--- title: "Quick introduction to plots created by our package" author: "Mateusz Nizwantowski" date: "`r Sys.Date()`" vignette: > %\VignetteIndexEntry{Quick introduction to plots created by our package} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} %\VignetteDepends{ggplot2} %\VignetteDepends{nplr} --- ```{r setup, include=FALSE} knitr::opts_chunk$set( collapse = FALSE, comment = "#>", warning = FALSE, message = FALSE, dpi = 50, out.width = "70%" ) ``` # Introduction The `PvSTATEM` package provides a variety of plots that can be used to visualize the Luminex data. In this vignette, we will show how to use them. To present the package's functionalities, we use a sample dataset from the Covid OISE study, which is pre-loaded into the package. Firstly, let us load the dataset as the `plate` object. ```{r} library(PvSTATEM) plate_filepath <- system.file("extdata", "CovidOISExPONTENT.csv", package = "PvSTATEM", mustWork = TRUE) # get the filepath of the csv dataset layout_filepath <- system.file("extdata", "CovidOISExPONTENT_layout.xlsx", package = "PvSTATEM", mustWork = TRUE) plate <- read_luminex_data(plate_filepath, layout_filepath) # read the data plate ``` # Plate layout We will omit some validation functionality in this vignette and focus on the plots. After successfully loading the plate, we should validate it by looking at some basic information using the summary function. However, we can obtain similar information more visually using the `plot_layout` function. It helps to quickly asses whether the layout of the plate is correctly read from Luminex or the layout file. The function takes the `plate` object as the argument. ```{r} plot_layout(plate) ``` The plot above shows the layout of the plate. The wells are coloured according to the type of sample. If the user is familiar with the colour scheme of this package, there is an option to turn off the legend. This can be done by setting the `show_legend` parameter to `FALSE`. If the plot window is resized, it is recommended that the function be rerun to adjust the scaling of the plot. Sometimes, the whole layout may be shifted when a legend is plotted. To solve this issue, one has to stretch the window toward the layout shift, and everything will be adjusted automatically. # Counts for a given analyte The `plot_counts` function allows us to visualize the counts of the analyte in the plate. This plot is useful for quickly spotting wells with a count that is too low to interpret results with high confidence. The function takes the `plate` object and the analyte name as the arguments. The function will return an error message if there is a typo in the analyte name. ```{r} plot_counts(plate, "Spike_B16172") ``` The plot above shows the the analyte "OC43_NP_NA" counts in the plate. The wells are coloured according to the count of the analyte. Too-low values are marked with red, values on the edge of the threshold are marked with yellow, and the rest are marked with green. There is an option to show legend by setting the `show_legend` parameter to `TRUE`. There is also an option to show the colours without the counts by setting the `show_counts` parameter to `FALSE`. This provides a cleaner plot without the counts. ```{r} plot_counts(plate, "FluA", plot_counts = FALSE) ``` # Distribution of MFI values The `plot_mfi_distribution` function allows us to visualize the distribution of the MFI values for test samples for the given analyte. And how they compare to standard curve samples on a given plate. This plot is helpful to asses if the standard curve samples cover the whole range of MFI of test samples. The function takes the `plate` object and the analyte name as the arguments. The function will return an error message if there is a typo in the analyte name. ```{r} plot_mfi_for_analyte(plate, "Spike_B16172") ``` This plot shows the distribution of the MFI values for test samples for the analyte "OC43_NP_NA". The test samples are coloured in blue, and the standard curve samples are coloured in red. The default plot type is violin, but there is an option to change it to the boxplot by setting the `plot_type` parameter to `boxplot`. ```{r} plot_mfi_for_analyte(plate, "FluA", plot_type = "boxplot") ``` Additionally, we can modify the scale of y-axis by setting the `scale_y` to the desired transformation from ggplot2 package. In case of `boxplot` type of plot, we may include the outliers by the `plot_outliers` parameter. ```{r} plot_mfi_for_analyte(plate, "FluA", plot_type = "boxplot", scale_y = "identity", plot_outliers = TRUE) ``` # Standard curve plots Finally, we arrive at the most crucial visualization in our package - the standard curve-related plots. Those plots help assess the quality of the fit, which will be crucial to us in the next step of package development. It comes in two flavors: `plot_standard_curve_analyte` and `plot_standard_curve_analyte_with_model`. The first does not incorporate the model, while the second does. ## Standard curve plot without model This plot should be used to assess the quality of the assay. If anything goes wrong during the plate preparation, it should be visible easily in this plot. ```{r} plot_standard_curve_analyte(plate, "Spike_B16172") ``` Above, we see the default plot for the analyte "Spike_B16172". We can modify this plot by setting the parameters of the function. For example, we can change the direction of the x-axis by setting `decreasing_rau_order` parameter to `FALSE`. Other parameters worth mentioning are `log_scale`, the default value is `c("all")`, which means that both the x and y axes are in the log scale. Other parameters worth mentioning are `log_scale`, the default value of which is `c("all")`, which means that both the x and y axes are on the log scale. There is also an option to turn off some parts of the plot by setting parameters `plot_line`, `plot_blank_mean` and `plot_rau_bounds` to `FALSE`. The first disables drawing the line between standard curve points, the second turns off plotting the mean of blank samples, and the last disables plotting the RAU value bounds. ## Standard curve plot with model This visualization is similar to the previous one but also incorporates the model. Thus, it carries more information at the cost of being more complex and crowded. ```{r} model <- create_standard_curve_model_analyte(plate, analyte_name = "Spike_B16172") plot_standard_curve_analyte_with_model(plate, model) ``` Here, we do not have to specify the analyte name, as the model already carries this information. The model is created by the `create_standard_curve_model_analyte` function, which takes the `plate` object and the analyte name as the arguments, but this is not the focus of this vignette. The arguments of this function are very similar to the previous one, except here there is a missing `plot_line` argument, and there are two new arguments: `plot_asymptote` and `plot_test_predictions`. The first turns off the asymptotes, and the second disables plotting the test samples' predictions. By default, both are set to `TRUE`.