--- title: "Load Single-Cell Proteomics data using `readSCP`" author: - name: Laurent Gatto - name: Christophe Vanderaa output: BiocStyle::html_document: self_contained: yes toc: true toc_float: true toc_depth: 2 code_folding: show bibliography: scp.bib date: "`r BiocStyle::doc_date()`" package: "`r BiocStyle::pkg_ver('scp')`" vignette: > %\VignetteIndexEntry{Load data using readSCP} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{r setup, include = FALSE} knitr::opts_chunk$set( collapse = TRUE, comment = "#>", crop = NULL ## cf to https://stat.ethz.ch/pipermail/bioc-devel/2020-April/016656.html ) ``` # The `scp` data framework Our data structure is relying on two curated data classes: `QFeatures` (@Gatto2020-ry) and `SingleCellExperiment` (@Amezquita2019-bf). `QFeatures` is dedicated to the manipulation and processing of MS-based quantitative data. It explicitly records the successive steps to allow users to navigate up and down the different MS levels. `SingleCellExperiment` is another class designed as an efficient data container that serves as an interface to state-of-the-art methods and algorithms for single-cell data. Our framework combines the two classes to inherit from their respective advantages. Because mass spectrometry (MS)-based single-cell proteomics (SCP) only captures the proteome of between one and a few tens of single-cells in a single run, the data is usually acquired across many MS batches. Therefore, the data for each run should conceptually be stored in its own container, that we here call an *assay*. The expected input for working with the `scp` package is quantification data of peptide to spectrum matches (PSM). These data can then be processed to reconstruct peptide and protein data. The links between related features across different assays are stored to facilitate manipulation and visualization of of PSM, peptide and protein data. This is conceptually shown below. ```{r, fig.cap="The `scp` framework relies on `SingleCellExperiment` and `QFeatures` objects", echo=FALSE, out.width='100%', fig.align='center'} knitr::include_graphics("./figures/SCP_framework.png") ``` There are two input tables required for starting an analysis with `scp`: 1. The input table 2. The sample table # Input table The input table is generated after the identification and quantification of the MS spectra by a pre-processing software such as MaxQuant, ProteomeDiscoverer or MSFragger (the [list](https://en.wikipedia.org/wiki/List_of_mass_spectrometry_software) of available software is actually much longer). We will here use as an example a data table that has been generated by MaxQuant. The table is available from the `scp` package and is called `mqScpData` (for MaxQuant generated SCP data). ```{r, message = FALSE} library(scp) data("mqScpData") dim(mqScpData) ``` In this toy example, there are 1361 rows corresponding to features (quantified PSMs) and 149 columns corresponding to different data fields recorded by MaxQuant during the processing of the MS spectra. There are three types of columns: - Feature quantification: 1 to n (depending on technology) - Feature annotations: *e.g.* peptide sequence, ion charge, protein name - Acquisition annotations: *e.g.* file name ```{r, echo=FALSE, out.width='60%', fig.cap="Conceptual representation of the input table", fig.align = 'center'} knitr::include_graphics('figures/readSCP_inputTable.png') ``` ### Feature quantifications The quantification data can be composed of one (in case of label-free acquisition) up to 16 columns (in case of TMT-16 multiplexing). The columns holding the quantification start with `Reporter.intensity.` followed by a number. ```{r} (quantCols <- grep("Reporter.intensity.\\d", colnames(mqScpData), value = TRUE)) ``` As you may notice, the example data was acquired using a TMT-16 protocol since we retrieve 16 quantification columns. Actually, some runs were acquired using a TMT-11 protocol (11 labels) but we will come back to this later. ```{r} head(mqScpData[, quantCols]) ``` ### Feature annotations Most columns in the `mqScpData` table contain information used or generated during the identification of the MS spectra. For instance, you may find the charge of the parent ion, the score and probability of a correct match between the MS spectrum and a peptide sequence, the sequence of the best matching peptide, its length, its modifications, the retention time of the peptide on the LC, the protein(s) the peptide originates from and much more. ```{r} head(mqScpData[, c("Charge", "Score", "PEP", "Sequence", "Length", "Retention.time", "Proteins")]) ``` ### Acquisition annotations This type of annotation is related to the MS instrument. In MaxQuant, only the file name generated by the MS instrument is stored. There is one file for each MS run, hence the file name can be used as a batch identifier. ```{r} unique(mqScpData$Raw.file) ``` # Sample table The sample table contains the experimental design generated by the researcher. The rows of the sample table correspond to a sample in the experiment and the columns correspond to the available annotations about the sample. We will here use the second example table: ```{r} data("sampleAnnotation") head(sampleAnnotation) ``` This table may contain any information about the samples. For example, useful information could be the type of sample that is analysed, a phenotype known from the experimental design, the MS batch, the acquisition date, MS settings used to acquire the sample, the LC batch, the sample preparation batch, etc... However, `scp` **requires** 2 specific fields in the sample annotations: 1. One column that tells `scp` the names of the columns in the feature data holds the quantification of the corresponding sample. 2. One column containing the MS run names (`Raw.file` in this case). It must have the same name as the name of the column containing the MS run names in the quantification table. These two columns allow `scp` to correctly split and match data that were acquired across multiple acquisition runs. ```{r echo=FALSE, out.width='60%', fig.cap="Conceptual representation of the sample table", fig.align = 'center'} knitr::include_graphics('figures/readSCP_sampleTable.png') ``` # `readSCP` `readSCP` is the function that converts the sample and the feature data into a `QFeatures` object following the data structure described above, that is storing the data belonging to each MS batch in a separate `SingleCellExperiment` object. We therefore provide the feature data, the sample data to the function as well as the name of the column that holds the batch name in both tables and the name of the column in the sample data that points to the quantification columns in the feature data. ## Sample names and `suffix` `readSCP` automatically assigns names that are unique across all samples in all assays. This is performed by appending the name of the batch where a given sample is found in with the name of the quantification column for that sample. Suppose a sample belongs to batch `190222S_LCA9_X_FP94BM` and the quantification values in the feature data are found in the column called `Reporter.intensity.3`, then the sample name will become `190222S_LCA9_X_FP94BMReporter.intensity.3`. Optionally, to improve the readability of sample names, `readSCP` can take a suffix instead of the quantification column name. For instance, in the example below, we will provide a short suffix with the TMT index to remind that samples were multiplexed using TMT. ## Special case: empty samples In some rare cases, it can be beneficial to remove empty samples (all quantifications are `NA`) from the assays. Such samples can occur when samples that were acquired with different multiplexing labels are merged in a single table. For instance, the SCoPE2 data we provide as an example contains runs that were acquired with two TMT protocols. The 3 first assays were acquired using the TMT-11 protocol and the last assay was acquired using a TMT-16 protocol. When exporting the table, the authors combined the data in a single table, were missing channels in the TMT-11 data are filled with `NA`. This is essential when working in table format, but since `scp` keeps the runs separated we can allow for different numbers of channels per run. When setting `removeEmptyCols = TRUE`, `readSCP` automatically detects and removes columns that contain only `NA`s, ## Running `readSCP` We convert the sample and the feature data into a `QFeatures` object in a single command thanks to `readSCP`. ```{r readSCP} scp <- readSCP(featureData = mqScpData, colData = sampleAnnotation, batchCol = "Raw.file", channelCol = "Channel", suffix = paste0("_TMT", 1:16), removeEmptyCols = TRUE) scp ``` We can see that the object returned by `readSCP` is a `QFeatures` object containing 4 `SingleCellExperiment` assays that have been named after the 4 MS batches. Each assay contains either 11 or 16 columns (samples) depending on the TMT labelling strategy and a variable number of rows (quantified PSMs). Each piece of information can easily be retrieved thanks to the `Qfeatures` architectures. As mentioned in the previous vignette, sample data is retrieved from the `colData`. Note that unique sample names were automatically generated by combining the batch name and the suffix provided to `readSCP`: ```{r colData} head(colData(scp)) ``` Notice that the sample names were suffixed with TMT indexes. The feature annotations are retrieved from the `rowData`. Since the feature annotations are specific to each assay, we need to tell from which assay we want to get the `rowData`: ```{r rowData} head(rowData(scp[["190222S_LCA9_X_FP94BM"]]))[, 1:5] ``` Finally, we can also retrieve the quantification matrix for an assay of interest: ```{r assay} head(assay(scp, "190222S_LCA9_X_FP94BM")) ``` ## Under the hood `readSCP` proceeds as follows: 1. If `featureData` is the path to a CSV file, it reads the file using `read.csv`. The table is converted to a `SingleCellExperiment` object. `readSCP` needs to know in which field(s) the quantitative data is stored. Those field name(s) is/are provided by the `channelCol` field in the annotation table (`colData` argument). So in this example, the column names holding the quantitative data in `mqScpData` are stored in the column named `Channel` in `sampleAnnotation`. ```{r echo=FALSE, out.width='60%', fig.cap="Step1: Convert the input table to a `SingleCellExperiment` object", fig.align = 'center'} knitr::include_graphics('figures/readSCP_step1.png') ``` 2. The `SingleCellExperiment` object is then split according to the acquisition run. The split is performed depending on the `batchCol` field in `featureData`, in this case the data will be split according to the `Raw.file` column in `mqScpData`. `Raw.file` contains the names of the acquisition runs that was used by MaxQuant to retrieve the raw data files. ```{r echo=FALSE, out.width='65%', fig.cap="Step2: Split by acquisition run", fig.align = 'center'} knitr::include_graphics('figures/readSCP_step2.png') ``` 3. The sample annotations is generated from the supplied sample table (`colData` argument). Note that in order for `readSCP` to correctly match the feature data with the annotations, `colData` must contain the same `batchCol` field with batch names. ```{r echo=FALSE, out.width='100%', fig.cap="Step3: Adding and matching the sample annotations", fig.align = 'center'} knitr::include_graphics('figures/readSCP_step3.png') ``` 4. Finally, the split feature data and the sample annotations are stored in a single `QFeatures` object. ```{r echo=FALSE, out.width='80%', fig.cap="Step4: Convert to a `QFeatures`", fig.align = 'center'} knitr::include_graphics('figures/readSCP_step4.png') ``` # What about label-free SCP? The `scp` package is meant for both label-free and multiplexed SCP data. Like in the example above, the label-free data should contain the batch names in both the feature data and the sample data. The sample data must also contain a column that points to the columns of the feature data that contains the quantifications. Since label-free SCP acquires one single-cell per run, this sample data column should point the same column for all samples. Moreover, this means that each PSM assay will contain a single column. # What about other input formats? `readSCP` should work with any PSM quantification table that is output by a pre-processing software. For instance, you can easily import the PSM tables generated by ProteomeDiscoverer. The batch names are contained in the `File ID` column (that should be supplied as the `batchCol` argument to `readSCP`). The quantification columns are contained in the columns starting with `Abundance `, eventually followed by a multiplexing tag name. These columns should be stored in a dedicated column of the sample data to be supplied as `channelCol` to `readSCP`. If your input cannot be loaded using the procedure described in this vignette, you can submit a feature request (see next section). # Need help? You can open an issue on the [GitHub repository](https://github.com/UCLouvain-CBIO/scp/issues) in case of troubles when loading your SCP data with `readSCP`. Any suggestion or feature request about the function or the documentation are also warmly welcome. # Session information {-} ```{r setup2, include = FALSE} knitr::opts_chunk$set( collapse = TRUE, comment = "", crop = NULL ) ``` ```{r sessioninfo, echo=FALSE} sessionInfo() ``` # License This vignette is distributed under a [CC BY-SA license](https://creativecommons.org/licenses/by-sa/2.0/) license. # Reference