 1)) %>%
select(Variable, starts_with("num_"), starts_with("mean_sd"), starts_with("p_"))
# Fraction of males and females
discrete.response.qualitative.3c.disc <- select(covars, pheno_3c, all_of(discrete.covars)) %>%
gather(key = Variable, value = value, -pheno_3c) %>%
group_by(pheno_3c, Variable) %>%
summarise(value = list(value)) %>%
spread(pheno_3c, value) %>%
rename("mNA" = "<NA>") %>%
group_by(Variable) %>%
mutate(across(.fns = list(num = ~length(unlist(.)),
mean_sd = ~paste(as.data.frame(table(factor(x = unlist(.)), useNA = "ifany")) %>% arrange(Var1) %>% mutate(Str = paste(Freq, Var1)) %>% pull(Str), collapse = " / ")),
.names = "{fn}_{col}"),
p_CTRL_PD_T2D = NA,
p_CTRL_PD = NA,
p_CTRL_T2D = NA,
p_PD_T2D = NA) %>%
select(Variable, starts_with("num_"), starts_with("mean_sd"), starts_with("p_"))
# combine results from above
discrete.response.qualitative.3c <- bind_rows(discrete.response.qualitative.3c.num, discrete.response.qualitative.3c.disc)
rm(discrete.response.qualitative.3c.num, discrete.response.qualitative.3c.disc)
discrete.response.qualitative.ctrl_pd.num <- select(covars, pheno_ctrl_pd, all_of(numeric.covars)) %>%
gather(key = Variable, value = value, -pheno_ctrl_pd) %>%
group_by(pheno_ctrl_pd, Variable) %>%
summarise(value = list(value)) %>%
spread(pheno_ctrl_pd, value) %>%
group_by(Variable) %>%
mutate(num_CTRL_PD = length(unlist(CTRL_PD)),
mean_CTRL_PD = mean(x = unlist(CTRL_PD), na.rm = T),
sd_CTRL_PD = sd(x = unlist(CTRL_PD), na.rm = T),
mean_sd_CTRL_PD = paste(round(x = mean_CTRL_PD, digits = 1), "?", round(x = sd_CTRL_PD, digits = 1))) %>%
mutate(p_CTRL_PD__T2D = formatC(x = pairwise.wilcox.test(x = c(unlist(CTRL_PD), unlist(T2D)),
g = c(rep("CTRL_PD", length(unlist(CTRL_PD))), rep("T2D", length(unlist(T2D)))),
p.adjust.method = "none", paired = FALSE, na.action = na.omit)$p.value[1,1],
format = "e", digits = 1)) %>%
select(Variable, starts_with("num_"), starts_with("mean_sd"), starts_with("p_"))
discrete.response.qualitative.ctrl_pd.disc <- select(covars, pheno_ctrl_pd, all_of(discrete.covars)) %>%
gather(key = Variable, value = value, -pheno_ctrl_pd) %>%
group_by(pheno_ctrl_pd, Variable) %>%
summarise(value = list(value)) %>%
spread(pheno_ctrl_pd, value) %>%
group_by(Variable) %>%
mutate(across(.fns = list(num = ~length(unlist(.)),
mean_sd = ~paste(as.data.frame(table(factor(x = unlist(.)), useNA = "ifany")) %>% arrange(Var1) %>% mutate(Str = paste(Freq, Var1)) %>% pull(Str), collapse = " / ")),
.names = "{fn}_{col}"),
p_CTRL_PD__T2D = NA) %>%
select(colnames(discrete.response.qualitative.ctrl_pd.num))
discrete.response.qualitative.ctrl_pd <- bind_rows(discrete.response.qualitative.ctrl_pd.num, discrete.response.qualitative.ctrl_pd.disc)
rm(discrete.response.qualitative.ctrl_pd.num, discrete.response.qualitative.ctrl_pd.disc)
discrete.response.qualitative.pd_t2d.num <- select(covars, pheno_pd_t2d, all_of(numeric.covars)) %>%
gather(key = Variable, value = value, -pheno_pd_t2d) %>%
group_by(pheno_pd_t2d, Variable) %>%
summarise(value = list(value)) %>%
spread(pheno_pd_t2d, value) %>%
group_by(Variable) %>%
mutate(num_PD_T2D = length(unlist(PD_T2D)),
mean_PD_T2D = mean(x = unlist(PD_T2D), na.rm = T),
sd_PD_T2D = sd(x = unlist(PD_T2D), na.rm = T),
mean_sd_PD_T2D = paste(round(x = mean_PD_T2D, digits = 1), "?", round(x = sd_PD_T2D, digits = 1))) %>%
mutate(p_PD_T2D__CTRL = formatC(x = pairwise.wilcox.test(x = c(unlist(PD_T2D), unlist(CTRL)),
g = c(rep("PD_T2D", length(unlist(PD_T2D))), rep("CTRL", length(unlist(CTRL)))),
p.adjust.method = "none", paired = FALSE, na.action = na.omit)$p.value[1,1],
format = "e", digits = 1)) %>%
select(Variable, starts_with("num_"), starts_with("mean_sd"), starts_with("p_"))
discrete.response.qualitative.pd_t2d.disc <- select(covars, pheno_pd_t2d, all_of(discrete.covars)) %>%
gather(key = Variable, value = value, -pheno_pd_t2d) %>%
group_by(pheno_pd_t2d, Variable) %>%
summarise(value = list(value)) %>%
spread(pheno_pd_t2d, value) %>%
group_by(Variable) %>%
mutate(across(.fns = list(num = ~length(unlist(.)),
mean_sd = ~paste(as.data.frame(table(factor(x = unlist(.)), useNA = "ifany")) %>% arrange(Var1) %>% mutate(Str = paste(Freq, Var1)) %>% pull(Str), collapse = " / ")),
.names = "{fn}_{col}"),
p_PD_T2D__CTRL = NA) %>%
select(colnames(discrete.response.qualitative.pd_t2d.num))
discrete.response.qualitative.pd_t2d <- bind_rows(discrete.response.qualitative.pd_t2d.num, discrete.response.qualitative.pd_t2d.disc)
rm(discrete.response.qualitative.pd_t2d.num, discrete.response.qualitative.pd_t2d.disc)
discrete.response.qualitative <- left_join(discrete.response.qualitative.3c, discrete.response.qualitative.ctrl_pd, by = 'Variable') %>%
left_join(discrete.response.qualitative.pd_t2d, by = 'Variable') %>%
select(Variable, num_CTRL, num_PD, num_T2D, num_mNA, mean_sd_CTRL, mean_sd_PD, mean_sd_CTRL_PD, mean_sd_T2D, mean_sd_PD_T2D, mean_sd_mNA, p_CTRL_PD, p_PD_T2D, p_CTRL_T2D, p_CTRL_PD_T2D, p_CTRL_PD__T2D, p_PD_T2D__CTRL)
colnames(discrete.response.qualitative) <- c(head(colnames(discrete.response.qualitative), 1), unlist(lapply(strsplit(x = tail(colnames(discrete.response.qualitative), -1), split = "_", fixed = T), function(i){
if (i[1] == "num") {
paste(tail(i, -1), collapse = "-")
} else if (i[1] == "mean") {
paste(i[1], " ? ", i[2], " (", paste(tail(i, -2), collapse = "-"), ")", sep = "")
} else if (i[1] == "p") {
paste("P (", paste(tail(i, -1), collapse = "-"), ")", sep = "")
}
})))
rm(discrete.response.qualitative.3c, discrete.response.qualitative.ctrl_pd, discrete.response.qualitative.pd_t2d)
covars.stats <- full_join(discrete.response.qualitative, numeric.response.p, by = 'Variable') %>%
ungroup() %>%
mutate(Variable = case_when(
Variable == "age" ~ "Age",
Variable == "bmi" ~ "BMI",
Variable == "cit.min" ~ "CIT (min)",
Variable == "days.in.icu" ~ "Days in ICU",
Variable == "gsis" ~ "GSIS",
Variable == "hba1c" ~ "HbA1c",
Variable == "islets.purity" ~ "Islets purity",
Variable == "antidiuretics" ~ "Antidiuretics",
Variable == "antiepileptic" ~ "Antiepileptics",
Variable == "antiinfectives" ~ "Antiinfectives",
Variable == "bba" ~ "Beta blocking agents",
Variable == "bsps" ~ "Blood substitutes and perfusion solutions",
Variable == "cardiac.therapy" ~ "Cardiac therapy",
Variable == "diuretics" ~ "Diuretics",
Variable == "glucocorticoids" ~ "Glucocorticoids",
Variable == "insulin" ~ "Insulin",
Variable == "sex" ~ "Sex",
Variable == "pheno_3c" ~ "CTRL v PD v T2D",
Variable == "pheno_ctrl_pd" ~ "CTRL+PD v T2D",
Variable == "pheno_pd_t2d" ~ "CTRL v PD+T2D",
TRUE ~ "TMP")) %>%
mutate_all(as.character) %>%
mutate_all(.funs = list(~replace_na(., "-"))) %>%
mutate_all(.funs = list(~gsub("?", "±", ., fixed = T)))
covars.stats <- magrittr::set_colnames(covars.stats, gsub("mNA", "NA", x = gsub("?", "±", x = gsub("-", " v ", x = gsub("--", "+", x = colnames(covars.stats), fixed = T), fixed = T), fixed = T), fixed = T))
if (file.exists(paste(output.directory, "covars.xlsx", sep = ""))) {
file.remove(paste(output.directory, "covars.xlsx", sep = ""))
}
write.to.excel(df_ = covars, file_ = paste(output.directory, "covars.xlsx", sep = ""), sheet_ = "covars", append_ = FALSE)
write.to.excel(df_ = covars.stats, file_ = paste(output.directory, "covars.xlsx", sep = ""), sheet_ = "stats", append_ = TRUE)
rm(numeric.response, discrete.response, exclude.cols, numeric.covars, discrete.covars, discrete.response.qualitative, numeric.response.p)
kable(covars.stats) %>%
kable_styling(full_width = T, bootstrap_options = "striped", font_size = 8.5) %>%
column_spec(1, bold = T, border_right = T)
rm(covars.stats)
```
_NB:_ there are a few not available (NA) values for some clinical variables.
## Reading, pre-processing and shaping the MS proteomics data
Read raw MS proteomics data. Create an index that contains the collection of unique custom identifiers for each protein, the UniProtKB identifier, the gene name and a brief description of it.
_NB:_ UniProtKB identifiers are resolved so that unique ones are maintained for downstream analysis. This is based on the order of the identifiers provided in the input file and whether they exist in the universe of identifiers from the GO terms and the pathways. For detailed description see the manuscript and explore the code
Pre-defined samples are filtered out, while proteins are filtered if <=`r paste(min.samples.threshold * 100, "%", sep = "")` of them is missing from samples.
The final object is a large list that contains all the data that allow the downstream analysis.
```{r reading.the.data.funs, echo=FALSE, warning=FALSE, message=FALSE, results='hide'}
# read input files
read.dia <- function(tissue_, file_) {
dia_ <- read_excel(path = file_, sheet = tissue_)
unique.tissue <- unique(unlist(lapply(strsplit(x = tail(colnames(dia_), -3), split = "_", fixed = T), function(x) tail(x = head(x = x, n = 2), n = 1))))
if (length(unique.tissue) > 1 || unique.tissue != tissue_) {
stop("Error: No unique tissue for colnames or tissue from colnames does not match given tissue name!")
}
colnames(dia_) <- c("Protein_Accession", "Gene", "Protein_Description", unlist(lapply(strsplit(x = tail(colnames(dia_), -3), split = "_", fixed = T), head, 1)))
return(mutate(dia_, AZpID = ifelse(row_number() < 10, paste(tissue_, ".AZp000", row_number(), sep = ""),
ifelse(row_number() < 100, paste(tissue_, ".AZp00", row_number(), sep = ""),
ifelse(row_number() < 1000, paste(tissue_, ".AZp0", row_number(), sep = ""), paste(tissue_, ".AZp", row_number(), sep = ""))))) %>%
select(AZpID, everything()))
}
# create index for all proteins
get.dia.index <- function(dia_, tissue_) {
return(dia_ %>%
select(AZpID, Gene, Protein_Accession, Protein_Description) %>%
rename(!!paste("AZpID", tissue_, sep = ".") := "AZpID",
!!paste("Gene", tissue_, sep = ".") := "Gene",
!!paste("Protein_Description", tissue_, sep = ".") := "Protein_Description"))
}
# shape proteomics measurements for downstream analysis
prepare.dia <- function(dia_, tissue_, dia.index_) {
azpid.name <- paste("AZpID", tissue_, sep = ".")
return(dia_ %>%
rename(!!azpid.name := "AZpID") %>%
left_join(select(dia.index_, AZpID, all_of(azpid.name)), by = azpid.name) %>%
select(AZpID, everything()) %>%
select(-all_of(azpid.name), -Gene, -Protein_Accession, -Protein_Description) %>%
mutate_all(as.character) %>%
map_df(~parse_guess(.)))
}
# resolve multiplicity of Gene names
unique.gene.id <- function(azpid.index_) {
list.of.geneid <- read.delim(file = gene.id.file, header = F, sep = "\t", stringsAsFactors = F) %>%
pull(V1)
azpid.to.remove <- azpid.index_ %>%
filter(is.na(Gene) | nchar(Gene) == 0) %>%
pull(AZpID)
spread.azpid.index <- azpid.index_ %>%
filter(nchar(Gene) > 0) %>%
filter(!is.na(Gene)) %>%
separate_rows(Gene, sep = ";") %>%
mutate(in.list = ifelse(Gene %in% list.of.geneid, "YES", "NO")) %>%
group_by(AZpID) %>%
filter(nchar(Gene) > 0) %>%
filter(!is.na(Gene)) %>%
distinct(Gene, .keep_all = T) %>%
mutate(ncol = n(),
indx = row_number())
id.in.list <- spread.azpid.index %>%
filter(in.list == "YES") %>%
slice(1) %>%
ungroup()
id.in.list.add <- spread.azpid.index %>%
filter_at(.vars = vars(in.list), all_vars(. == "NO")) %>%
ungroup() %>%
filter(!(AZpID %in% id.in.list$AZpID)) %>%
separate_rows(Gene, sep = "-") %>%
mutate(in.list = ifelse(Gene %in% list.of.geneid, "YES", "NO")) %>%
group_by(AZpID) %>%
filter(nchar(Gene) > 0) %>%
filter(!is.na(Gene)) %>%
distinct(Gene, .keep_all = T) %>%
mutate(ncol = n(),
indx = row_number()) %>%
filter(in.list == "YES") %>%
slice(1) %>%
ungroup()
id.not.in.list <- spread.azpid.index %>%
filter_at(.vars = vars(in.list), all_vars(. == "NO")) %>%
slice(1) %>%
ungroup() %>%
filter(!(AZpID %in% c(id.in.list$AZpID, id.in.list.add$AZpID)))
unique.azpid <- bind_rows(id.in.list, id.in.list.add, id.not.in.list) %>%
arrange(AZpID) %>%
select(-in.list, -ncol, -indx)
duplicated.gene.id <- unique.azpid %>%
arrange(Gene, AZpID) %>%
group_by(Gene) %>%
mutate(ncol = n()) %>%
filter(ncol > 1) %>%
mutate(tmp.AZpID = AZpID[1]) %>%
ungroup()
azpid.index <- bind_rows(unique.azpid %>% anti_join(select(duplicated.gene.id, AZpID), by = 'AZpID'),
unique.azpid %>% inner_join(select(duplicated.gene.id, tmp.AZpID) %>% distinct(tmp.AZpID), by = c("AZpID" = "tmp.AZpID")))
return(list(azpindex = azpid.index, duplicated = duplicated.gene.id, toremove = azpid.to.remove))
}
# filter proteins based on the % of samples the protein was identified on
filter.proteins <- function(dia_, min.samples.thresh_) {
keep.overall <- dia_ %>%
column_to_rownames("AZpID") %>%
transpose() %>%
lapply(., function(i) 1 - (sum(is.na(i)) / length(i))) %>%
unlist() %>%
enframe(., value = "V1") %>%
mutate(AZpID = dia_$AZpID) %>%
filter(V1 >= min.samples.thresh_)
return(filter(dia_, AZpID %in% keep.overall$AZpID))
}
# filter pre-defined samples
filter.samples <- function(dia_, samples.to.remove.manually_) {
keep.me <- tail(colnames(dia_), -1)
keep.me <- keep.me[!(keep.me %in% samples.to.remove.manually_)]
return(dia_ %>% select(AZpID, all_of(keep.me)))
}
# update index so that the custom identifiers refer to the same protein across tissues
update.index <- function(dia.index.all_, dia.filtered_, tissues_) {
s <- lapply(tissues_, function(i) dia.index.all_ %>%
select(AZpID, all_of(paste("AZpID", i, sep = "."))) %>%
magrittr::set_names(c("AZpID", "tissue")) %>%
mutate(!!i := ifelse(AZpID %in% dia.filtered_[[i]]$AZpID, tissue, NA)) %>%
select(-tissue))
si <- plyr::join_all(dfs = s, by = 'AZpID', type = 'left') %>%
as_tibble()
colnames(si) <- c(head(colnames(si), 1), paste("AZpID", tail(colnames(si), -1), sep = "."))
return(left_join(select(dia.index.all_, AZpID, Gene, Protein_Accession, Protein_Description), si, by = 'AZpID'))
}
# average proteins that have identified twice within the same tissue
average.duplicate.protein.accessions <- function(dia_, dpa_) {
average.of.duplicated <- dia_ %>%
inner_join(select(dpa_, AZpID, tmp.AZpID), by = 'AZpID') %>%
arrange(tmp.AZpID) %>%
group_by(tmp.AZpID) %>%
summarise_at(.vars = vars(starts_with("p")), .funs = list(~mean(., na.rm = T))) %>%
mutate_all(.funs = list(~ifelse(is.nan(.), NA, .))) %>%
rename("AZpID" = "tmp.AZpID")
return(dia_ %>%
anti_join(select(dpa_, AZpID), by = 'AZpID') %>%
bind_rows(average.of.duplicated))
}
```
```{r reading.the.data, echo=FALSE, warning=FALSE, message=FALSE, results='hide'}
dia <- list() # dia is the list that will keep all the data
dia$raw <- lapply(tissues, read.dia, proteomics.file) # read the raw data for all the tissues from the input excel file
names(dia$raw) <- tissues # name the list of raw data according to the tissues
# make the index for Gene names, protein accessions etc.
dia$index <- list() # create an index list that will keep matches of IDs to protein accessions and gene names
dia$index$all <- lapply(seq_along(dia$raw), function(i) get.dia.index(dia_ = dia$raw[[i]], tissue_ = names(dia$raw)[i])) # retrieve all protein IDs
names(dia$index$all) <- tissues #name the list
dia$index$all <- plyr::join_all(dfs = dia$index$all, by = 'Protein_Accession', type = 'full') %>%
unite(Gene, starts_with("Gene"), sep = '|', na.rm = T, remove = F) %>%
unite(Protein_Description, starts_with("Protein_Description"), sep = '|', na.rm = T, remove = F) %>%
mutate(Gene = unlist(lapply(lapply(strsplit(x = Gene, split = "|", fixed = T), unique), paste, collapse = "|")),
Protein_Description = unlist(lapply(lapply(strsplit(x = Protein_Description, split = "|", fixed = T), unique), paste, collapse = "|")),
AZpID = case_when(
row_number() < 10 ~ paste("AZp000", row_number(), sep = ""),
row_number() < 100 ~ paste("AZp00", row_number(), sep = ""),
row_number() < 1000 ~ paste("AZp0", row_number(), sep = ""),
TRUE ~ paste("AZp", row_number(), sep = ""))) %>%
select(AZpID, Gene, Protein_Accession, Protein_Description, starts_with("AZpID.")) %>%
mutate_at(.vars = vars(starts_with("AZpID.")), .funs = list(Present = ~ifelse(!is.na(.), "X", NA))) %>%
rename_at(.vars = vars(ends_with("Present")), .funs = list(~paste("Present", unlist(lapply(strsplit(x = unlist(lapply(strsplit(x = ., split = "_", fixed = T), head, 1)), split = ".", fixed = T), function(x) paste(tail(x, 1), collapse = "."))), sep = "."))) %>%
as_tibble()
dia$index$AZpID <- dia$index$all %>% select(AZpID, Gene, Protein_Accession, Protein_Description, starts_with("AZpID."))
# prepare the final set of raw data in a format proteins x samples
dia$raw <- lapply(seq_along(dia$raw), function(i) prepare.dia(dia_ = dia$raw[[i]], tissue_ = names(dia$raw)[i], dia.index_ = dia$index$AZpID))
names(dia$raw) <- tissues
# fix unique protein accessions
tmp.azpid <- unique.gene.id(azpid.index_ = dia$index$AZpID)
dia$index$AZpID <- tmp.azpid$azpindex
dia$index$all <- dia$index$all %>%
select(-Protein_Accession) %>%
inner_join(select(dia$index$AZpID, AZpID, Protein_Accession), by = 'AZpID') %>%
select(AZpID:Gene, Protein_Accession, everything())
dia$index$index <- dia$index$all %>% select(AZpID, Gene, Protein_Accession, Protein_Description)
dia$index$Present <- dia$index$all %>% select(AZpID, Gene, Protein_Accession, Protein_Description, starts_with("Present"))
# filter the data set for proteins in min.samples.threshold
dia$raw.filtered <- lapply(dia$raw, filter.proteins, min.samples.thresh_ = min.samples.threshold)
# filter samples with large deviations from the average median and samples that we know are outliers
dia$raw.filtered <- lapply(dia$raw.filtered, filter.samples, samples.to.remove.manually_ = samples.to.remove.manually)
# remove the temporary AZpID per tissue we set when we first read the data into the unique global IDs
dia$index$all <- dia$index$all %>%
mutate_at(.vars = vars(starts_with("AZpID.")), .funs = list(~ifelse(!is.na(.), paste(unlist(lapply(strsplit(x = ., split = ".", fixed = T), head, 1)), AZpID, sep = "."), NA)))
if (file.exists(paste(output.directory, "proteins_in_tissues.xlsx", sep = ""))) {
file.remove(paste(output.directory, "proteins_in_tissues.xlsx", sep = ""))
}
write.to.excel(df_ = dia$index$all, file_ = paste(output.directory, "proteins_in_tissues.xlsx", sep = ""), sheet_ = "Sheet1", append_ = F)
dia$index$AZpID <- update.index(dia.index.all_ = dia$index$all %>% select(AZpID, Gene, Protein_Accession, Protein_Description, starts_with("AZpID.")), dia.filtered_ = dia$raw.filtered , tissues_ = tissues)
# average duplicate rotein accessions in tissue datasets
dia$raw.filtered <- lapply(dia$raw.filtered, average.duplicate.protein.accessions, dpa_ = tmp.azpid$duplicated)
# store other variables/results
dia$other <- list()
# make data frames proteins x samples and samples x proteins
dia$raw.filtered.unique.proteins <- merge.unique.proteins.dia(dia.all_ = dia$raw.filtered)
dia$raw.filtered.unique.samples <- merge.unique.samples.dia(dia.all_ = dia$raw.filtered)
dia$other$raw.unique.proteins <- merge.unique.proteins.dia(dia.all_ = dia$raw)
dia$other$raw.unique.samples <- merge.unique.samples.dia(dia.all_ = dia$raw)
# cleanup
rm(read.dia, get.dia.index, prepare.dia, unique.protein.accession, filter.proteins, filter.samples, update.index, average.duplicate.protein.accessions, dia.tmp, tmp.azpid)
```
## Exploring identified proteins in tissues and phenotypes
The table below shows the number of proteins identified in each tissue, the total number of identified proteins and the unique number of identified proteins before and after filtering. The same is shown for samples, including the dustribution of samples over phenotypes.
```{r show.table.protein.stats, echo=FALSE, message=FALSE, warning=FALSE, results='asis'}
# calculate distribution of samples over phenotypes
sample.distribution.in.phenos <- plyr::join_all(dfs = lapply(seq_along(dia$raw.filtered), function(i) covars %>% filter(id %in% tail(colnames(dia$raw.filtered[[i]]), -1)) %>% group_by(pheno_3c) %>% summarise(!!names(dia$raw.filtered)[i] := n())),
by = 'pheno_3c', type = 'full') %>%
as_tibble() %>%
mutate(pheno_3c = replace(pheno_3c, is.na(pheno_3c), "NA"),
Total = rowSums(.[2:6]),
Unique = qlcMatrix::rowMax(.[2:6] %>% as.matrix()) %>% as.numeric()) %>%
column_to_rownames("pheno_3c") %>%
t() %>%
as.data.frame() %>%
rownames_to_column("Tissue")
# prepare data frame
number.proteins.per.tissue <- data.frame(Tissue = c(tissues, "Total", "Unique"),
Samples.Raw = c(unname(unlist(lapply(dia$raw, ncol))) - 1, ncol(dia$other$raw.unique.proteins) - 1, nrow(dia$other$raw.unique.samples)),
Proteins.Raw = c(unname(unlist(lapply(dia$raw, nrow))), ncol(dia$other$raw.unique.samples) - 1, nrow(dia$other$raw.unique.proteins)),
Missing.Raw = c(unlist(lapply(dia$raw, function(i) paste(round(x = ((sum(is.na(i %>% column_to_rownames("AZpID"))) * 100) / (ncol(i %>% column_to_rownames("AZpID")) * nrow(i %>% column_to_rownames("AZpID")))), digits = 2), "%", sep = ""))), paste(round(x = (sum(unlist(lapply(dia$raw, function(i) sum(is.na(i %>% column_to_rownames("AZpID")))))) * 100) / sum(unlist(lapply(dia$raw, function(i) ncol(i %>% column_to_rownames("AZpID")) * nrow(i %>% column_to_rownames("AZpID"))))), digits = 2), "%", sep = ""), ""),
Samples.Filtered = c(unname(unlist(lapply(dia$raw.filtered, ncol))) - 1, ncol(dia$raw.filtered.unique.proteins) - 1, nrow(dia$raw.filtered.unique.samples)),
Proteins.Filtered = c(unname(unlist(lapply(dia$raw.filtered, nrow))), ncol(dia$raw.filtered.unique.samples) - 1, nrow(dia$raw.filtered.unique.proteins)),
Missing.Filtered = c(unlist(lapply(dia$raw.filtered, function(i) paste(round(x = ((sum(is.na(i %>% column_to_rownames("AZpID"))) * 100) / (ncol(i %>% column_to_rownames("AZpID")) * nrow(i %>% column_to_rownames("AZpID")))), digits = 2), "%", sep = ""))), paste(round(x = (sum(unlist(lapply(dia$raw.filtered, function(i) sum(is.na(i %>% column_to_rownames("AZpID")))))) * 100) / sum(unlist(lapply(dia$raw, function(i) ncol(i %>% column_to_rownames("AZpID")) * nrow(i %>% column_to_rownames("AZpID"))))), digits = 2), "%", sep = ""), ""),
Details.Samples.Removed = c(unlist(lapply(seq_along(tissues), function(i) paste(filter(covars, id %in% colnames(dia$raw[[i]])[!(colnames(dia$raw[[i]]) %in% colnames(dia$raw.filtered[[i]]))]) %>% mutate(pheno_3c = replace(pheno_3c, is.na(pheno_3c), "NA"), tmp = paste(id, " (", pheno_3c, ")", sep = "")) %>% pull(tmp), collapse = ", "))), "", paste(data.frame(s = unlist(lapply(seq_along(tissues), function(i) colnames(dia$raw[[i]])[!(colnames(dia$raw[[i]]) %in% colnames(dia$raw.filtered[[i]]))])), stringsAsFactors = F) %>% group_by(s) %>% summarise(n = n()) %>% mutate(p = paste(s, " (", n, ")", sep = "")) %>% pull(p), collapse = ", ")),
stringsAsFactors = F) %>%
mutate(Samples.Removed = Samples.Raw - Samples.Filtered,
Percentage.Samples.Removed = paste(round((Samples.Removed * 100) / Samples.Raw, digits = 2), "%", sep = ""),
Proteins.Removed = Proteins.Raw - Proteins.Filtered,
Percentage.Proteins.Removed = paste(round((Proteins.Removed * 100) / Proteins.Raw, digits = 2), "%", sep = "")) %>%
select(Tissue, ends_with("Raw"), ends_with("Filtered"), Samples.Removed, Percentage.Samples.Removed, Details.Samples.Removed, Proteins.Removed, Percentage.Proteins.Removed) %>%
left_join(sample.distribution.in.phenos, by = 'Tissue')
if (file.exists(paste(output.directory, "number_of_proteins_per_tissue.xlsx", sep = ""))) {
file.remove(paste(output.directory, "number_of_proteins_per_tissue.xlsx", sep = ""))
}
write.to.excel(df_ = number.proteins.per.tissue, file_ = paste(output.directory, "number_of_proteins_per_tissue.xlsx", sep = ""), sheet_ = "Sheet1", append_ = F)
kable(number.proteins.per.tissue, col.names = unlist(lapply(strsplit(x = colnames(number.proteins.per.tissue), split = ".", fixed = T), head, 1)), align = 'r') %>%
kable_styling(full_width = F, bootstrap_options = "striped") %>%
column_spec(1, bold = T, border_right = T) %>%
row_spec(6:7, bold = T) %>%
add_header_above(c(" " = 1, "Raw" = 3, "Filtered" = 3, "Removed Samples" = 3, "Removed Proteins" = 2, "Phenotypes" = 4))
dia$other$raw.filtered.unique.proteins <- dia$raw.filtered.unique.proteins
dia$raw.filtered.unique.proteins <- NULL
dia$other$raw.filtered.unique.samples <- dia$raw.filtered.unique.samples
dia$raw.filtered.unique.samples <- NULL
rm(number.proteins.per.tissue)
```
The plot below shows the distributon of samples over phenotypes for each tissue in a doughnut plot. The inner doughnut represents samples in _CTRL_, _PD_ and _T2D_. The next outer doughnut represents the merged class of _CTRL_ and _PD_, while the outmost one the merged class of _PD_ and _T2D_.
```{r show.plots.protein.stats, echo=FALSE, message=FALSE, warning=FALSE, results='hide', fig.show='asis', fig.keep='last', fig.width=9, fig.height=7}
sample.distribution.in.phenos %>%
filter(Tissue %in% tissues) %>%
select(Tissue, all_of(c("CTRL", "PD", "T2D"))) %>%
mutate(Total = rowSums(.[2:4]),
CTRL_PD = CTRL + PD,
CTRL_PD__T2D = T2D,
PD_T2D = PD + T2D,
PD_T2D__CTRL = CTRL) %>%
select(Tissue, CTRL, PD, T2D, CTRL_PD, CTRL_PD__T2D, PD_T2D, PD_T2D__CTRL, Total) %>%
mutate_at(.vars = vars(CTRL:PD_T2D__CTRL), .funs = list(~((. * 100)/Total))) %>%
select(-Total) %>%
gather("Pheno", "Percentage", -Tissue) %>%
mutate(Resolution = case_when(
Pheno %in% c("CTRL", "PD", "T2D") ~ "3C",
Pheno %in% c("CTRL_PD", "CTRL_PD__T2D") ~ "2C2",
Pheno %in% c("PD_T2D", "PD_T2D__CTRL") ~ "2C1",
TRUE ~ "TMP"),
Resolution = replace(Resolution, Resolution == "TMP", NA)) %>%
mutate_all(as.character) %>%
map_df(~parse_guess(.)) %>%
select(Tissue, Resolution, Pheno, Percentage) %>%
arrange(Tissue, desc(Resolution), Pheno, Percentage) %>%
mutate(Pheno = factor(x = Pheno, levels = c("CTRL", "PD", "T2D", "CTRL_PD", "CTRL_PD__T2D", "PD_T2D__CTRL", "PD_T2D")),
Resolution = factor(x = Resolution, levels = c("3C", "2C2", "2C1")),
Tissue = factor(x = Tissue, levels = tissues),
Text.Percentage = paste(as.character(round(x = Percentage, digits = 0)), "%", sep = "")) %>%
ggplot(aes(x = Resolution, y = Percentage, fill = Pheno)) +
geom_bar(stat = "identity", color = "white", width = 0.7) +
scale_x_discrete(limits = c(" ", "3C", "2C2", "2C1")) +
geom_text(aes(label = Text.Percentage), position = position_stack(vjust = 0.5)) +
scale_fill_manual(values = setNames(object = phenotypes.colors$colors, nm = phenotypes.colors$names)) +
coord_polar(theta = "y", start = 0) +
facet_wrap(~ Tissue) +
theme_void()
rm(sample.distribution.in.phenos)
```
# Quality control (QC), normalization and imputation of the data
## QC of the data
The figure below shows the log~2~ FC of the median sample intensity over the average intensity of the medians of all samples. The left column shows the raw data (raw) and the right one for the fileterd set of median samples intensities deviating less than log~2~5 (filtered).
```{r plot.median.diff, echo=FALSE, message=FALSE, warning=FALSE, results='hide', fig.show='asis', fig.keep='last', fig.width=12, fig.height=16}
plot.median.diff <- function(dia_, tissue_, pheno.col_) {
return(select(dia_, -AZpID) %>%
summarise_all(.funs = list(~median(x = ., na.rm = TRUE))) %>%
t() %>%
as.data.frame() %>%
rename("median.intensity" = "V1") %>%
rownames_to_column("id") %>%
mutate(intensity.diff = abs(median.intensity) - mean(median.intensity),
id = as.character(id)) %>%
left_join(select(covars, id, all_of(pheno.col_)) %>% rename("pheno" = all_of(pheno.col_)), by = 'id') %>%
filter(!is.na(pheno)) %>%
arrange(intensity.diff) %>%
mutate(id = factor(id, levels = id)) %>%
ggplot(aes(x = id, y = intensity.diff, color = pheno, fill = pheno)) +
geom_bar(stat="identity") +
theme_minimal() +
# geom_hline(yintercept=c(-log2(5), log2(5)), linetype="dashed", color = "red") +
scale_y_continuous(breaks = seq(from = -5, to = 5, by = 1), labels = seq(from = -5, to = 5, by = 1), limits = c(-5, 5),
sec.axis = sec_axis(~sign(.)*(2^abs(.)), breaks = c(-25, -10, -5, 0, 5, 10, 25), labels = c(-25, -10, -5, 0, 5, 10, 25), name = "linear scale")) +
labs(title = tissue_, y = expression('log'[2]*'(sample'['avg']*') - log'[2]*'(total'['avg']*')')) +
scale_fill_manual(name = "Phenotype", values=setNames(object = phenotypes.colors$colors, nm = phenotypes.colors$names)) +
scale_color_manual(name = "Phenotype", values=setNames(object = phenotypes.colors$colors, nm = phenotypes.colors$names)) +
theme(axis.ticks.x = element_blank(),
axis.text.x = element_text(angle = 90),
axis.title.x = element_blank(),
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank(),
plot.title = element_text(hjust = 0.5)))
}
plot_grid(plot_grid(plotlist = lapply(seq_along(dia$raw), function(i) plot.median.diff(dia_ = dia$raw[[i]], tissue_ = names(dia$raw)[i], pheno.col_ = covars.pheno)), ncol = 1),
plot_grid(plotlist = lapply(seq_along(dia$raw.filtered), function(i) plot.median.diff(dia_ = dia$raw.filtered[[i]], tissue_ = names(dia$raw.filtered)[i], pheno.col_ = covars.pheno)), ncol = 1), ncol = 2,
labels = c("Raw", "Filtered"))
rm(plot.median.diff)
```
## Missingness of the data
The plots below show the characteristic of the missing data from the dataset. The left panels show the intensity distributions of the proteins that have missing values, and the right panels show the cumulative fraction of proteins with and without missing values. Every row represents a tissues.
```{r explore.missingness, echo=FALSE, message=FALSE, warning=FALSE, results='hide', fig.show='asis', fig.keep='last', fig.width=6, fig.height=10}
plot.miss <- function(dia_, tissue_) {
dia.tmp_ <- dia_ %>%
mutate(total.tmp = select(., -AZpID) %>% rowSums(., na.rm = T),
count = select(., -AZpID) %>% is.na(.) %>% `!` %>% rowSums(., na.rm = T),
total = length(colnames(.)) - 1,
mean = total.tmp / count,
miss = ifelse(count < total, TRUE, FALSE)) %>%
select(mean, miss) %>%
group_by(miss) %>%
arrange(mean) %>%
mutate(num = 1, cs = cumsum(num), cs_frac = cs/n()) %>%
ungroup()
# Plot the densities and cumalitive fractions for
# proteins with and without missing values
p1 <- ggplot(dia.tmp_, aes(mean, col = miss)) +
geom_density(na.rm = TRUE) +
xlim(10, 30) +
ylim(0, 0.5) +
labs(x = expression(log[2]~"Intensity"), y = "Density", title = tissue_) +
guides(col = guide_legend(title = "Missing values")) +
theme_minimal() +
theme(panel.grid.minor = element_blank())
p2 <- ggplot(dia.tmp_, aes(mean, cs_frac, col = miss)) +
geom_line() +
xlim(10, 30) +
ylim(0, 1) +
labs(x = expression(log[2]~"Intensity"), y = "Cumulative fraction", title = tissue_) +
guides(col = guide_legend(title = "Missing values")) +
theme_minimal() +
theme(panel.grid.minor = element_blank())
return(plot_grid(p1, p2, nrow = 1, labels = c(tissue_, "")))
}
plot_grid(plotlist = lapply(tissues, function(i) plot.miss(dia_ = dia$raw.filtered[[i]], tissue_ = names(dia$raw.filtered)[i])), ncol = 1)
rm(plot.miss)
```
## Normalization of the data
```{r normalizing.the.data, warning=FALSE, message=FALSE, echo=FALSE, results='hide'}
normalize.median <- function(dia_) {
return(mutate_at(dia_, .vars = vars(-AZpID), .funs = list(~(. - median(., na.rm = T)))))
}
dia$normalized <- lapply(dia$raw.filtered, normalize.median)
dia$other$normalized.unique.proteins <- merge.unique.proteins.dia(dia.all_ = dia$normalized)
dia$other$normalized.unique.samples <- merge.unique.samples.dia(dia.all_ = dia$normalized)
rm(normalize.median)
```
The figure below shows the raw (left) and normalized protein intensities (right). Normalization is applied to remove systematic differences in protein abundance due to different sample concentrations, or different amounts of sample loaded in the columns. Normalization is important, so that true differentially expressed proteins can be detected. To do this the z-score of the log~2~ transformed intensities is computed, which is updated by the average of the standard deviation of the log~2~ transformed intensities in all samples. After normalization all samples have a similar distribution.
```{r explore.normalization, echo=FALSE, message=FALSE, warning=FALSE, results='hide', fig.show='asis', fig.keep='last', fig.width=16, fig.height=35}
plot.intensity.distributions <- function(dia_, pheno.col_, tissue_, ylim.min_, ylim.max_) {
return(dia_ %>%
gather("variable", "value", -AZpID) %>%
left_join(select(covars, id, all_of(pheno.col_)) %>% rename("pheno" = all_of(pheno.col_)), by = c("variable" = "id")) %>%
filter(!is.na(pheno)) %>%
arrange(match(pheno, sort(unique(pheno))), variable) %>%
mutate(pheno = factor(x = pheno, levels = c(sort(unique(pheno)), "NA")),
variable = factor(x = variable, levels = unique(variable))) %>%
ggplot(aes(x = variable, y = value, color = pheno, fill = pheno)) +
labs(title = tissue_, y = expression('log'[2]*' intensity')) +
geom_boxplot(alpha = 0.5) +
theme_minimal() +
coord_flip() +
ylim(ylim.min_, ylim.max_) +
scale_fill_manual(name = "Phenotype", values=setNames(object = phenotypes.colors$colors, nm = phenotypes.colors$names)) +
scale_color_manual(name = "Phenotype", values=setNames(object = phenotypes.colors$colors, nm = phenotypes.colors$names)) +
theme(axis.ticks.y = element_blank(),
axis.title.y = element_blank(),
plot.title = element_text(hjust = 0.5)))
}
plot_grid(plot_grid(plotlist = lapply(tissues, function(i) plot.intensity.distributions(dia_ = dia$raw.filtered[[i]], pheno.col_ = covars.pheno, tissue_ = i, ylim.min_ = plyr::round_any(x = min(dia$other$raw.filtered.unique.proteins %>% select(-AZpID), na.rm = T), accuracy = 5, f = ceiling), ylim.max_ = plyr::round_any(x = max(dia$other$raw.filtered.unique.proteins %>% select(-AZpID), na.rm = T), accuracy = 5, f = ceiling))), ncol = 1),
plot_grid(plotlist = lapply(tissues, function(i) plot.intensity.distributions(dia_ = dia$normalized[[i]], pheno.col_ = covars.pheno, tissue_ = i, ylim.min_ = min(unlist(lapply(dia$normalized, function(i) plyr::round_any(x = min(i %>% select(-AZpID), na.rm = T), accuracy = 5, f = ceiling)))), ylim.max_ = max(unlist(lapply(dia$normalized, function(i) plyr::round_any(x = max(i %>% select(-AZpID), na.rm = T), accuracy = 5, f = ceiling)))))), ncol = 1),
ncol = 2, labels = c("Raw", "Normalized"))
rm(plot.intensity.distributions)
```
## Imputation of the data
The density plots below summarize the imputed data using QRILC and compare it to the non-imputed normalized set (top row). Every column repsesents one tissue and every colored line represents one sample.
```{r explore.imputation, echo=FALSE, message=FALSE, warning=FALSE, results='hide', fig.show='asis', fig.keep='last', fig.width=15, fig.height=9}
impute.and.plot <- function(dia_, tissue_) {
dia.imputed.QRILC <- imputeLCMD::impute.QRILC(dataSet.mvs = dia_ %>% column_to_rownames("AZpID"))
p1 <- dia_ %>%
gather("sample", "value", -AZpID) %>%
ggplot(data = ., aes(x = value, color = sample)) +
geom_density(na.rm = TRUE) +
labs(title = tissue_, y = ifelse(tissue_ == tissues[1], "Normalized", "")) +
xlim(-15, 15) +
ylim(0, 0.25) +
theme_minimal() +
theme(legend.position = "none",
axis.title.x = element_blank(),
panel.grid.minor = element_blank())
p2 <- dia.imputed.QRILC[[1]] %>%
rownames_to_column("AZpID") %>%
gather("sample", "value", -AZpID) %>%
ggplot(data = ., aes(x=value, color = sample)) +
geom_density() +
labs(y = ifelse(tissue_ == tissues[1], "QRILC Imputed", "")) +
xlim(-15, 15) +
ylim(0, 0.25) +
theme_minimal() +
theme(legend.position = "none",
axis.title.x = element_blank(),
panel.grid.minor = element_blank())
p <- plot_grid(p1, p2, ncol = 1)
return(list(QRILC = dia.imputed.QRILC[[1]] %>% rownames_to_column("AZpID"), plot = p))
}
imputation.result <- lapply(seq_along(tissues), function(i) impute.and.plot(dia_ = dia$normalized[[i]], tissue_ = names(dia$normalized)[i]))
dia$normalized.imputed <- lapply(imputation.result, function(i) i[["QRILC"]])
names(dia$normalized.imputed) <- tissues
dia$normalized.imputed.unique.proteins <- merge.unique.proteins.dia(dia.all_ = dia$normalized.imputed)
dia$normalized.imputed.unique.samples <- merge.unique.samples.dia(dia.all_ = dia$normalized.imputed)
if(file.exists(paste(output.directory, "normalized_imputed_dataset.xlsx", sep = ""))) {
file.remove(paste(output.directory, "normalized_imputed_dataset.xlsx", sep = ""))
}
lapply(seq_along(tissues), function(i) write.to.excel(df_ = dia$normalized.imputed[[i]] %>% left_join(dia$index$index, by = 'AZpID'), file_ = paste(output.directory, "normalized_imputed_dataset.xlsx", sep = ""), sheet_ = names(dia$normalized.imputed)[i], append_ = ifelse(i == 1, FALSE, TRUE)))
plot_grid(plotlist = lapply(imputation.result, function(i) i[["plot"]]), ncol = length(tissues))
rm(impute.and.plot, imputation.result, merge.unique.proteins.dia, merge.unique.samples.dia, merge.full.joins)
```
# Visualizing the data and the results
## Shared proteins across tissues
The figure shows shared proteins among tissues. The left panel (A) shows shared proteins among all tissues, while the right one (B) among those with more than 1000 identified proteins.
```{r venn.diagram.shared.proteins, echo=FALSE, message=FALSE, warning=FALSE, results='hide', fig.show='asis', fig.keep='last', fig.ext='png', fig.width=14, fig.height=7, dev='png'}
suppressPackageStartupMessages(library(VennDiagram))
draw.venn.diagram.5.tissues <- function(index.azpid_, filename_) {
venn.diagram(x = list(Fat = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[1], sep = ".")))) %>% pull(AZpID),
Liver = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[2], sep = ".")))) %>% pull(AZpID),
Muscle = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[3], sep = ".")))) %>% pull(AZpID),
Pancreas = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[4], sep = ".")))) %>% pull(AZpID),
Serum = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[5], sep = "."))))%>% pull(AZpID)),
category.names = c(paste("VAT", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[1], sep = "."))))), ")", sep = ""),
paste("Liver", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[2], sep = "."))))), ")", sep = ""),
paste("Muscle", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[3], sep = "."))))), ")", sep = ""),
paste("Islets", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[4], sep = "."))))), ")", sep = ""),
paste("Serum", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[5], sep = "."))))), ")", sep = "")),
filename = filename_, imagetype = "tiff", lwd = 1, cex = 0.7, fontfamily = "sans",
col = venn.cols, fill = c(alpha(colour = venn.cols[1], alpha = 0.3), alpha(colour = venn.cols[2], alpha = 0.3), alpha(colour = venn.cols[3], alpha = 0.3),
alpha(colour = venn.cols[4], alpha = 0.3), alpha(colour = venn.cols[5], alpha = 0.3)),
cat.cex = 1.5, cat.default.pos = "outer", cat.pos = c(-65, 163, -100, 130, 20), cat.dist = c(0.27, -0.24, 0.25, 0.22, 0.23), cat.fontfamily = "sans", cat.col = venn.cols)
}
draw.venn.diagram.4.tissues <- function(index.azpid_, filename_) {
venn.diagram(x = list(Fat = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[1], sep = ".")))) %>% pull(AZpID),
Liver = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[2], sep = ".")))) %>% pull(AZpID),
Muscle = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[3], sep = ".")))) %>% pull(AZpID),
Pancreas = filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[4], sep = ".")))) %>% pull(AZpID)),
category.names = c(paste("VAT", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[1], sep = "."))))), ")", sep = ""),
paste("Liver", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[2], sep = "."))))), ")", sep = ""),
paste("Muscle", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[3], sep = "."))))), ")", sep = ""),
paste("Islets", "\n(n=", nrow(filter(index.azpid_, !is.na(!!as.symbol(paste("AZpID", tissues[4], sep = "."))))), ")", sep = "")),
filename = filename_, imagetype = "tiff", lwd = 1, cex = 2, fontfamily = "sans", label.col = 'grey30',
col = venn.cols[1:4], fill = c(alpha(colour = venn.cols[1], alpha = 0.3), alpha(colour = venn.cols[2], alpha = 0.3), alpha(colour = venn.cols[3], alpha = 0.3), alpha(colour = venn.cols[4], alpha = 0.3)),
cat.cex = 1.5, cat.default.pos = "outer", cat.pos = c(-13, 193, -13, 193), cat.dist = c(0.24, -0.24, 0.14, -0.14), cat.fontfamily = "sans", cat.col = venn.cols[1:4])
}
venn.tiff.files <- c(paste(output.directory, "venn.diagram.shared.proteins.5.tissues-1.tiff", sep = ""), paste(output.directory, "venn.diagram.shared.proteins.4.tissues-1.tiff", sep = ""))
venn.cols <- unname(tissue.colors)
draw.venn.diagram.5.tissues(index.azpid_ = dia$index$AZpID, filename_ = venn.tiff.files[1])
draw.venn.diagram.4.tissues(index.azpid_ = dia$index$AZpID, filename_ = venn.tiff.files[2])
plot_grid(ggdraw() + draw_image(venn.tiff.files[1]), ggdraw() + draw_image(venn.tiff.files[2]), labels = c("A", "B"), label_size = 20, ncol = 2, scale = 0.95)
file.remove(paste(output.directory, list.files(path = output.directory, pattern = glob2rx("venn*.log")), sep = ""))
rm(venn.tiff.files, venn.cols, draw.venn.diagram.4.tissues, draw.venn.diagram.5.tissues)
```
## Clustering the proteomics into tissues