match-standards-serum-mix19.Rmd

---
title: "Identifying mix19 standards in serum"
author: "Marilyn De Graeve, Andrea Vicini, Vinicius Verri Hernandes, Johannes Rainer"
output:
  rmarkdown::html_document:
    highlight: pygments
    toc: true
    toc_float: true
    toc_depth: 3
    fig_width: 5
---

```{r, include = FALSE, cached = FALSE}
knitr::read_chunk("R/match-standards-chunks.R")
MIX <- 19
MATRIX <- "Serum"
POLARITY <- "POS"
## settings to find MS2 spectra for features. Be rather inclusive here.
FEATURE_MS2_PPM <- 20
FEATURE_MS2_TOLERANCE <- 0.05
```

```{r, libraries, echo = FALSE, message = FALSE}
```
```{r, general-settings, echo = FALSE, message = FALSE}
```

**Current version: 0.10.1, 2024-10-29.**

# Introduction

Mixes of standards have been solved in serum or added to human serum sample
pools in two different concentration and these samples were measured with the
LC-MS setup from Eurac (used also to generate the CHRIS untargeted metabolomics
data). Assignment of retention time and observed ion can be performed for each
standard based on MS1 information such as expected m/z but also based on the
expected difference in signal between samples with low and high concentration of
the standards. Finally, experimental fragment spectra provide the third level of
evidence. Thus, the present data set allows to annotate features to standards
based on 3 levels of evidence: expected m/z, difference in measured intensity
and MS2-based annotation.

A detailed description of the approach is given in
[match-standards-introduction.Rmd](match-standards-introduction.Rmd).


The list of standards that constitute the present sample mix are listed below
along with the expected retention time and the most abundant adduct for positive
and negative polarity as defined manually in a previous analysis by Mar
Garcia-Aloy.

```{r, standards-table, echo = FALSE, results = "asis"}
```

# Data import

We next load the data and split it according to matrix and polarity.

```{r, data-import}
```

We next also load the reference databases we will use to compare the
experimental MS2 spectra against. Below we thus load HMDB and MassBank data. We
create in addition *neutral loss* versions of the databases. Since HMDB does not
provide precursor m/z values we assume the fragment spectra to represent `[M+H]+`
ions (respectively `[M-H]-` ions for negative polarity) and add the m/z for
these adducts as *precursor m/z*.

```{r, load-reference-databases, message = FALSE}
```

```{r, setup-ion-db, message = FALSE, echo = FALSE}
```

# Initial evaluation of available MS2 data

Before performing the actual analysis that involves chromatographic peak
detection and evaluation of feature abundances, we compare all experimental MS2
spectra against the reference libraries. This provides already a first hint for
which standards a valid MS2 spectrum was recorded (and hence would allow
annotation based on evidence 3) and what related retention time might be.

We first extract all MS2 spectra from the respective data files and process them
by first removing all peaks with an intensity below 5% of the highest peak
signal, then scaling all intensities to a value range between 0 and 100 and
finally remove all MS2 spectra with less than 2 peaks.

```{r prepare-all-ms2}
fls <- fileNames(data_all)[data_all$polarity == "POS"]
```
```{r, all-ms2}
```

We next match these spectra against all reference spectra from HMDB for the
standards in `r MIX_NAME`. 

The settings for the matching are defined below. These will be used for all MS2
spectra similarity calculations. All matching spectra with a similarity larger
0.7 are identified.

```{r, compare-spectra-param}
```
 
Positive polarity:

Standards for which no MS2 spectrum in HMDB is present are listed in
the table below.

```{r, echo = FALSE, results = "asis"}
tmp <- std_dilution[!(std_dilution$HMDB %in% hmdb_std$compound_id), ]
pandoc.table(tmp[, c("name", "HMDB", "formula")], style = "rmarkdown",
             caption = "Standards for which no reference spectrum is available")
```

The table below lists all standards and the number of matching reference spectra
(along with their retention times).

```{r, table-all-ms2-hmdb, echo = FALSE, results = "asis"}
```

Negative polarity:

We repeat the same analysis for the negative polarity data (code not shown
again).

```{r, echo = FALSE}
fls <- fileNames(data_all)[data_all$polarity == "NEG"]
```
```{r, all-ms2, echo = FALSE}
```
```{r, table-all-ms2-hmdb, echo = FALSE, results = "asis"}
```

We next perform the full analysis of the data set involving MS1 and MS2 data.


# Serum, positive polarity

We first perform the analysis on the samples with the standards solved in pure
serum acquired in positive polarity mode.

```{r}
POLARITY <- "POS"
data <- filterFile(data_all, which(data_all$polarity == POLARITY))
hmdb <- hmdb_pos
hmdb_nl <- hmdb_pos_nl
```

## Data pre-processing

We perform the pre-processing of our data set which consists of the
chromatographic peak detection followed by a peak refinement step to reduce peak
detection artifacts, the correspondence analysis to group peaks across samples
and finally the gap-filling to fill in missing peak data from samples in which
no chromatographic peak was detected. For details on settings please refer to
the analysis of *mix 01* samples.

```{r, preprocessing, eval = !file.exists(paste0(RDATA_PATH, "processed_data_", POLARITY, ".RData")), message = FALSE}
```

```{r, echo = FALSE}
load(paste0(RDATA_PATH, paste0("processed_data_", POLARITY, ".RData")))
data_FS <- filterFile(data, which(data$mode == "FS"),
                      keepFeatures = TRUE)
```

## Signal intensity difference

We compute the difference in (log2) signals between samples with high and low
concentration of the standards and calculate the p-value for this difference
using the Student's t-test.

```{r, abundance-difference}
```

## Identification of features matching standards

We now identify all features matching the any of the pre-defined set of adducts
for the standards of mix `r MIX`. Matching features are further subsetted to
those that show an at least twice as high signal in samples with high
concentration compared to those with low concentrations. Also, features with a
signal in low concentration but no detectable signal in high concentration are
removed. Lastly, from mixture 12 onward the maximum retention time deviation 
from the expected 'Target_RT' is less than 10 seconds.

```{r, define-adducts-pos}
```
```{r, match-features}
```

For most of the standards (`r length(unique(mD$target_name))` out of 
`r nrow(std_dilution)`) in mix `r MIX` at least one feature was found
matching the standards adducts' m/z. 

```{r, table-standard-no-feature, results = "asis", echo = FALSE}
```

In the next sections we investigate for each standard which assignment would be
the correct one or, for those for which no signal was detected, why that was the
case.

## Standards evaluation (all, incl not found/matched)

While for some standards a matching feature was found we still need to evaluate
whether this matching is correct. For each standard we thus first evaluate the
EICs for all matching features, then we match their MS2 spectra (if available)
against reference libraries. To ensure correct assignment of a feature, its
retention time and eventually related MS2 spectra, we consider the following
criteria to determine the annotation confidence:

- feature(s) was/were found with m/z matching those of adduct(s) of the 
  standard.
- signal is higher for samples with higher concentration.
- MS2 spectra matches reference spectra for the standard (if available).
- MS2 spectra don't match MS2 spectra of other compounds.

### Adenine

```{r, echo = FALSE}
std <- "Adenine"
```

```{r, table-feature-matches, results = "asis", echo = FALSE, message = FALSE}
```

Many features have been assigned to this standard based on m/z matching. Based
on their retention times, these seem however to represent ions from different
compounds.

We next plot the EIC for the assigned feature and visually inspect these.

EICs (check D?):

```{r, echo = FALSE}
plot_eics(data, std, feature_table, "WP", std_ms2)
```

The cleaned MS2 spectra for the features matched to `r std` are shown below
(peaks with an intensity below 5% of the maximum peak and spectra with less
than 2 peaks were removed).

```{r mix19-serum-pos-Adenine-ms2, echo = FALSE}
plot_spectra(std_ms2)
```

We next match the extracted MS2 spectra against the reference spectra for 
`r std` from HMDB and MassBank.

Heat maps:

```{r mix19-serum-pos-Adenine-ms2-hmdb, echo = FALSE, fig.width = 5, fig.height = 5}
hmdb_id <- std_dilution$HMDB[std_dilution$name == std]
std_hmdb <- Spectra(cdb, filter = ~ compound_id == hmdb_id)
sim_hmdb <- plot_ms2_similarity_heatmap(std_ms2, std_hmdb, csp)
```

```{r mix19-serum-pos-Adenine-ms2-mbank, echo = FALSE, fig.width = 5, fig.height = 5}
std_mbank <- get_mbank(mbank, inchikey = unique(std_hmdb$inchikey))
sim_mbank <- plot_ms2_similarity_heatmap(std_ms2, std_mbank, csp)
```

For spectra from untargeted features that match with good similarity to some of 
the HMDB reference spectra for `r std`, we show the mirror plots of the best 
matches.

Mirror plots (check D -> C or B?):

```{r mix19-serum-pos-Adenine-mirror-hmdb, echo = TRUE, fig.cap = "Mirror plots"}
std_ms2_sel <- plot_select_ms2(std_ms2, std_hmdb, 0.7, ppm = csp@ppm,
                               tolerance = csp@tolerance, sim_hmdb)
```

```{r mix19-serum-pos-Adenine-mirror-mbank, echo = TRUE, fig.cap = "Mirror plots"}
tmp_sel <- plot_select_ms2(std_ms2, std_mbank, 0.7,
                           ppm = csp@ppm, tolerance = csp@tolerance, sim_mbank)
```

In addition we match (**all**) the MS2 spectra for the matched features against
all spectra from HMDB or MassBank identifying reference spectra with a
similarity larger than 0.7. The results (if any spectra matched) are shown in
the two following tables.

Highest confidence matches (check B -> A?):

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, hmdb, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, mbank, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

#### Summary

- good spectral match, one non-unique match with Adenosine so keep A for
  FT01064. But in massbank also less similar hits >0.7, so back to B. 

| label |   feature  | mzmed | ppm_error | rtmed | target_RT | feature_group |  adduct    | n_ms2 | mean_high | mean_low |
|:-----:|:----------:|:-----:|:---------:|:-----:|:---------:|:-------------:|:----------:|:-----:|:---------:|:--------:|
| **B** | FT01064    | 136.1 | 1.504     | 133.2 | 133       | FG.001.001    | [M+H]+     | 2     | 23.07     | 21.14    |
| **D** | FT00763    | 119   | 2.986     | 133.2 | 133       | FG.001.001    | [M+H-NH3]+ | 0     | 18.67     | 14.62    |

- Reference MS2: FT01064: F08.S0417, F07.S0427

```{r, echo = FALSE}
fts <- data.frame(feature_id = c("FT01064", "FT00763"),
                  confidence_level = c("B", "D"))
## Get MS2 spectra:
ms2 <- std_ms2[match(c("FT01064_F08.S0417", "FT01064_F07.S0427"),
                     spectraNames(std_ms2))]
ms2$confidence <- c("high", "high")
```

```{r, add-ions, echo = FALSE, results = "asis"}
```

```{r, add-ms2-spectra, echo = FALSE, results = "asis"}
```


### ATP

```{r, echo = FALSE}
std <- "ATP"
```

```{r, table-feature-matches-ATP, results = "asis", echo = FALSE, message = FALSE}
feature_table <- as.data.frame(mD[mD$target_name == std, ])
pandoc.table(feature_table, style = "rmarkdown",
             split.tables = Inf,
             caption = paste0("Feature to standards matches for ", std))
```

#### Summary

- No filtered feature table, no potential peaks found.
- No MS2 spectra available (because no potential peaks for `r std`).


### Homocysteinethiolactone

```{r, echo = FALSE}
std <- "Homocysteinethiolactone"
```

```{r, table-feature-matches, results = "asis", echo = FALSE, message = FALSE}
```

Many features have been assigned to this standard based on m/z matching. Based
on their retention times, these seem however to represent ions from different
compounds.

We next plot the EIC for the assigned feature and visually inspect these.

EICs (check D?):

```{r, echo = FALSE}
plot_eics(data, std, feature_table, "WP", std_ms2)
```

The cleaned MS2 spectra for the features matched to `r std` are shown below
(peaks with an intensity below 5% of the maximum peak and spectra with less
than 2 peaks were removed).

```{r mix19-serum-pos-Homocysteinethiolactone-ms2, echo = FALSE}
plot_spectra(std_ms2)
```

We next match the extracted MS2 spectra against the reference spectra for 
`r std` from HMDB and MassBank.

Heat maps:

```{r mix19-serum-pos-Homocysteinethiolactone-ms2-hmdb, echo = FALSE, fig.width = 5, fig.height = 5}
hmdb_id <- std_dilution$HMDB[std_dilution$name == std]
std_hmdb <- Spectra(cdb, filter = ~ compound_id == hmdb_id)
sim_hmdb <- plot_ms2_similarity_heatmap(std_ms2, std_hmdb, csp)
```

```{r mix19-serum-pos-Homocysteinethiolactone-ms2-mbank, echo = FALSE, fig.width = 5, fig.height = 5}
std_mbank <- get_mbank(mbank, inchikey = unique(std_hmdb$inchikey))
sim_mbank <- plot_ms2_similarity_heatmap(std_ms2, std_mbank, csp)
```

FYI: heatmap with 1 row / feature gives non-visual output.

For spectra from untargeted features that match with good similarity to some of 
the HMDB reference spectra for `r std`, we show the mirror plots of the best 
matches.

Mirror plots (check D -> C or B?):

```{r mix19-serum-pos-Homocysteinethiolactone-mirror-hmdb, echo = TRUE, fig.cap = "Mirror plots"}
std_ms2_sel <- plot_select_ms2(std_ms2, std_hmdb, 0.1, ppm = csp@ppm,
                               tolerance = csp@tolerance, sim_hmdb)
```

```{r mix19-serum-pos-Homocysteinethiolactone-mirror-mbank, echo = TRUE, fig.cap = "Mirror plots"}
tmp_sel <- plot_select_ms2(std_ms2, std_mbank, 0.5,
                           ppm = csp@ppm, tolerance = csp@tolerance, sim_mbank)
```

In addition we match (**all**) the MS2 spectra for the matched features against
all spectra from HMDB or MassBank identifying reference spectra with a
similarity larger than 0.7. The results (if any spectra matched) are shown in
the two following tables.

Highest confidence matches (check B -> A?):

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, hmdb, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, mbank, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

#### Summary

- Mirror plots D -> C for FT00746, but <0.7 match. not nicest EIC though.

| label |   feature  | mzmed | ppm_error | rtmed | target_RT | feature_group |  adduct   | n_ms2 | mean_high | mean_low |
|:-----:|:----------:|:-----:|:---------:|:-----:|:---------:|:-------------:|:---------:|:-----:|:---------:|:--------:|
| **C** | FT00746    | 118   | 1.501     | 125.7 | 124       | FG.001.001    | [M+H]+    | 1     | 21.95     | 17.8     |

```{r, echo = FALSE}
fts <- data.frame(feature_id = c("FT00746"),
                  confidence_level = c("C"))
## Get MS2 spectra:
ms2 <- std_ms2[match(c("FT00746_F08.S0388"),
                     spectraNames(std_ms2))]
ms2$confidence <- c("high")
```

```{r, add-ions, echo = FALSE, results = "asis"}
```

```{r, add-ms2-spectra, echo = FALSE, results = "asis"}
```


### Methylamine

```{r, echo = FALSE}
std <- "Methylamine"
```

```{r, table-feature-matches-Methylamine, results = "asis", echo = FALSE, message = FALSE}
feature_table <- as.data.frame(mD[mD$target_name == std, ])
pandoc.table(feature_table, style = "rmarkdown",
             split.tables = Inf,
             caption = paste0("Feature to standards matches for ", std))
```

#### Summary

- No filtered feature table, no potential peaks found.
- No MS2 spectra available (because no potential peaks for `r std`).


### Trimethylamine

```{r, echo = FALSE}
std <- "Trimethylamine"
```

```{r, table-feature-matches-Trimethylamine, results = "asis", echo = FALSE, message = FALSE}
feature_table <- as.data.frame(mD[mD$target_name == std, ])
pandoc.table(feature_table, style = "rmarkdown",
             split.tables = Inf,
             caption = paste0("Feature to standards matches for ", std))
```

#### Summary

- No filtered feature table, no potential peaks found.
- No MS2 spectra available (because no potential peaks for `r std`).


### UMP

```{r, echo = FALSE}
std <- "UMP"
```

```{r, table-feature-matches, results = "asis", echo = FALSE, message = FALSE}
```

Many features have been assigned to this standard based on m/z matching. Based
on their retention times, these seem however to represent ions from different
compounds.

We next plot the EIC for the assigned feature and visually inspect these.

EICs (check D?):

```{r, echo = FALSE}
plot_eics(data, std, feature_table, "WP", std_ms2)
```

The cleaned MS2 spectra for the features matched to `r std` are shown below
(peaks with an intensity below 5% of the maximum peak and spectra with less
than 2 peaks were removed).

```{r mix19-serum-pos-UMP-ms2, echo = FALSE}
plot_spectra(std_ms2)
```

We next match the extracted MS2 spectra against the reference spectra for 
`r std` from HMDB and MassBank.

Heat maps:

```{r mix19-serum-pos-UMP-ms2-hmdb, echo = FALSE, fig.width = 5, fig.height = 5}
hmdb_id <- std_dilution$HMDB[std_dilution$name == std]
std_hmdb <- Spectra(cdb, filter = ~ compound_id == hmdb_id)
sim_hmdb <- plot_ms2_similarity_heatmap(std_ms2, std_hmdb, csp)
```

```{r mix19-serum-pos-UMP-ms2-mbank, echo = FALSE, fig.width = 5, fig.height = 5}
std_mbank <- get_mbank(mbank, inchikey = unique(std_hmdb$inchikey))
sim_mbank <- plot_ms2_similarity_heatmap(std_ms2, std_mbank, csp)
```

For spectra from untargeted features that match with good similarity to some of 
the HMDB reference spectra for `r std`, we show the mirror plots of the best 
matches.

Mirror plots (check D -> C or B?):

```{r mix19-serum-pos-UMP-mirror-hmdb, echo = TRUE, fig.cap = "Mirror plots"}
std_ms2_sel <- plot_select_ms2(std_ms2, std_hmdb, 0.1, ppm = csp@ppm,
                               tolerance = csp@tolerance, sim_hmdb)
```

```{r mix19-serum-pos-UMP-mirror-mbank, echo = TRUE, fig.cap = "Mirror plots"}
tmp_sel <- plot_select_ms2(std_ms2, std_mbank, 0.7,
                           ppm = csp@ppm, tolerance = csp@tolerance, sim_mbank)
```

In addition we match (**all**) the MS2 spectra for the matched features against
all spectra from HMDB or MassBank identifying reference spectra with a
similarity larger than 0.7. The results (if any spectra matched) are shown in
the two following tables.

Highest confidence matches (check B -> A?):

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, hmdb, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, mbank, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

#### Summary

- FYI: UMP and URIDINEMONOPHOSPHATE are synonyms
- URIDINE5DIPHOSPHATE is very similar, so keep A for FT04738, but in hmdb also
  other matches, so back to B.
- Reference MS2: FT04738: F08.S0740, F08.S0793, F07.S0765, F07.S0806,
  F07.S0834

| label |   feature  | mzmed | ppm_error | rtmed | target_RT | feature_group |  adduct   | n_ms2 | mean_high | mean_low |
|:-----:|:----------:|:-----:|:---------:|:-----:|:---------:|:-------------:|:---------:|:-----:|:---------:|:--------:|
| **B** | FT04738    | 325   | 2.39      | 222.6 | 225       | FG.001.001    | [M+H]+    | 5     | 18.7      | 11.87    |
| **C** | FT09502    | 649.1 | 2.453     | 222.8 | 225       | FG.001.001    | [2M+H]+   | 2     | 15.11     | 2.721    |
| **D** | FT05284    | 347   | 4.308     | 222.9 | 225       | FG.001.001    | [M+Na]+   | 2     | 17.04     | 7.324    |
| **D** | FT05610    | 363   | 12.08     | 222.9 | 225       | FG.001.001    | [M+K]+    | 2     | 15.16     | 5.591    |

```{r, echo = FALSE}
fts <- data.frame(feature_id = c("FT04738", "FT09502", 'FT05284', 'FT05610'),
                  confidence_level = c("B", "C", 'D', 'D'))
## Get MS2 spectra:
ms2 <- std_ms2[match(c("FT04738_F08.S0740", "FT04738_F08.S0793", "FT04738_F07.S0765", "FT04738_F07.S0806", "FT04738_F07.S0834",
                       'FT09502_F08.S0746'),
                     spectraNames(std_ms2))]
ms2$confidence <- c("low", 'high', 'low', 'low', 'low',
                    "high")
```

```{r, add-ions, echo = FALSE, results = "asis"}
```

```{r, add-ms2-spectra, echo = FALSE, results = "asis"}
```


# Serum, negative polarity

We now perform the analysis on the samples with the standards still solved in
pure serum but acquired in negative polarity mode.

```{r}
POLARITY <- "NEG"
data <- filterFile(data_all, which(data_all$polarity == POLARITY))
hmdb <- hmdb_neg
hmdb_nl <- hmdb_neg_nl
```

```{r, preprocessing, eval = !file.exists(paste0(RDATA_PATH, "processed_data_", POLARITY, ".RData")), message = FALSE}
```

```{r, echo = FALSE}
load(paste0(RDATA_PATH, paste0("processed_data_", POLARITY, ".RData")))
data_FS <- filterFile(data, which(data$mode == "FS"),
                      keepFeatures = TRUE)
```


## Signal intensity difference

We compute the difference in (log2) signals between samples with high and low
concentration of the standards and calculate the p-value for this difference
using the Student's t-test.

```{r, abundance-difference}
```

## Identification of features matching standards

We now identify all features matching the any of the pre-defined set of adducts
for the standards of mix `r MIX`. Matching features are further subsetted to
those that show an at least twice as high signal in samples with high
concentration compared to those with low concentrations. Also, features with a
signal in low concentration but no detectable signal in high concentration are
removed.

```{r, define-adducts-neg}
```
```{r, match-features}
```

For most of the standards (`r length(unique(mD$target_name))` out of 
`r nrow(std_dilution)`) in mix `r MIX` at least one feature was found
matching the standards adducts' m/z. 

```{r, table-standard-no-feature, results = "asis", echo = FALSE}
```

In the next sections we investigate for each standard which assignment would be
the correct one or, for those for which no signal was detected, why that was the
case.

## Standards evaluation (all, incl not found/matched)

While for some standards a matching feature was found we still need to evaluate
whether this matching is correct. For each standard we thus first evaluate the
EICs for all matching features, then we match their MS2 spectra (if available)
against reference libraries. To ensure correct assignment of a feature, its
retention time and eventually related MS2 spectra, we consider the following
criteria to determine the annotation confidence:

- feature(s) was/were found with m/z matching those of adduct(s) of the 
  standard.
- signal is higher for samples with higher concentration.
- MS2 spectra matches reference spectra for the standard (if available).
- MS2 spectra don't match MS2 spectra of other compounds.

### Adenine

```{r, echo = FALSE}
std <- "Adenine"
```

```{r, table-feature-matches, results = "asis", echo = FALSE, message = FALSE}
```

Many features have been assigned to this standard based on m/z matching. Based
on their retention times, these seem however to represent ions from different
compounds.

We next plot the EIC for the assigned feature and visually inspect these.

EICs (check D?):

```{r, echo = FALSE}
plot_eics(data, std, feature_table, "WP", std_ms2)
```

The cleaned MS2 spectra for the features matched to `r std` are shown below
(peaks with an intensity below 5% of the maximum peak and spectra with less
than 2 peaks were removed).

```{r mix19-serum-neg-Adenine-ms2, echo = FALSE}
plot_spectra(std_ms2)
```

We next match the extracted MS2 spectra against the reference spectra for 
`r std` from HMDB and MassBank.

Heat maps:

```{r mix19-serum-neg-Adenine-ms2-hmdb, echo = FALSE, fig.width = 5, fig.height = 5}
hmdb_id <- std_dilution$HMDB[std_dilution$name == std]
std_hmdb <- Spectra(cdb, filter = ~ compound_id == hmdb_id)
sim_hmdb <- plot_ms2_similarity_heatmap(std_ms2, std_hmdb, csp)
```

```{r mix19-serum-neg-Adenine-ms2-mbank, echo = FALSE, fig.width = 5, fig.height = 5}
std_mbank <- get_mbank(mbank, inchikey = unique(std_hmdb$inchikey))
sim_mbank <- plot_ms2_similarity_heatmap(std_ms2, std_mbank, csp)
```

For spectra from untargeted features that match with good similarity to some of 
the HMDB reference spectra for `r std`, we show the mirror plots of the best 
matches.

Mirror plots (check D -> C or B?):

```{r mix19-serum-neg-Adenine-mirror-hmdb, echo = TRUE, fig.cap = "Mirror plots"}
std_ms2_sel <- plot_select_ms2(std_ms2, std_hmdb, 0.7, ppm = csp@ppm,
                               tolerance = csp@tolerance, sim_hmdb)
```

```{r mix19-serum-neg-Adenine-mirror-mbank, echo = TRUE, fig.cap = "Mirror plots"}
tmp_sel <- plot_select_ms2(std_ms2, std_mbank, 0.7,
                           ppm = csp@ppm, tolerance = csp@tolerance, sim_mbank)
```

In addition we match (**all**) the MS2 spectra for the matched features against
all spectra from HMDB or MassBank identifying reference spectra with a
similarity larger than 0.7. The results (if any spectra matched) are shown in
the two following tables.

Highest confidence matches (check B -> A?):

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, hmdb, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, mbank, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

#### Summary

- good spectral match, one non-unique match with Adenosine and Adenosine-related 
  so keep A for FT0283. But in massbank also less similar hits >0.7, so back to
  B. also high ppm.

| label |   feature  | mzmed | ppm_error | rtmed | target_RT | feature_group |  adduct   | n_ms2 | mean_high | mean_low |
|:-----:|:----------:|:-----:|:---------:|:-----:|:---------:|:-------------:|:---------:|:-----:|:---------:|:--------:|
| **B** | FT0283     | 134   | 15.57     | 133.6 | 133       | FG.001.001    | [M-H]-    | 2     | 21        | 19.3     |
| **D** | FT0703     | 180   | 17.74     | 132.5 | 133       | FG.001.001    | [M+CHO2]- | 0     | 15.61     | 13.41    |

- FYI: FT1742 has only good EIC for low abundance samples, so not asign D.

```{r, echo = FALSE}
fts <- data.frame(feature_id = c("FT0283", "FT0703"),
                  confidence_level = c("B", "D"))
## Get MS2 spectra:
ms2 <- std_ms2[match(c("FT0283_F15.S0343", "FT0283_F16.S0372"),
                     spectraNames(std_ms2))]
ms2$confidence <- c("high", 'high')
```

```{r, add-ions, echo = FALSE, results = "asis"}
```

```{r, add-ms2-spectra, echo = FALSE, results = "asis"}
```


### ATP

```{r, echo = FALSE}
std <- "ATP"
```

```{r, table-feature-matches-neg-ATP, results = "asis", echo = FALSE, message = FALSE}
feature_table <- as.data.frame(mD[mD$target_name == std, ])
pandoc.table(feature_table, style = "rmarkdown",
             split.tables = Inf,
             caption = paste0("Feature to standards matches for ", std))
```

#### Summary

- No filtered feature table, no potential peaks found.
- No MS2 spectra available (because no potential peaks for `r std`).


### Homocysteinethiolactone

```{r, echo = FALSE}
std <- "Homocysteinethiolactone"
```

```{r, table-feature-matches-Homocysteinethiolactone, results = "asis", echo = FALSE, message = FALSE}
feature_table <- as.data.frame(mD[mD$target_name == std, ])
pandoc.table(feature_table, style = "rmarkdown",
             split.tables = Inf,
             caption = paste0("Feature to standards matches for ", std))
```

#### Summary

- No filtered feature table, no potential peaks found.
- No MS2 spectra available (because no potential peaks for `r std`).


### Methylamine

```{r, echo = FALSE}
std <- "Methylamine"
```

```{r, table-feature-matches-neg-Methylamine, results = "asis", echo = FALSE, message = FALSE}
feature_table <- as.data.frame(mD[mD$target_name == std, ])
pandoc.table(feature_table, style = "rmarkdown",
             split.tables = Inf,
             caption = paste0("Feature to standards matches for ", std))
```

#### Summary

- No filtered feature table, no potential peaks found.
- No MS2 spectra available (because no potential peaks for `r std`).


### Trimethylamine

```{r, echo = FALSE}
std <- "Trimethylamine"
```

```{r, table-feature-matches-neg-Trimethylamine, results = "asis", echo = FALSE, message = FALSE}
feature_table <- as.data.frame(mD[mD$target_name == std, ])
pandoc.table(feature_table, style = "rmarkdown",
             split.tables = Inf,
             caption = paste0("Feature to standards matches for ", std))
```

#### Summary

- No filtered feature table, no potential peaks found.
- No MS2 spectra available (because no potential peaks for `r std`).


### UMP

```{r, echo = FALSE}
std <- "UMP"
```

```{r, table-feature-matches, results = "asis", echo = FALSE, message = FALSE}
```

Many features have been assigned to this standard based on m/z matching. Based
on their retention times, these seem however to represent ions from different
compounds.

We next plot the EIC for the assigned feature and visually inspect these.

EICs (check D?):

```{r, echo = FALSE}
plot_eics(data, std, feature_table, "WP", std_ms2)
```

The cleaned MS2 spectra for the features matched to `r std` are shown below
(peaks with an intensity below 5% of the maximum peak and spectra with less
than 2 peaks were removed).

```{r mix19-serum-neg-UMP-ms2, echo = FALSE}
plot_spectra(std_ms2)
```

We next match the extracted MS2 spectra against the reference spectra for 
`r std` from HMDB and MassBank.

Heat maps:

```{r mix19-serum-neg-UMP-ms2-hmdb, echo = FALSE, fig.width = 5, fig.height = 5}
hmdb_id <- std_dilution$HMDB[std_dilution$name == std]
std_hmdb <- Spectra(cdb, filter = ~ compound_id == hmdb_id)
sim_hmdb <- plot_ms2_similarity_heatmap(std_ms2, std_hmdb, csp)
```

```{r mix19-serum-neg-UMP-ms2-mbank, echo = FALSE, fig.width = 5, fig.height = 5}
std_mbank <- get_mbank(mbank, inchikey = unique(std_hmdb$inchikey))
sim_mbank <- plot_ms2_similarity_heatmap(std_ms2, std_mbank, csp)
```

For spectra from untargeted features that match with good similarity to some of 
the HMDB reference spectra for `r std`, we show the mirror plots of the best 
matches.

Mirror plots (check D -> C or B?):

```{r mix19-serum-neg-UMP-mirror-hmdb, echo = TRUE, fig.cap = "Mirror plots"}
std_ms2_sel <- plot_select_ms2(std_ms2, std_hmdb, 0.5, ppm = csp@ppm,
                               tolerance = csp@tolerance, sim_hmdb)
```

```{r mix19-serum-neg-UMP-mirror-mbank, echo = TRUE, fig.cap = "Mirror plots"}
tmp_sel <- plot_select_ms2(std_ms2, std_mbank, 0.4,
                           ppm = csp@ppm, tolerance = csp@tolerance, sim_mbank)
```

In addition we match (**all**) the MS2 spectra for the matched features against
all spectra from HMDB or MassBank identifying reference spectra with a
similarity larger than 0.7. The results (if any spectra matched) are shown in
the two following tables.

Highest confidence matches (check B -> A?):

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, hmdb, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

```{r, echo = FALSE, message = FALSE, results = "asis"}
perform_match(std_ms2, mbank, sv = c("rtime", "target_name", "score"),
              name = "target_name", param = csp)
```

#### Summary

- good sprectra for FT2528, but in massbank got 2 other hits, so B. high ppm

| label |   feature  | mzmed | ppm_error | rtmed | target_RT | feature_group |  adduct   | n_ms2 | mean_high | mean_low |
|:-----:|:----------:|:-----:|:---------:|:-----:|:---------:|:-------------:|:---------:|:-----:|:---------:|:--------:|
| **B** | FT2528     | 323   | 12.19     | 226.4 | 225       | FG.001.001    | [M-H]-    | 5     | 21.85     | 15.86    |
| **D** | FT6303     | 647.1 | 4.172     | 226.8 | 225       | FG.001.001    | [2M-H]-   | 2     | 20.64     | 2.679    |

```{r, echo = FALSE}
fts <- data.frame(feature_id = c("FT2528", "FT6303"),
                  confidence_level = c("B", "D"))
## Get MS2 spectra:
ms2 <- std_ms2[match(c("FT2528_F16.S0655", "FT2528_F15.S0629", "FT2528_F15.S0676", "FT2528_F15.S0699", "FT2528_F16.S0701"),
                     spectraNames(std_ms2))]
ms2$confidence <- c("high", 'high', "high", "high", "high")
```

```{r, add-ions, echo = FALSE, results = "asis"}
```

```{r, add-ms2-spectra, echo = FALSE, results = "asis"}
```


# Summary on the ion database

Summarizing the content that was added to the `IonDb`.

```{r, iondb-summary, echo = FALSE, results = "asis"}
```

# Session information

The R version and packages used in this analysis are listed below.

```{r sessioninfo}
sessionInfo()
```