-
Notifications
You must be signed in to change notification settings - Fork 0
/
p24073.Rmd
263 lines (197 loc) · 8.87 KB
/
p24073.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
---
title: "p24073 spectral angle using AlphaPept_ms2_generic"
author: "Christian Panse <cp@fgcz.ethz.ch>"
date: "`r date()`"
output:
html_document:
toc_float: true
bibliography: fgcz.bib
---
To further validate all peptide spectrum matches (PSMs) with a mascot (version 2.7) [@mascot] score below 40, we compared each spectrum with an `AlphaPept_ms_generic` [@alphapept] fragment-ion prediction model using R (version 4.4.1), Bioconductor (version 3.19), and the koinar package [@koina; @bioc.koinar]. The following model parameters were used: QE for Thermo Fischer Scientific (TFS) QExactive mass spectrometer and a normalized collision energy (CE) setting of 28eV. The rawrr package [@rawrr] was used to extract spectral data from the TFS generated raw files. For the CE calibration [@prosit], we predicted fragment ion intensities using the model by iterating over a CE range between 15eV and 36eV, and determined the local maxima of the Pearson correlation of matching peaks from experimental and predicted fragment ions (mass window of 0.1Da). PSMs with spectral angle Pearson scores of over 0.70 are considered to pass, while the cut-off score for PSMs with methionine or proline oxidation (Unimod:35) was 0.60. We chose a lower score for PTMs because the current prediction models have limited prediction power with respect to post-translational modifications. Alphabet is the model considered most capable of predicting PTM fragments at the moment. The mirror plots were generated in plain R (source code: github.com/fgcz/koina-paleo/) with red and blue peaks corresponding to the predicted and experimental fragment ions, respectively.
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```
## Required R packages
```{r requiredPackages}
stopifnot(require(readr),
require(httr),
require(jsonlite),
require(rawrr))
```
## Input - read PSM information
```{r readInput}
df <- readr::read_csv("__wilkin_2023_NEW.csv",
col_types = cols(scan = col_integer(),
charge = col_integer(), SpectrumID = col_integer()))
df[1:3,] -> df.psm
```
```{r printInput}
knitr::kable(df, row.names = T)
```
```{r configKoinaRpkg}
koina_AlphaPept_ms2_generic <- koinar::Koina$new(
# model_name = "Prosit_2020_intensity_HCD",
model_name = "AlphaPept_ms2_generic",
# server_url = "dlomix.fgcz.uzh.ch:443",
ssl = TRUE
)
n <- nrow(df.psm)
input <- list(
peptide_sequences = array(df.psm$sequence,
dim = c(n, 1)),
collision_energies = array(rep(35, n), dim = c(n, 1)),
precursor_charges = array(df.psm$charge, dim = c(n, 1)),
instrument_types = array(rep("QE", n), dim = c(n,1))
)
koina_AlphaPept_ms2_generic$predict(input,
pred_as_df = FALSE) -> rv_AlphaPept_ms2_generic
extractKoinaPrediction <- function(p, idx){
df <- data.frame(mz = p$mz[idx, ],
intensities = p$intensities[idx, ],
annotation = p$annotation[idx, ])
df[df$intensities > -1, ]
}
```
## Predict peptide fragment intensities using koina
```{r rawrrReadSpectra, eval = FALSE}
S$rawfile <- paste0(S$file, ".raw", sep = "")
spectra <- lapply(1:n, FUN = function(idx){
rf <- S$rawfile[idx]
scan <- S$scan[idx]
message("Processing ", rf, " scan ", scan, " using rawrr ...")
rawrr::readSpectrum(rf, scan = scan)[[1]]
})
save(spectra, S, file = "p24073_spectra.RData")
```
## Load spectra and input from Rdata file
```{r loadSpectra}
"p24073_spectra.RData" -> rdafn
stopifnot(file.exists(rdafn))
load(rdafn)
```
```{r someProofOfConcept}
spectra |> length()
spectra[1]
stopifnot(length(spectra) == nrow(S))
S |> head()
```
## Define R-Helper functions
Define Determine match between predicted and measured spectra function
```{r defineDetermineMatch, eval=TRUE}
.determineMatch <- function(s0, s1, plot, massRange){
s1 <- s1[s1$mZ>0, ]
idx <- order(s1$mZ)
s1 <- s1[idx, ]
idxNN <- protViz::findNN(q = s1$mZ, s0$mZ)
idxHits <- (abs(s1$mZ - s0$mZ[idxNN]) < 0.05)
s1idx <- which(idxHits)
s0idx <- idxNN[idxHits]
if(plot && length(s0idx) > 0 && length(s1idx) > 0){
## mirror plot
plot(s0$mZ, s0$intensity/max(s0$intensity),
lwd = 1.25,
col = rgb(0.1, 0.1, 0.9, alpha = 0.5),
xlim = massRange,
ylim = c(-1, 1),
type = 'n',
xlab = "m/z",
ylab = "Relative Intensities")
points(s0$mZ[s0idx], -s0$intensity[s0idx] / max(s0$intensity[s0idx]),
type = 'h', col='blue', lwd = 3)
points(s1$mZ[s1idx], s1$intensity[s1idx] / max(s1$intensity[s1idx]),
type = 'h', col='red', lwd = 3)
axis(3, s1$mZ[s1idx], s1$annotation[s1idx], las = 2, cex.axis = 0.5)
}
data.frame(pearson = cor(s0$intensity[s0idx], s1$intensity[s1idx], method = "pearson"),
spearman = cor(s0$intensity[s0idx], s1$intensity[s1idx], method = "spearman"),
nMatches = length(s0idx),
coverage = round(100 * length(s0idx) / length(s1$intensity)))
}
```
```{r defineDetermineCov, fig.retina=3}
.determineCov <- function(CE, plot = FALSE){
n <- length(spectra)
input <- list(
peptide_sequences = array(S$sequence,
dim = c(n, 1)),
collision_energies = array(rep(CE, n), dim = c(n, 1)),
precursor_charges = array(S$charge, dim = c(n, 1)),
instrument_types = array(rep("QE", n), dim = c(n,1))
)
#browser()
koina_AlphaPept_ms2_generic$predict(input, pred_as_df = FALSE) -> rv_AlphaPept_ms2_generic
rv <- lapply(1:n,
FUN = function(idx){
# message("idx", idx, " ...")
s <- spectra[[idx]]
s0 <- data.frame(mZ = s$mZ, intensity = s$intensity / max(s$intensity))
s1 <- data.frame(mZ = rv_AlphaPept_ms2_generic$mz[idx, ],
intensity = rv_AlphaPept_ms2_generic$intensities[idx, ],
annotation = rv_AlphaPept_ms2_generic$annotation[idx, ])
extractKoinaPrediction(rv_AlphaPept_ms2_generic, idx) -> predictedSpectrum
if (plot){
op <- par(mfrow = c(2, 1))
plot(s, sub = paste(S$sequence[idx],
"charge", S$charge[idx]))
points(predictedSpectrum$mz, predictedSpectrum$intensities,
col = rgb(0.9, 0.1, 0.1, alpha = 0.5), type = 'h', lwd = 1.25)
axis(3, s1$mZ, s1$annotation, las = 2, cex.axis = 0.5)
}
pc <- .determineMatch(s0, s1, plot = plot, massRange = s$massRange)
if (plot){
legend("topright", c(paste0("pearson = ", round(pc$pearson,2)),
paste0("spearman = ", round(pc$spearman,2)),
paste0("#matches =", pc$nMatches),
paste0("coverage in %", pc$coverage)),
cex = 0.5)
}
#data.frame(pearson = pc$pearson, sequence = S$sequence[idx])
pc$sequence <- S$sequence[idx]
pc$rawfile <- S$rawfile[idx]
pc$scan <- S$scan[idx]
pc$charge <- S$charge[idx]
pc$scanId <- S$scan[idx]
pc
})
#if (plot) dev.off()
rv |> Reduce(f = rbind) -> df
df$CE <- CE
df
}
```
here we generate the mirror plots
```{r determineBestCEBestOf, fig.retina=3, fig.height=8}
pdf("p24073_spectra.pdf", width = 19/2, height = 12/2)
.determineCov(CE = 28, plot = TRUE) -> rv
write.csv(rv, "p24073_spectra_CE28_QE_AlphaPeptGeneric.csv", row.names = FALSE)
dev.off()
```
## Perform CE normalization
```{r determineCeNormalization}
lapply(15:36, FUN = .determineCov) |>
Reduce(f = rbind) -> df
```
```{r boxplotFun, fig.height = 12, fig.retina=3, fig.cap = "CE normalization: TRUE represents specs having an identified sequence containing P[UNIMOD:35].", fig.width=8}
lattice::bwplot(df$pearson ~ as.character(df$CE) | grepl("P\\[UNIMOD:35", df$sequence),
ylab = "median spectral angle (pearson)",
xlab = "CE [in eV]",
scales=list(y=list(at=seq(0, 1, .05)), x = list(rot = 45)),
layout = c(2, 1),
horizontal = FALSE)
lattice::bwplot(df$spearman ~ as.character(df$CE) | grepl("P\\[UNIMOD:35", df$sequence),
ylab = "median spectral angle (spearmann)",
xlab = "CE [in eV]",
scales=list(y=list(at=seq(0, 1, .05)), x = list(rot = 45)),
layout = c(2, 1),
horizontal = FALSE)
```
```{r hist}
hist(rv$pearson)
hist(rv$spearman)
```
# References
<div id="refs"></div>
# Session info
```{r sessionInfo}
sessionInfo()
```