ShinySperm: navigating the sperm proteome landscape

Application development and user interface

The ShinySperm application was built using the shiny package (ver. 1.8.1) on RStudio (ver. 2023.12.1), with base R (ver. 4.3.3). Supporting the functionality and aesthetics of this application are several packages, including: DT, ploty, RColorBrewer, dplyr, ggplot2, and shinythemes. The user interface is defined using Shiny’s fluidPage function, which allows for a responsive layout.

The title panel includes a custom logo. The navigation of the application is facilitated by tabsetPanel allowing the addition of multiple interlinked tabs, including ‘Epididymal Sperm Proteome’, ‘Volcano Plot’, ‘Phenotypes’, and ‘Functions’. Each subpanel contains interactive elements such as search boxes, dropdown menus for filtering, and checkboxes to enable specific selections based on gene symbols, protein types, and sperm cell-specific locations.

Dataset source

The epididymal sperm proteomic datasets uploaded to the ShinySperm app, alongside their in silico components, were sourced from Skerrett-Byrne et al. (2022). These table/files were imported into the ShinySperm application as comma-separated value (CSV) files (Supplementary Tables S1–S3).

Data handling, visualisation, and export

ShinySperm integrates several R packages for data handling, visualisation, and export functionality. The DT package provides interactive tables, which are utilised to display searchable and sortable data tables. The dplyr package is used for data interrogation, including filtering and sorting datasets based on user input. For static and interactive visualisations of these datasets, ggplot2 and plotly are employed, respectively. These software packages enable the generation of pie charts, volcano plots, and heatmaps, all with the embedded capacity to display additional information as the user hovers a mouse over the relevant element. User interactions, such as text inputs for gene symbol searches and selections from dropdown menus, trigger reactive expressions defined in the server logic. These expressions dynamically update the data displayed in both the data tables and the visual plots, reflecting user-specified filters and searches.

The ShinySperm application offers data export functionality, allowing users to download filtered datasets and specific visualisations (Table 1). This is achieved using Shiny’s downloadHandler function and the package openxlsx, which together dynamically generate downloadable Microsoft Excel files based on the current state of data filtering, along with the capacity to add a descriptor sheet to the Excel file.

Main panel: interactive proteomic dataset

Three files are required to achieve the sum of this application: the original proteomic dataset (Table S1), the associated phenotypes of each protein (Table S2), and the Ingenuity Pathway Analysis (IPA) reproductive outputs (Table S3). Each file can be modified easily at the user’s discretion, but in this case, we focused on alignment with key aspects of the original publication (Skerrett-Byrne et al. 2022). In the main Panel of the application, the user is presented with the interaction table, a ‘live’ version of the proteomic dataset (Table S1), which contains unique identifiers (UniProt Accessions), protein names, gene symbols, protein types, known sperm location, and the comparative quantitative data of the maturing sperm cells: caput (immature) and cauda (mature) (Fig. 2a). Each column has a search feature, whilst the comparison data of fold change and P-value have a sliding scale indicative of the available range. Simultaneously, the dataset is graphed as two pie charts below the table, one for ‘protein type’ and another for ‘sperm cell location’. Both pie charts harbour interactive qualities, such that hovering a mouse over a pie slice reveals both the number and percentage of proteins mapping to that protein type or sperm cell location. The supporting legend facilitates the ability to deselect any groups not of interest (single click), or to focus on a group of interest (double click).

Fig. 2.

Illustrative example of a user enquiry. (a) Filtered interactive table for sperm acrosome proteins leading to reactive data visualisation of (b) volcano plot, (c) associated phenotypes, and (d) reproductive functions. The golden boxes highlight the points of interest discussed in user example section.

Supporting the table and pie charts is a panel on the left side, providing a search and filter feature. Importantly, any searches or filters performed are reflected instantly in both the table and pie charts. The user has the capacity to search specific gene symbols (e.g. RhoA), filter by protein type (e.g. kinases), and sperm location (e.g. acrosome). In accordance with the findings of the original publication, the user can filter for the full complement of proteins identified in populations of either caput or cauda epididymal sperm, as well as the core sperm proteins retained during maturation. Additionally, the panel provides the capacity to rest all filtering, export the filtered or unfiltered data (Table 1), and importantly, a feedback mechanism facilitating contact with the author of the application.

This panel is the core hub supporting the subsequent tabs. Any searches or filters in the left-hand side panel are promptly adjusted across each of the subsequent tabs.

Side panels: reactive data visualisation

Extending the data presented on the main panel are three side tabs, which are composed of ‘Volcano Plot’, ‘Phenotypes’, and ‘Functions’ (Fig. 2b–d). As indicated by the name, the ‘Volcano Plot’ graphs the comparative data from the interactive table (Fig. 2b). Without filtering, the ggplot package renders the full comparative proteome as an interactive volcano plot. Complemented with aesthetical features, the code provides the capacity to indicate thresholds of significance (FC ± 1.5, P-value ≤ 0.05), colouration to groups of proteins of interest, and visual guides on the graph (arrows and text). Hovering a mouse over each dot point provides the user with the associate gene symbol of the corresponding protein. Additionally, the graph can be exported as a PNG file (Table 1). Integral to the interactive aspect of the application, the volcano plot will react in live time to any filters or searches imposed on the main panel.

Table S2 provides the information that serves as the basis for the phenotypes tab, which links back to the interactive table via a common column, UniProt Accession. The application renders a pie chart of all mapped phenotypes to the identified epididymal sperm proteins, with the same features as the main panel (i.e. hovering a mouse over pie slices) (Fig. 2c). The full repertoire of phenotypes can be exported as an Excel file (Table 1), and likewise, is reactive to any filters imposed on the main panel. The exported file integrates the interactive table with any mapping phenotypes, whereby each sperm protein with its associated phenotype(s) is provided as individual rows. Furthermore, the pie chart can be exported as a PNG file.

The final subpanel leverages the data contained in Table S3, again unified by the common column of UniProt Accession. This IPA file also contains function annotation (reproductive focused), enrichment score of function (−log₁₀P-value), and the predictive activation score (z-score) (Krämer et al. 2013). The application renders these data as a heatmap, displaying the top 20 functions, reactive to filtering and searches performed, with the colour gradient indicative of the enrichment value of that function (Fig. 2d). As with the previous panels, this figure can be exported as either a PNG or Excel file (Table 1). Likewise, the exported Excel file integrates the interactive table with any mapped reproductive function and its associated enrichment and z-score where applicable.

User example

To illustrate the capacity of ShinySperm, Fig. 2 sets out an example case of a user interested in epididymal sperm proteins known to be located in the acrosome. By filtering for acrosome via the side panel under ‘Sperm location’, the interactive table returns 69 proteins and adjusts the pie charts below reflecting the protein type and other known sperm cell locations of those proteins (Fig. 2a). This refined proteome can be extracted via the export button, which reveals six proteins that are apparently lost from the sperm proteome during epididymal maturation and a further four proteins that are gained by the maturing sperm cells. The remaining 59 proteins can be further explored under the ‘Volcano Plot’ tab, which has been adjusted to reflect that filtered dataset (Fig. 2b). All proteins which meet those significant thresholds are coloured by their respective decrease or increase coincident with epididymal maturation. The user can hover a mouse over each protein dot to obtain the gene symbol as highlighted in the gold box, revealing milk fat globule-EGF factor 8 (MFGE8) protein ligand, an important mediator of macromolecular cargo exchange between extracellular vesicles, known as epididymosomes, and maturing epididymal spermatozoa (Nixon et al. 2019b; Trigg et al. 2021).

Progressing to the next subpanel, ‘Phenotypes’, the user is presented with a modified pie chart of all phenotypes mapping to the 69 sperm acrosome proteins, chief amongst them being impaired acrosome reaction (Fig. 2c). Hovering a mouse over the red slice informs the user that only one protein matches to this phenotype. To provide a more detailed summary of phenotypes, a modified Excel file can be exported that includes all of the phenotypic relationships linked to the subset of targeted sperm proteins. The final subpanel renders all significant reproductive functions related to those acrosome sperm proteins as a heatmap (Fig. 2d). Notably, in this example the ‘Functions’ panel identified crucial functions pertaining to acrosome reaction, including ‘capacitation’, ‘fertilisation’, and ‘binding’ of sperm. Finally, the user once more can export these data as an Excel file, providing a stratified approach to their biological exploration of the data.