-
Notifications
You must be signed in to change notification settings - Fork 0
/
full_CP.Rmd
1233 lines (1012 loc) · 42 KB
/
full_CP.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: "Capitainer analysis"
author: "Philippine Louail"
date: "2024-04-24"
output: html_document
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```
```{r packages, message=FALSE, warning=FALSE}
library(MsExperiment)
library(xcms)
library(Spectra)
library(RColorBrewer)
library(pander)
library(readxl)
library(MetaboCoreUtils)
library(pheatmap)
library(MsBackendSql)
library(readxl)
library(Biobase)
library(SummarizedExperiment)
library(openxlsx)
library(vioplot)
```
```{r parallel-process}
#' Set up parallel processing using 2 cores
if (.Platform$OS.type == "unix") {
register(MulticoreParam(2))
} else
register(SnowParam(2))
```
# Introduction
Here I will evaluate the effect of different solvent used for extraction and
their effect on the metabolite profile of the samples.
This markdown is for the Capitainer device, and I will compare the results
between the solvent throughout the analysis.
Below we load the data with its respective phenodata.
```{r load_data}
#' No Phenodata - to be added
MZML_PATH <- getwd()
pd <- read_xls("phenodata.xls") |>
as.data.frame()
full <- readMsExperiment(paste0(MZML_PATH, "/", pd$file), sampleData = pd)
full
sampleData(full)|>
as.data.frame() |>
pandoc.table(style = "simple", caption = "Samples from the data set.")
```
#CP
```{r}
# Extract each sovent of CP
full <- full[sampleData(full)$sample_type == "Capitainer" &
sampleData(full)$sample_role %in% c("Sample", "Blank")]
sampleData(full) |>
as.data.frame() |>
pandoc.table(style = "simple", caption = "Samples from the data set.")
dir <- "CP_results/full/"
dir.create(dir, recursive = TRUE, showWarnings = FALSE)
```
```{r quick-checks}
#' Retention time range for entire dataset
spectra(full) |>
rtime() |>
range()
# Check number of samples
length(full)
#' Check Ms level
spectra(full) |>
msLevel() |>
split(fromFile(full)) |>
lapply(table)
```
```{r}
#' Define colors for the different sample role
leg_sample <- brewer.pal(8, name = "Dark2")[c(2, 8)]
names(leg_sample) <- unique(sampleData(full)$sample_role)
col_sample <- leg_sample[sampleData(full)$sample_role]
#' Define colors for the differen solvent
leg_solvent <- brewer.pal(8, name = "Dark2")[c(3, 5, 6, 7)]
names(leg_solvent) <- c("MetOH_H2O", "ACN", "MetOH", "ACN_MetOH")
col_solvent <- leg_solvent[sampleData(full)$solvent]
```
Below we compute and plot the BPC of the dataset.
```{r plot_bpc}
#' filter the beginning of the retention time
full <- filterRt(full, c(40, 840))
#' First extract and plot bpc
bpc <- chromatogram(full, aggregationFun = "max", msLevel = 1, chunkSize = 2)
plot(bpc, main = "BPC", col = col_sample,
lwd = 1)
grid()
legend("topright", col = leg_sample,
legend = names(leg_sample), lty = 1, horiz = TRUE, bty = "n")
```
BPC comments:
There seem to be contamination throughout the entire RT range but especially
after 500s. This needs to be taken in account in the analysis.
I import a known compound list (curation of annotated ions from MsDial using
HMDB database). I will use these compounds all throughout the analysis to
evaluate the different preprocessing steps.
I then extract eics for each of these compounds and evaluate them regarding:
- peak shape, width
- intensity variation related to sample-type,...
- retention time variation
```{r include=FALSE}
#import known_compound list
compounds <- read_xlsx("Annotation_List.xlsx") |>
as.data.frame()
# plot eic and coloring per solvent - see if I need to extend rtmin and rtmax
# full
eics <- chromatogram(
full,
rt = as.matrix(compounds[, c("rtmin", "rtmax")]),
mz = as.matrix(compounds[, c("mzmin", "mzmax")]), msLevel = 1, chunkSize = 2)
fData(eics)$mz <- compounds$Average_Mz
fData(eics)$rt <- compounds$Average_Rt
fData(eics)$name <- compounds$Metabolite_Name
rownames(eics) <- compounds$Metabolite_Name
tmpdr <- paste0(dir, "raw/")
dir.create(tmpdr, recursive = TRUE, showWarnings = FALSE)
for (i in seq_len(nrow(compounds))) {
png(paste0(tmpdr, "EIC_", fData(eics)$name[i], ".png"),
width = 12, height = 8, units = "cm", res = 600, pointsize = 4)
plot(eics[i, ], main = fData(eics)$name[i],
col = paste0(col_sample, 80))
grid()
legend("topright", col = leg_sample,
legend = names(leg_sample), lty = 1)
abline(v = fData(eics)$rt[i], col = "red", lty = 3)
dev.off()
}
```
There is wide intensity variation between the study samples. Not related to
solvents. This also needs to be taken in account in the rest of the analysis
# preprocessing
The parameters for peak picking are based on previous data that I received and
tested using the list of known ions given by the laboratory.
The parameters of the CentWaveParam method are the followings:
- `peakwidth`: the expected peak width in seconds. Here looking at our internal
standard and other peaks in the dataset we estimate that they are between 10
to 20 second wide
- `ppm`: The accepted m/z deviation in ppm. We set it to 50 ppm.
- some peaks did not have enough datapoints and therefor we use
`extendLengthMSW = TRUE` to extend these signals.
```{r}
#' Peakpicking
param <- CentWaveParam(peakwidth = c(10, 20), ppm = 50, integrate = 2,
snthresh = 5, extendLengthMSW = TRUE)
full <- findChromPeaks(full, param = param, chunkSize = 2L)
```
Again i will plot the EICs to observe how well the peak picking went.
```{r eic2, include=FALSE}
#full
eics <- chromatogram(
full,
rt = as.matrix(compounds[, c("rtmin", "rtmax")]),
mz = as.matrix(compounds[, c("mzmin", "mzmax")]),
msLevel = 1,
chunkSize = 2)
fData(eics)$mz <- compounds$Average_Mz
fData(eics)$rt <- compounds$Average_Rt
fData(eics)$name <- compounds$Metabolite_Name
rownames(eics) <- compounds$Metabolite_Name
tmpdr <- paste0(dir, "chrompeaks/")
dir.create(tmpdr, recursive = TRUE, showWarnings = FALSE)
for (i in seq_len(nrow(compounds))) {
png(paste0(tmpdr, "EIC_", fData(eics)$name[i], ".png"),
width = 12, height = 8, units = "cm", res = 600, pointsize = 4)
plot(eics[i, ], main = fData(eics)$name[i],
col = paste0(col_sample, 80),
peakBg = paste0(col_sample, 40)[chromPeaks(eics[i])[, "column"]])
grid()
legend("topright", col = leg_sample,
legend = names(leg_sample), lty = 1)
abline(v = fData(eics)$rt[i], col = "red", lty = 3)
dev.off()
}
```
Refinement: Merge neighboring peaks. Remove artifacts that can be created
during peak picking. Especially necessary when not enough MS1 data point (which
is the cases here for some peaks)
- `expandRt` and `expandMz` are the maximum allowed distance between two peaks
in the retention time and m/z dimensions, respectively. We set them up based
on the peakwidth set up from the peak picking step and observing how well the
peak picking step performed.
- The other parameters are left as default and their definitions can be found
in `?MergeNeighboringPeaksParam` documentation
```{r}
param <- MergeNeighboringPeaksParam(expandRt = 10,
expandMz = 0.01,
ppm = 10,
minProp = 0.75)
full <- refineChromPeaks(full, param = param, chunkSize = 2)
chromPeakData(full)$merged |>
table()
```
```{r include=FALSE}
eics <- chromatogram(
full,
rt = as.matrix(compounds[, c("rtmin", "rtmax")]),
mz = as.matrix(compounds[, c("mzmin", "mzmax")]),
msLevel = 1,
chunkSize = 2)
fData(eics)$mz <- compounds$Average_Mz
fData(eics)$rt <- compounds$Average_Rt
fData(eics)$name <- compounds$Metabolite_Name
rownames(eics) <- compounds$Metabolite_Name
tmpdr <- paste0(dir, "refine/")
dir.create(tmpdr, recursive = TRUE, showWarnings = FALSE)
for (i in seq_len(nrow(compounds))) {
png(paste0(tmpdr, "EIC_", fData(eics)$name[i], ".png"),
width = 12, height = 8, units = "cm", res = 600, pointsize = 4)
plot(eics[i, ], main = fData(eics)$name[i],
col = paste0(col_sample, 80),
peakBg = paste0(col_sample, 40)[chromPeaks(eics[i])[, "column"]])
grid()
legend("topright", col = leg_sample,
legend = names(leg_sample), lty = 1)
abline(v = fData(eics)$rt[i], col = "red", lty = 3)
dev.off()
}
```
Alignment: Very little variation in RT between samples, but always good to run
it.
We first need to run a correspondence analysis using all samples. This will
allow us to select the best anchor peaks for the alignment. For this we set
`minFraction = 1` meaning that we want to keep peaks that are present in
all samples to base the alignment on these.
```{r}
#' perform quick correspondence analysis - do not take in account Blanks
f <- factor(sampleData(full)$solvent,
levels = unique(sampleData(full)$solvent))
idx_B <- sampleData(full)$sample_role == "Blank"
f[idx_B] <- NA
param <- PeakDensityParam(sampleGroups = f,
minFraction = 1,
binSize = 0.01, ppm = 10,
bw = 2)
full <- groupChromPeaks(full, param = param)
#' align the data
param <- PeakGroupsParam(minFraction = 0.75, extraPeaks = 50, span = 0.5)
#' Input in the function
full <- adjustRtime(full, param = param)
#' See retention time variation
plotAdjustedRtime(full, col = paste0(col_sample, 80), peakGroupsPch = 1)
grid()
legend("topright", col = leg_sample,
legend = names(leg_sample), lty = 1, bty = "n")
full <- applyAdjustedRtime(full)
```
We evaluate the efficacy of the alignment on our internal standard.
```{r include=FALSE}
eics <- chromatogram(
full,
rt = as.matrix(compounds[, c("rtmin", "rtmax")]),
mz = as.matrix(compounds[, c("mzmin", "mzmax")]),
msLevel = 1,
chunkSize = 2)
fData(eics)$mz <- compounds$Average_Mz
fData(eics)$rt <- compounds$Average_Rt
fData(eics)$name <- compounds$Metabolite_Name
rownames(eics) <- compounds$Metabolite_Name
tmpdr <- paste0(dir, "aligned/")
dir.create(tmpdr, recursive = TRUE, showWarnings = FALSE)
for (i in seq_len(nrow(compounds))) {
png(paste0(tmpdr, "EIC_", fData(eics)$name[i], ".png"),
width = 12, height = 8, units = "cm", res = 600, pointsize = 4)
plot(eics[i, ], main = fData(eics)$name[i],
col = paste0(col_sample, 80),
peakBg = paste0(col_sample, 40)[chromPeaks(eics[i])[, "column"]])
grid()
legend("topright", col = leg_sample,
legend = names(leg_sample), lty = 1)
abline(v = fData(eics)$rt[i], col = "red", lty = 3)
dev.off()
}
```
Correspondence step:
Correspondence is the step were features are defined based on how often a peak
is repeated in the dataset. We set up the following parameters:
- `sampleGroups`: the factor to group the samples. Here we remove blanks from
being considered in this step. And separate the dataset per triplicates of
solvents.
- `minFraction`: the minimum fraction of samples of a certain device in which a
peak must be present to be considered a feature. We set it to 2/3. which mean
in this case that a peak must be present in at least 2 out of 3 samples of
a device to be considered a feature.
- `bw`: the bandwidth of the kernel density estimation. We set it to 2 after
testing a checking that is it wide enough to capture the peaks for one
feature.
```{r}
# correspondence - use same factor as before
param <- PeakDensityParam(sampleGroups = f,
minFraction = 2/3, binSize = 0.015, bw = 2.0, ppm = 10)
plotChromPeakDensity(
eics["1_18_1_lysophosphatidylcholine"], param = param,
col = paste0(col_sample, "80"),
peakCol = col_sample[
chromPeaks(eics["1_18_1_lysophosphatidylcholine"])[, "column"]],
peakBg = paste0(col_sample[
chromPeaks(eics["1_18_1_lysophosphatidylcholine"])[, "column"]], 20),
peakPch = 16)
plotChromPeakDensity(eics["LysoPhosphatidylcholine_16_0"], param = param,
col = paste0(col_sample, "80"),
peakCol = col_sample[
chromPeaks(eics["LysoPhosphatidylcholine_16_0"])[, "column"]],
peakBg = paste0(
col_sample[chromPeaks(
eics["LysoPhosphatidylcholine_16_0"])[, "column"]], 20),
peakPch = 16)
# the tests look great, lets apply to entire dataset
full <- groupChromPeaks(full, param = param)
# how many features:
nrow(featureDefinitions(full))
```
We will now use gapfilling to fill in missing values in the dataset. We will
use the `ChromPeakAreaParam` method with defualt parameters for this.
```{r}
#' gap filing
#' Number of missing values
sum(is.na(featureValues(full)))
full <- fillChromPeaks(full, param = ChromPeakAreaParam(), chunkSize = 2)
#' How many missing values after
sum(is.na(featureValues(full)))
```
Extract the intensity values for the features and save the data.
```{r}
#' Extract results as a SummarizedExperiment
library(SummarizedExperiment)
res_full <- quantify(full, method = "sum", filled = FALSE)
assays(res_full)$raw_filled <- featureValues(full, method = "sum",
filled = TRUE )
```
# flag features in blanks
We will now flag features that are highly present in blanks. We will use the
`filterFeatures` with the `BlankFlag` method to do this. By setting up a
`threshold = 2` we will flag the features that have an intensity in the blanks
that is at least half the intensity in the study samples.
We finalize preprocessing by generating a summarizedExperiment object. We also
remove the previously flagged features from the data for subsequent analysis.
```{r}
idx <- sampleData(full)$sample_role == "Blank"
full <- filterFeatures(full, BlankFlag(blankIndex = idx, qcIndex = !idx))
featureDefinitions(full)$possible_contaminants[is.na(featureDefinitions(full)$possible_contaminants)] <- FALSE
featureDefinitions(full)$possible_contaminants <- as.logical(featureDefinitions(full)$possible_contaminants)
# we actually remove them for downstream analysis in the sumexp object
nrow(res_full)
res_full <- res_full[!featureDefinitions(full)$possible_contaminants, ]
nrow(res_full) # check that it is working
de <- list(values = as.data.frame(featureValues(full)), definitions = featureDefinitions(full))
write.xlsx(de, paste0(dir, "feature_results.xlsx"), rowNames = TRUE)
# just remove blanks also
full
full <- full[sampleData(full)$sample_role != "Blank", keepFeatures = TRUE]
full
res_full
res_full <- res_full[, res_full$sample_role != "Blank"]
res_full
```
Save files at the end of preprocessing
```{r}
save(res_full, file = paste0(dir, "SumExp_full_preprocessing.RData"))
load(file = paste0(dir, "SumExp_full_preprocessing.RData"))
save(full, file = paste0(dir, "full_preprocessing.RData"))
load(file = paste0(dir, "full_preprocessing.RData"))
```
# Noise comparison: we will compare the amount of noise between the solvents.
Below we compare the noise signals between devices. We first calculate the
overall signal in the dataset and then calculate the signal that is in the
chromatographic peaks detection. We then subtract the two to get the noise
signal.
```{r}
# overall signal in the dataset
#' - for each file calculate the sum of intensities
background <- spectra(full) |>
split(fromFile(full)) |>
lapply(tic) |>
lapply(sum) |>
unlist()
# Overall signal that is in the chromatographic peaks detection
# - check "into" definition first, mioght need to multiply it by something
detected <- apply(assay(res_full), 2, function(x) sum(x, na.rm = TRUE))
# substract and plot ? Also i'm removing blanks bc i think we don't need it
names(background) <- names(detected) <- res_full$solvent
#remove blanks
noise <- background - detected
f <- factor(names(noise), levels = unique(names(noise)))
group <- split(noise, f)
plot(NULL, xlim = c(1, length(group)), ylim = range(unlist(group)),
xaxt = "n", xlab = "Solvents", ylab = "Noise",
main = "Noise comparison between solvents")
for (i in seq_along(group)) {
points(rep(i, length(group[[i]])), group[[i]], pch = 19, col = leg_solvent[i])
}
axis(1, at = seq_along(group), labels = names(group))
```
# Normalisation
```{r extra-packages, message=FALSE, warning=FALSE}
library(ggfortify)
library(SummarizedExperiment)
library(RColorBrewer)
col_solvent <- leg_solvent[res_full$solvent]
```
We first need to evaluate the data distribution and try to see any technical
related bias.
```{r counts1, fig.height=5, fig.width=5, include=TRUE}
layout(mat = matrix(1:3, ncol = 1), height = c(0.2, 0.2, 0.8))
par(mar = c(0.2, 4.5, 0.2, 3))
barplot(apply(assay(res_full, "raw"), MARGIN = 2,
function(x) sum(!is.na(x))),
col = col_solvent, ylab = "features raw data", xaxt = "n",
space = 0.012)
barplot(apply(assay(res_full, "raw_filled"), MARGIN = 2,
function(x) sum(!is.na(x))),
col = col_solvent, ylab = "features filled data", xaxt = "n",
space = 0.012)
boxplot(log2(assay(res_full, "raw_filled")), xaxt = "n",
ylab = expression(log[2]~abundance~filled~data),
col = col_solvent, outline=FALSE, medlty = "blank",
border = col_solvent, varwidth = TRUE)
points(colMedians(log2(assay(res_full, "raw_filled")),
na.rm = TRUE), type = "b", pch = 16)
grid(nx = NA, ny = NULL)
legend("topright", col = leg_solvent,
legend = names(leg_solvent), lty=1, lwd = 2, xpd = TRUE, ncol = 4,
cex = 0.8, bty = "n")
```
```{r rla-plot raw and filled1, fig.cap = "RLA plot for the raw data and filled data. Note: outliers are not drawn."}
par(mfrow = c(1, 1), mar = c(0.2, 4.5, 2.5, 3))
boxplot(rowRla(assay(res_full, "raw_filled"),
group = res_full$solvent),
cex = 0.5, pch = 16, boxwex = 1,
col = col_solvent, ylab = "RLA",
border = paste0(col_solvent, 40),
outline = FALSE, xaxt = "n", main = "Relative log abundance",
cex.main = 1)
grid(nx = NA, ny = NULL)
abline(h = 0, lty=3, lwd = 1, col = "black")
legend("topright", col = leg_solvent,
legend = names(leg_solvent), lty=1, lwd = 2, xpd = TRUE, ncol = 3,
cex = 0.8, bty = "n")
```
We will run a median scaling step to see how well we can correct the
variation between samples.
## quick median scaling
```{r}
#' Compute median and generate normalization factor, we compute per solvent as
#' to keep the technical variation relatedto the solvent.
assays(res_full)$norm <- assay(res_full, "raw_filled")
for (i in res_full$solvent) {
idx <- res_full$solvent == i
mdns <- apply(assay(res_full, "raw_filled")[, idx], 2, median, na.rm = TRUE)
nf_mdn <- mdns / median(mdns)
assays(res_full)$norm[, idx] <- sweep(assay(res_full, "raw_filled")[, idx], MARGIN = 2,
nf_mdn, '/')
}
```
We also do not use the median from blank samples as we don't want them to
influence our study samples. Because we separated the blanks when defining the
features (correspondence step) we will have much *less* features in blanks and
therefore the intensity distribution for these is not reliable.
```{r rla-plot after norm2, include = TRUE, fig.cap = "RLA plot before and after normalization. Note: outliers are not drawn.", fig.height= 7, fig.width=5.5}
par(mfrow = c(2, 1), mar = c(1, 4, 3, 1))
boxplot(rowRla(assay(res_full, "raw_filled"),
group = res_full$solvent),
cex = 0.5, pch = 16, boxwex = 1,
col = col_solvent, ylab = "RLA",
border = paste0(col_solvent, 40),
outline = FALSE, xaxt = "n",
main = "Relative log abundance before normalisation",
cex.main = 1)
grid(nx = NA, ny = NULL)
abline(h = 0, lty=3, lwd = 1, col = "black")
legend("topright", col = leg_solvent,
legend = names(leg_solvent), lty=1, lwd = 2, xpd = TRUE, ncol = 3,
cex = 0.8, bty = "n")
boxplot(rowRla(assay(res_full, "norm"),
group = res_full$solvent),
cex = 0.5, pch = 16, boxwex = 1,
col = col_solvent, ylab = "RLA",
border = paste0(col_solvent, 40),
outline = FALSE, xaxt = "n",
main = "Relative log abundance after normalization",
cex.main = 1, ylim = c(-1.0 , 2.0))
grid(nx = NA, ny = NULL)
abline(h = 0, lty=3, lwd = 1, col = "black")
legend("topright", col = leg_solvent,
legend = names(leg_solvent), lty=1, lwd = 2, xpd = TRUE, ncol = 3,
cex = 0.8, bty = "n")
```
## Coeffcient of variation
The coefficient of variation (or Relative standanrd deviation) evaluate how
close data is to each other. It is especially interesting in our case as we
have triplicate.
Therefore here we compute the RSD for each feature across the samples in each
solvent.
The RSD table below therefore give us an information on how the triplicate are
close to each other overall per solvent.
```{r include=TRUE, results = "asis"}
# indices of each solvents' triplicates
idx_mh <- res_full$solvent == "MetOH_H2O"
idx_m <- res_full$solvent == "MetOH"
idx_a <- res_full$solvent == "ACN"
idx_am <- res_full$solvent == "ACN_MetOH"
# Compute Rsds on all data per solvents
sample_res <- cbind(
Raw_MH = rowRsd(assay(res_full, "raw_filled")[, idx_mh],
na.rm = TRUE, mad = TRUE),
Norm_MH = rowRsd(assay(res_full, "norm")[, idx_mh],
na.rm = TRUE, mad = TRUE),
Raw_M = rowRsd(assay(res_full, "raw_filled")[, idx_m],
na.rm = TRUE, mad = TRUE),
Norm_M = rowRsd(assay(res_full, "norm")[, idx_m],
na.rm = TRUE, mad = TRUE),
Raw_A = rowRsd(assay(res_full, "raw_filled")[, idx_a],
na.rm = TRUE, mad = TRUE),
Norm_A = rowRsd(assay(res_full, "norm")[, idx_a],
na.rm = TRUE, mad = TRUE),
Raw_AM = rowRsd(assay(res_full, "raw_filled")[, idx_am],
na.rm = TRUE, mad = TRUE),
Norm_AM = rowRsd(assay(res_full, "norm")[, idx_am],
na.rm = TRUE, mad = TRUE)
)
#' Compute quantile for better data visualisation
res_df <- data.frame(
Raw_MH = quantile(sample_res[, "Raw_MH"], na.rm = TRUE),
Norm_MH = quantile(sample_res[, "Norm_MH"], na.rm = TRUE),
Raw_M = quantile(sample_res[, "Raw_M"], na.rm = TRUE),
Norm_M = quantile(sample_res[, "Norm_M"], na.rm = TRUE),
Raw_A = quantile(sample_res[, "Raw_A"], na.rm = TRUE),
Norm_A = quantile(sample_res[, "Norm_A"], na.rm = TRUE),
Raw_AM = quantile(sample_res[, "Raw_AM"], na.rm = TRUE),
Norm_AM = quantile(sample_res[, "Norm_AM"], na.rm = TRUE)
)
cpt <- paste0("Distribution of RSD values across samples for the raw and ",
"normalized data.")
pandoc.table(res_df, style = "rmarkdown", caption = cpt)
```
Both the RLA plot and the CV values show that the normalization step was
successful. The RLA shows all samples median got closer to each other and 75%
of our features have a CV below 30% for each solvent which is great.
# Comparison on overall data
- summary plot: the plot below is one of the summary plot that compare results
after prepossessing and normalization.
```{r fig.height=8, fig.width=6}
#' Quantile of RSD values after norm - non contaminant
res_df <- data.frame(
MetOH_H2O = quantile(sample_res[, "Norm_MH"], na.rm = TRUE),
ACN = quantile(sample_res[, "Norm_A"], na.rm = TRUE),
MetOH = quantile(sample_res[, "Norm_M"], na.rm = TRUE),
ACN_MetOH = quantile(sample_res[, "Norm_AM"], na.rm = TRUE)
)
# Intensity and missing values
res_mh <- res_full[, res_full$solvent == "MetOH_H2O"]
res_m <- res_full[, res_full$solvent == "MetOH"]
res_a <- res_full[, res_full$solvent == "ACN"]
res_am <- res_full[, res_full$solvent == "ACN_MetOH"]
idx_fts <- cbind(
MetOH_H2O = rowSums(is.na(assay(res_mh, "norm"))) < 2,
ACN = rowSums(is.na(assay(res_a, "norm"))) < 2,
MetOH = rowSums(is.na(assay(res_m, "norm"))) < 2,
ACN_MetOH = rowSums(is.na(assay(res_am, "norm"))) < 2)
res_mh <- res_mh[rowSums(is.na(assay(res_mh, "norm"))) < 2,]
res_m <- res_m[rowSums(is.na(assay(res_m, "norm"))) < 2,]
res_a <- res_a[rowSums(is.na(assay(res_a, "norm"))) < 2,]
res_am <- res_am[rowSums(is.na(assay(res_am, "norm"))) < 2,]
intensity <- cbind(
MetOH_H2O = log2(as.numeric(assay(res_mh, "norm"))),
ACN = log2(as.numeric(assay(res_a, "norm"))),
MetOH = log2(as.numeric(assay(res_m, "norm"))),
AVN_MetOH = log2(as.numeric(assay(res_am, "norm"))))
num_features <- cbind(
MetOH_H2O = nrow(res_mh),
ACN = nrow(res_a),
MetOH = nrow(res_m),
ACN_MetOH = nrow(res_am)
)
missing_values <- cbind(
MetOH_H2O = sum(is.na(assay(res_mh, "raw")))/length(assay(res_mh, "raw")) * 100,
ACN = sum(is.na(assay(res_a, "raw")))/length(assay(res_a, "raw")) * 100,
MetOH = sum(is.na(assay(res_m, "raw")))/length(assay(res_m, "raw")) * 100,
ACN_MetOH = sum(is.na(assay(res_am, "raw")))/length(assay(res_am, "raw")) * 100
)
#General plot
#General plot - now without flagged features
layout(mat = matrix(1:3, ncol = 1), height = c(0.3, 0.3, 0.8))
par(mar = c(1, 4.5, 1, 3))
barplot(colSums(num_features),
col = leg_solvent,
ylab = "Number of features", space = 0.05, ylim = c(0, 6000))
barplot(c(missing_values),
ylab = "% of missing values", col = leg_solvent, space = 0.05, ylim = c(0, 60))
vioplot(intensity,
ylab = "Log2 intensity", col = leg_solvent, space = 0.05)
```
```{r}
cpt <- paste0("RSD values distributionacross samples for the ",
"normalized data for each solvent type.")
pandoc.table(res_df, style = "rmarkdown", caption = cpt)
```
- number of features / intensity per rt slices
```{r fig.height=9, fig.width=8}
# Bin features per RT slices
vc <- rowData(res_full)$rtmed
breaks <- seq(0, max(vc, na.rm = TRUE) + 1, length.out = 15) |>
round(0)
cuts <- cut(vc, breaks = breaks, include.lowest = TRUE)
table(cuts)
num_features_solvent <- apply(idx_fts, MARGIN = 2, function(x) table(cuts[x]))
idx_fts <- as.data.frame(idx_fts)
ftc <-function(res_solvent, fts_idx) {
tmp <- rowSums(assay(res_solvent, "norm"), na.rm = TRUE)
cuts_tmp <- cuts[fts_idx]
t <- split(tmp, cuts_tmp) |>
lapply(sum, na.rm = TRUE, simplify = TRUE)
unlist(t)
}
intensity_solvent <- list(
MetOH_H20 = ftc(res_mh, idx_fts$MetOH_H2O),
ACN = ftc(res_a, idx_fts$ACN),
MetOH = ftc(res_m, idx_fts$MetOH),
ACN_MetOH = ftc(res_am, idx_fts$ACN_MetOH)
)
# Transform intensity to log2 scale
intensity_solvent <- lapply(intensity_solvent, log2)
# Plot
layout(mat = matrix(1:2, ncol = 1), heights = c(0.5, 0.5))
par(mar = c(0.5, 4.5, 2, 3))
# Plot number of features
ylim_features <- c(0, max(unlist(num_features_solvent)))
plot(num_features_solvent[,1], col = leg_solvent[1], ylab = "Number of features",
xlab = "", type = "b", pch = 16, xaxt = "n", ylim = ylim_features,
main = "Analysis along the RT axis for each solvent")
for (i in 2:ncol(num_features_solvent)) {
lines(num_features_solvent[,i], col = leg_solvent[i], type = "b", pch = 16)
}
axis(1, at = 1:length(num_features_solvent[,1]), labels = FALSE)
grid()
legend("top", legend = names(intensity_solvent), col = leg_solvent, pch = 16,
cex = 1, horiz = TRUE, bty = "n")
# Plot intensity
par(mar = c(5, 4.5, 2, 3))
ylim_intensity <- range(unlist(intensity_solvent))
plot(intensity_solvent[[1]], type = "b", pch = 16, xlab = "",
ylab = "Log2 intensity", col = leg_solvent[1], xaxt = "n", ylim = ylim_intensity)
for (i in 2:length(intensity_solvent)) {
lines(intensity_solvent[[i]], type = "b", pch = 16, col = leg_solvent[i])
}
axis(1, at = 1:length(intensity_solvent[[1]]),
labels = names(intensity_solvent[[1]]), las = 2, cex.axis = 0.8)
grid()
mtext("Retention time (s)", side = 1, line = 4, cex = 1)
```
- meadians of the medians
- median of median
```{r fig.height=9, fig.width=8}
ftc <-function(res_solvent, fts_idx) {
tmp <- rowMedians(assay(res_solvent, "norm"), na.rm = TRUE)
cuts_tmp <- cuts[fts_idx]
t <- split(tmp, cuts_tmp) |>
lapply(median, na.rm = TRUE, simplify = TRUE)
unlist(t)
}
intensity_solvent <- list(
MetOH_H20 = ftc(res_mh, idx_fts$MetOH_H2O),
ACN = ftc(res_a, idx_fts$ACN),
MetOH = ftc(res_m, idx_fts$MetOH),
ACN_MetOH = ftc(res_am, idx_fts$ACN_MetOH)
)
# Transform intensity to log2 scale
intensity_solvent <- lapply(intensity_solvent, log2)
# Plot
layout(mat = matrix(1:2, ncol = 1), heights = c(0.5, 0.5))
par(mar = c(0.5, 4.5, 2, 3))
# Plot number of features
ylim_features <- c(0, max(unlist(num_features_solvent)))
plot(num_features_solvent[,1], col = leg_solvent[1], ylab = "Number of features",
xlab = "", type = "b", pch = 16, xaxt = "n", ylim = ylim_features,
main = "Analysis along the RT axis for each solvent")
for (i in 2:ncol(num_features_solvent)) {
lines(num_features_solvent[,i], col = leg_solvent[i], type = "b", pch = 16)
}
axis(1, at = 1:length(num_features_solvent[,1]), labels = FALSE)
grid()
legend("top", legend = names(intensity_solvent), col = leg_solvent, pch = 16,
cex = 1, horiz = TRUE, bty = "n")
# Plot intensity
par(mar = c(5, 4.5, 2, 3))
ylim_intensity <- range(unlist(intensity_solvent))
plot(intensity_solvent[[1]], type = "b", pch = 16, xlab = "",
ylab = "Log2 intensity", col = leg_solvent[1], xaxt = "n", ylim = ylim_intensity)
for (i in 2:length(intensity_solvent)) {
lines(intensity_solvent[[i]], type = "b", pch = 16, col = leg_solvent[i])
}
axis(1, at = 1:length(intensity_solvent[[1]]),
labels = names(intensity_solvent[[1]]), las = 2, cex.axis = 0.8)
grid()
mtext("Retention time (s)", side = 1, line = 4, cex = 1)
```
- overlap of features between solvents
```{r echo=TRUE}
# Create a data frame for the UpSet plot - need to fix
upset_df <- lapply(idx_fts, as.integer)
names(leg_solvent) <- names(upset_df)
# Plot the UpSet plot
library(UpSetR)
upset(as.data.frame(upset_df), sets = c("MetOH_H2O", "ACN", "MetOH", "ACN_MetOH"),
sets.bar.color = leg_solvent, mainbar.y.label = "Number of common features", keep.order = TRUE, mainbar.y.max= 3000)
```
```{r}
save(res_full, file = paste0(dir, "SumExp_full_normalisation.RData"))
```
# MS/MS annotation
First need to prepare the different spectra input
```{r prep-spectra}
## dataset
# get spectra data and change their backend
idx_fts <- rownames(featureDefinitions(full)) %in% rownames(res_full)
full_spectra <- featureSpectra(full, msLevel = 2L, features = idx_fts) # not annotating for contamination
#' Remove peaks with an intensity below 5% or the spectra's BPC
low_int <- function(x, ...) {
x > max(x, na.rm = TRUE) * 0.05
}
full_spectra <- filterIntensity(full_spectra, intensity = low_int)
length(full_spectra)
full_spectra$feature_id |>
table() |>
quantile()
full_spectra |>
lengths() |>
quantile()
#' Remove peaks with an m/z > the precursor m/z. For single-charged ions
#' no fragment peak can have an m/z >= the precursor
full_spectra <- full_spectra |>
filterPrecursorPeaks(mz = "==", ppm = 50)
#' Remove spectra with a single peak.
full_spectra <- full_spectra[lengths(full_spectra) > 1]
#' Add Spectra index
full_spectra$spectra_idx <- seq_len(length(full_spectra))
full_spectra <-setBackend(full_spectra, MsBackendMemory())
full_spectra <- applyProcessing(full_spectra)
save(full_spectra, file = paste0(dir, "full_spectra.RData"))
```
```{r loadlibrary}
library(MetaboAnnotation)
library(MsBackendSql)
library(RSQLite)
```
For this analysis we use the GNPS database. We first load the database from a local sqlite file.
```{r loadgnps}
#load gnps library
mb <- Spectra(file.path("/home/plouail/pilot_study/MsBackendSql.GNPS.matchms.cleaned.v1.sqlite"),
drv = SQLite(), source = MsBackendOfflineSql())
```
Below we prepare the database spectra the same way as we prepared our own
spectra.
```{r filtergnps, echo=TRUE}
mb <- setBackend(mb, MsBackendMemory())
low_int <- function(x, ...) {
x > max(x, na.rm = TRUE) * 0.05
}
#' remove negative polarity
mb <- mb[mb$polarity == 1]
#' Do same filtering as for our spectra data
mb <- filterIntensity(mb, intensity = low_int)
mb <- filterPrecursorPeaks(mb, mz = "==", ppm = 50)
mb <- mb[lengths(mb) > 1]
```
We then compare our dataset spectra with the database and select matches with
the score above 0.8.
```{r matching, echo=TRUE}
#' remove parallel processing
register(SerialParam())
#' Matching
prm <- CompareSpectraParam(ppm = 10, requirePrecursor = TRUE,
THRESHFUN = function(x) which(x >= 0.8))
mtch_full <- matchSpectra(full_spectra, mb, param = prm)
mtch_full
#' really low percentage of MS2 spectra matched.
length(whichQuery(mtch_full)) / length(mtch_full) * 100
#' for how many features do we have MS2 spectra
length(unique(mtch_full$feature_id))
#' Keep only the query that got matches
mtch_full <- mtch_full[whichQuery(mtch_full)]
#' for how many features do we have MS2 spectra WITH db matches?
length(unique(mtch_full$feature_id))
# Extract results
md_full <- matchedData(mtch_full, c("rtime", "precursorMz", "feature_id",
"target_inchikey", "target_compound_name",
"score", "spectra_idx"))
md_full
save(md_full, file = paste0(dir, "md_full.RData"))
```
Below we remove duplicate matches (using inchikey) and keep the keep best
scoring match for each inchikey.
```{r echo=TRUE}
rmv_duplicate <- function(md) {
res <- lapply(split(md, md$feature_id), function(x) {
lapply(split(x, x$target_inchikey), function(z) {
z[which.max(z$score), ]
}) |>
do.call(what = rbind)
}) |>
do.call(what = rbind) |>
as.data.frame()
}
md_full <- rmv_duplicate(md_full)
md_full
toberefined <- cbind(md_full, assay(res_full)[md_full$feature_id,])
write.csv(toberefined, "toberefined_CP_solvent.csv")
```
I need them to do some refinement on the annotation before doing proper plotting.
The plots below are just to prepare the codes, they are not the actual results !!
# plot resulting compounds
```{r eval=FALSE, include=FALSE}