---
title: "Usage of Annotation Resources with the CompoundDb Package"
output:
    BiocStyle::html_document:
        toc_float: true
vignette: >
    %\VignetteIndexEntry{Usage of Annotation Resources with the CompoundDb Package}
    %\VignetteEngine{knitr::rmarkdown}
    %\VignetteEncoding{UTF-8}
    %\VignettePackage{CompoundDb}
    %\VignetteDepends{CompoundDb,RSQLite,Spectra,BiocStyle}
bibliography: references.bib
---

```{r style, echo = FALSE, results = 'asis', message=FALSE}
BiocStyle::markdown()
```

**Authors**: `r packageDescription("CompoundDb")[["Author"]] `<br />
**Last modified:** `r file.info("CompoundDb-usage.Rmd")$mtime`<br />
**Compiled**: `r date()`

```{r, echo = FALSE, message = FALSE}
library(CompoundDb)
library(BiocStyle)
knitr::opts_chunk$set(echo = TRUE, message = FALSE)
```


# Introduction

The `r Biocpkg("CompoundDb")` package provides the functionality to create
chemical *compound* databases from a variety of sources and to use such
annotation databases (`CompDb`) [@rainer_modular_2022]. A detailed description
on the creation of annotation resources is given in the *Creating CompoundDb
annotation resources* vignette. This vignette focuses on how annotations can be
search for and retrieved.


# Installation

The package (including dependencies) can be installed with the code below:

```{r, eval = FALSE}
install.packages("BiocManager")
BiocManager::install("CompoundDb")
```


# General usage

In this vignette we use a small `CompDb` database containing annotations for a
small number of metabolites build using
[MassBank](https://massbank.eu/MassBank/) release *2020.09*. The respective
`CompDb` database which is loaded below contains in addition to general compound
annotations also MS/MS spectra for these compounds.

```{r load}
library(CompoundDb)
cdb <- CompDb(system.file("sql/CompDb.MassBank.sql", package = "CompoundDb"))
cdb
```

General information about the database can be accessed with the `metadata`
function.

```{r}
metadata(cdb)
```


## Querying compound annotations

The `CompoundDb` package is designed to provide annotation resources for small
molecules, such as metabolites, that are characterized by an exact mass and
additional information such as their IUPAC International Chemical Identifier
[InChI](https://en.wikipedia.org/wiki/International_Chemical_Identifier) or
their chemical formula. The available annotations (*variables*) for compounds
can differ between databases. The `compoundVariables()` function can be used to
retrieve a list of all available compound annotations for a specific `CompDb`
database.

```{r}
compoundVariables(cdb)
```

The actual compound annotations can then be extracted with the `compounds()`
function which returns by default all columns listed by
`compoundVariables()`. We can also define specific columns we want to extract
with the `columns` parameter.

```{r}
head(compounds(cdb, columns = c("name", "formula", "exactmass")))
```

As a technical detail, `CompDb` databases follow a very simple database layout
with only few constraints to allow data import and representation for a variety
of sources (e.g. MassBank, HMDB, MoNa, ChEBI). For the present database, which
is based on MassBank, the mapping between entries in the *ms_compound* database
table and MS/MS spectra is for example 1:1 and the *ms_compound* table contains
thus highly redundant information. Thus, if we would include the column
`"compound_id"` in the query we would end up with redundant values:

```{r}
head(compounds(cdb, columns = c("compound_id", "name", "formula")))
```

By default, `compounds()` extracts the data for **all** compounds stored in the
database. The function supports however also *filters* to get values for
specific entries only. These can be defined as *filter expressions* which are
similar to the way how e.g. a `data.frame` would be subsetted in R. In the
example below we extract the compound ID, name and chemical formula for a
compound *Mellein*.

```{r}
compounds(cdb, columns = c("compound_id", "name", "formula"),
          filter = ~ name == "Mellein")
```

Note that a filter expression always has to start with `~` followed by the
*variable* on which the data should be subsetted and the condition to select the
entries of interest. An overview of available filters for a `CompDb` can be
retrieved with the `supportedFilter()` function which returns the name of the
filter and the database column on which the filter selects the values:

```{r}
supportedFilters(cdb)
```

Also, filters can be combined to create more specific filters in the same manner
this would be done in R, i.e. using `&` for *and*, `|` for *or* and `!` for
*not*. To illustrate this we extract below all compound entries from the table
for compounds with the name *Mellein* and that have a `"compound_id"` which is
either 1 or 5.

```{r}
compounds(cdb, columns = c("compound_id", "name", "formula"),
          filter = ~ name == "Mellein" & compound_id %in% c(1, 5))
```

Similarly, we can define a filter expression to retrieve compounds with an exact
mass between 310 and 320.

```{r}
compounds(cdb, columns = c("name", "exactmass"),
          filter = ~ exactmass > 310 & exactmass < 320)
```

In addition to *filter expressions*, we can also define and combine filters
using the actual filter classes. This provides additional conditions that would
not be possible with regular filter expressions. Below we fetch for examples
only compounds from the database that contain a *H14* in their formula. To this
end we use a `FormulaFilter` with the condition `"contains"`. Note that all
filters that base on character matching (i.e. `FormulaFilter`, `InchiFilter`,
`InchikeyFilter`, `NameFilter`) support as conditions also `"contains"`,
`"startsWith"` and `"endsWith"` in addition to `"="` and `"!="`.

```{r}
compounds(cdb, columns = c("name", "formula", "exactmass"),
          filter = FormulaFilter("H14", "contains"))
```

It is also possible to combine filters if they are defined that way, even if it
is a little less straight forward than with the filter expressions. Below we
combine the `FormulaFilter` with the `ExactmassFilter` to retrieve only
compounds with an `"H14"` in their formula and an exact mass between 310 and
320.

```{r}
filters <- AnnotationFilterList(
    FormulaFilter("H14", "contains"),
    ExactmassFilter(310, ">"),
    ExactmassFilter(320, "<"),
    logicOp = c("&", "&"))
compounds(cdb, columns = c("name", "formula", "exactmass"),
          filter = filters)
```


## Additional functionality for `CompDb` databases

*CompoundDb* defines additional functions to work with `CompDb` databases. One
of them is the `mass2mz()` function that allows to directly calculate ion
(adduct) m/z values for exact (monoisotopic) masses of compounds in a
database. Below we use this function to calculate `[M+H]+` and `[M+Na]+` ions
for all unique chemical formulas in our example `CompDb` database.

```{r}
mass2mz(cdb, adduct = c("[M+H]+", "[M+Na]+"))
```

To get a `matrix` with adduct m/z values for discrete compounds (identified
by their InChIKey) we specify `name = "inchikey"`.

```{r}
mass2mz(cdb, adduct = c("[M+H]+", "[M+Na]+"), name = "inchikey")
```

Alternatively we could also use `name = "compound_id"` to get a value for each
row in the compound database table, but for this example database this would
result in highly redundant information.

`mass2mz()` bases on the `MetaboCoreUtils::mass2mz` function and thus supports
all pre-defined adducts from that function. These are (for positive polarity):

```{r}
MetaboCoreUtils::adductNames()
```

and for negative polarity:

```{r}
MetaboCoreUtils::adductNames(polarity = "negative")
```

In addition, user-supplied adduct definitions are also supported (see the help
of `mass2mz()` in the `r Biocpkg("MetaboCoreUtils")` package for details).


## Accessing and using MS/MS data

`CompDb` database can also store and provide MS/MS spectral data. These can be
accessed *via* a `Spectra` object from the `r Biocpkg("Spectra")`
Bioconductor. Such a `Spectra` object for a `CompDb` can be created with the
`Spectra()` function as in the example below.

```{r}
sps <- Spectra(cdb)
sps
```

This `Spectra` object uses a `MsBackendCompDb` to *represent* the MS data of the
`CompDb` database. In fact, only the compound identifiers and the precursor m/z
values from all spectra are stored in memory while all other data is retrieved
on-the-fly from the database when needed.

The `spectraVariables()` function lists all available annotations for a spectrum
from the database, which includes also annotations of the associated compounds.

```{r}
spectraVariables(sps)
```

Individual variables can then be accessed with `$` and the variable name:

```{r}
head(sps$adduct)
```

For more information on how to use `Spectra` objects in your analysis have also
a look at the package
[vignette](https://rformassspectrometry.github.io/Spectra/articles/Spectra.html)
or a [tutorial](https://jorainer.github.io/SpectraTutorials/) on how to perform
MS/MS spectra matching with `Spectra`.

Similar to the `compounds()` function, a call to `Spectra()` will give access to
**all** spectra in the database. Using the same filtering framework it is
however also possible to *extract* only specific spectra from the
database. Below we are for example accessing only the MS/MS spectra of the
compound *Mellein*. Using the `filter` in the `Spectra()` call can be
substantially faster than first initializing a `Spectra` with the full data and
then subsetting that to selected spectra.

```{r}
mellein <- Spectra(cdb, filter = ~ name == "Mellein")
mellein
```

Instead of all spectra we extracted now only a subset of `r length(mellein)`
spectra from the database.

As a simple toy example we perform next pairwise spectra comparison between the
5 spectra from *Mellein* with all the MS/MS spectra in the database.

```{r, message = FALSE}
library(Spectra)
cormat <- compareSpectra(mellein, sps, ppm = 40)
```

Note that the `MsBackendCompDb` does not support parallel processing, thus,
while `compareSpectra()` would in general support parallel processing, it gets
automatically be disabled if a `Spectra` with a `MsBackendCompDb` is used.

```{r}
cormat <- compareSpectra(mellein, sps, ppm = 40, BPPARAM = MulticoreParam(2))
```


# Ion databases

The `CompDb` database layout is designed to provide compound annotations, but in
mass spectrometry (MS) ions are measured. These ions are generated e.g. by
electro spray ionization (ESI) from the original compounds in a sample. They are
characterized by their specific mass-to-charge ratio (m/z) which is measured by
the MS instrument. Eventually, also a retention time is available. Also, for the
same compound several different ions (adducts) can be formed and measured, all
with a different m/z. This type of data can be represented by an `IonDb`
database, which extends the `CompDb` and hence inherits all of its properties
but adds additional database tables to support also ion annotations. Also,
`IonDb` objects provide functionality to add new ion annotations to an existing
database. Thus, this type of database can be used to build lab-internal
annotation resources containing ions, m/z and retention times for pure standards
measured on a specific e.g. LC-MS setup.

`CompDb` databases, such as the `cdb` from this example, are however by default
*read-only*, thus, we below create a new database connection, copy the content
of the `cdb` to that database and convert the `CompDb` to an `IonDb`.

```{r}
library(RSQLite)
## Create a temporary database
con <- dbConnect(SQLite(), tempfile())
## Create an IonDb copying the content of cdb to the new database
idb <- IonDb(con, cdb)
idb
```

The `IonDb` defines an additional function `ions` that allows to retrieve ion
information from the database.

```{r}
ions(idb)
```

The present database does not yet contain any ion information. Below we define a
data frame with ion annotations and add that to the database with the
`insertIon()` function. The column `"compound_id"` needs to contain the
identifiers of the compounds to which the ion should be related to. In the
present example we add 2 different ions for the compound with the ID 1
(*Mellein*). Note that the specified m/z values as well as the retention times
are completely arbitrary.

```{r}
ion <- data.frame(compound_id = c(1, 1),
                  ion_adduct = c("[M+H]+", "[M+Na]+"),
                  ion_mz = c(123.34, 125.34),
                  ion_rt = c(196, 196))
idb <- insertIon(idb, ion)
```

These ions have now be added to the database.

```{r}
ions(idb)
```

Ions can also be deleted from a database with the `deleteIon` function (see the
respective help page for more information).

Note that we can also retrieve compound annotation information for the
ions. Below we extract the associated compound name and its exact mass.

```{r}
ions(idb, columns = c("ion_adduct", "name", "exactmass"))
```


# Session information

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

# References