A tutorial hub for realising the potential of tools within the SPARC ecosystem

Events and goals

This publication is prepared as the outcome of the 2022 SPARC (Stimulating Peripheral Activity to Relieve Conditions) FAIR (Findable, Accessible, Interoperable, and Reusable Codeathon, held 6-8 August. The first edition of this event was held in July 2021. The results of the groups participating in the July 2021 SPARC FAIR Codeathon are collected and accessible on this resource page. In the 2022 SPARC FAIR Codeathon, five groups participated, and more information can be found here.

The goal of the SPARC FAIR Codeathons is to find novel ways of using the SPARC resources, data, or tools to demonstrate and enhance the FAIRness of the data.

Introduction

The SPARC program is a platform supported by the National Institutes of Health (NIH) Common Fund to advance bioelectronic medicine devices.¹ This is done by identifying the spatial distribution of neurons on the surface of organs using flatmaps.² The SPARC Portal hosts over a hundred datasets and resources (continuously being updated) to assist researchers in learning about organ functions, devise medical tools, and ultimately treat diseases.

The SPARC Portal emphasizes on a FAIR (Findable, Accessible, Interoperable and Reusable) repository of curated resources available to the researchers globally. The FAIR Data Principles were proposed in 2016 by Wilkinson et al. in order to increase the reusability of data.³ The SPARC project incorporates multiple open-source tools and platforms with their own documentation and tutorials. However, well-documented tutorials on how to properly utilize these tools in conjunction with SPARC data are lacking and could significantly aid users in seeing the advantages of using FAIR data.

In this Codeathon we aimed at practically improving the reusability of the SPARC Portal resources by giving visual and written instructions on how the available datasets and models can be employed, integrated, and visualized. One of the core platforms of Data Resource Center (DRC) of the SPARC program is ${\mathrm{o}}^2{\mathrm{S}}^2\text{PARC}$ (Open Online Simulations for Stimulating Peripheral Activity to Relieve Conditions).⁴ Researchers can, for example, use ${\mathrm{o}}^2{\mathrm{S}}^2\text{PARC}$ for simulating the peripheral nerve system and instantly observe its impact on pertaining organs. However, the utilization of ${\mathrm{o}}^2{\mathrm{S}}^2\text{PARC}$ is not necessarily straightforward for general users. Therefore, to increase the FAIRness of the SPARC Portal resources, we aimed to show how one can search and use the SPARC datasets using available SPARC tools, in conjunction with other open-access tools, in a structured and comprehensible manner. We lower the barrier for the general user who might be unfamiliar with the SPARC integrated platforms (like ${\mathrm{O}}^2{\mathrm{S}}^2\text{PARC}$) to acquire data, process it, and visualize it using common, open-source scientific computing tools.

The purpose of this article is to present the tutorials that were developed during the 2022 SPARC FAIR Codeathon on the procedure of utilizing the SPARC portal tools and resources. These tutorials demonstrated the ease with which a range of tools in the SPARC ecosystem could be used in concert to access, manipulate and interpret SPARC datasets. As a case study, we provide an example of generating a 3D scaffold of the rat stomach and projecting the coordination data of the vagal afferent and efferent. We discuss how this tutorial differs from the existing tutorials and documentation on SPARC Portal. Finally, we describe how our tutorial style can impact the FAIRness of the SPARC Portal and possible future developments.

Methods

This section discusses the improvement of tutorials and documentation on the utilization of the SPARC Portal tools along with the available open-access software packages such as SciPy. Figure 1 demonstrates the general workflow for the tutorials. The first step is to obtain SPARC datasets from the portal using Pennsieve. This free tool is a scalable, graph-based, and cloud-based data management platform that is used for managing scientific datasets on the SPARC Portal. It is used here to download.xlsx files. Once the files are downloaded from a project, the data can be manipulated as needed in order to perform computational analysis on it by employing a programming language such as Python or open-source tools from the SPARC Portal. After the processing step, the data can be visualised in 2D or 3D graphs. In the following, we explain these steps and the cases demonstrated in the tutorial.

Figure 1. The general workflow of utilizing the available datasets on the SPARC Portal.

The.xlsx data files are downloaded using Pennsieve and fed to a program written in Python which uses open-source tools, like those available from SPARC, for preprocessing and analysing the data. After being processed, the data can be visualised in different manners.

Data acquisition

Figure 2 illustrates the above-mentioned workflow using a previously published model of the rat stomach.⁵ We illustrate two applications of the acquired datasets and give detailed instructions on how they can be processed for different applications. Three datasets constitute the basis of these tutorials. They provide the 2D coordinates for different neuronal populations in the rat stomach, namely the efferent vagal neurons and afferent vagal neuron sub-populations of intraganglionic laminar endings (IGLEs) and intramuscular arrays (IMAs). These data are derived from biomedical imaging of stained rat stomach sections. The information regarding these neurons is contained in.xlsx files that are accessible on the SPARC portal from the following links (see also Underlying data):

• • Spatial distribution and morphometric characterization of vagal afferents associated with the myenteric plexus of the rat stomach: IGLE_data.xlsx

• • Spatial distribution and morphometric characterization of vagal efferents associated with the myenteric plexus of the rat stomach: Efferent_data.xlsx

• • Spatial distribution and morphometric characterization of vagal afferents (intramuscular arrays (IMAs)) within the longitudinal and circular muscle layers of the rat stomach: IMA_analyzed_data.xlsx

Figure 2. The workflow of utilizing the available datasets on the SPARC Portal.

The upper panel (the general approach): The.xlsx data files are downloaded from Pennsieve and fed to Python for preprocessing (data extraction and cleaning). Using the available Scaffold Mapping Tools on the SPARC Portal, a geometric 3D scaffold of an organ is obtained. The preprocessed data can be projected on the 3D scaffold to create a data-enriched 3D image of the organ. The lower panel (the rat stomach example): Three different datasets of the spatial distribution of neurons in the rat stomach are downloaded from Pennsieve. The required 2D data (spatial coordinates of x and y) are extracted using Python. A 3D scaffold of the rat stomach is achieved using the SPARC Scaffold Mapping Tools. The preprocessed data of the spatial 2D distribution of neurons is projected on the rat stomach scaffold. The output is a 3D image of the rat stomach in which the neurons are mapped on the surface. The dots correspond to the spatial distribution of neurons and the three colors represent the three different datasets.

In this tutorial, we demonstrate a dataset acquisition mechanism that includes searching for datasets using keywords, e.g., ‘vagal,’ using the search API provided by Pennsieve with output datasets’ id, version, name, and tags. Using the search results, we can now download datasets manually or automatically, where the automatic method is facilitated by a script utilizing the discovery API that can be accessed freely. This mechanism is then helpful in searching for other datasets, which will later be used for different analysis purposes.

Visualization: Mapping 2D data to a 3D organ scaffold

Moreover, one can project the 2D points representing the location of neurons to a scaffold of the organ generated by the Scaffold Mapping Tools of SPARC as demonstrated in the lower panel of Figure 2. The scaffold organ obtained is a 3D mesh converted as an STL file to handle it using a Python-based package easily. Figure 3 is the representation of the 2D dataset that we map to the 3D organ scaffold. 3D mapping can be done because the datasets contain information on the anatomical location of neurons, dorsal and ventral.

Figure 3. A 2D map representing the location of neurons in the stomach.

The color difference indicates the type of neuron, and the color intensity indicates the area of the neuron (retrieved from [1]).

The step-by-step instructions on how to use the Python code are available on GitHub (see Software availability).

Results

In this section, we demonstrate the graphical results from the two tutorials that were developed during the Codeathon.

Figure 4 illustrates the results of the first tutorial, data resampling for density estimation simulations using KDE. The simulation includes all three datasets: vagal efferents, IGLEs, and IMAs. Within the plots, the green dot signifies the resampled point regarding vagal, while the backdrop color denotes the density of the data points. Areas in dark red indicate high density, while lighter ones represent lower density. It can be seen here that the highest concentration of vagal efferents is at 50% and above the x-axis approaching the forestomach. As for IGLEs, the distribution is relatively even, with the highest density between 10% and 50% of the x-axis. Meanwhile, IMA is concentrated in the antrum and scattered in the proximal stomach. These plots confirm the distribution of the stomach’s vagal afferent and efferent populations.⁵ They could be an alternative to visualize the distribution more effectively and efficiently. Here, we demonstrate that the datasets and tools provided by SPARC can be easily used and combined with other open-source tools. This tutorial module does not consider dorsal and ventral positions, naturally allowing for more precise visualization. We provide this space for learners as a learning process.

Figure 4. A probability density estimation of neurons on the surface of the stomach.

(a) efferent (b) IGLE (c) IMA. Statistical processes are used to compute a probability density estimation of neurons. This density function can be used to generate and simulate neuronal activity in a network.

Figure 5 illustrates the 3D mapping of the spatial distribution data to the rat stomach scaffold. The datasets for this module are the same as for the previous module, using data features on the ventral and dorsal position as well as the area of innervation. With this visualization, we can see the extent of innervation and later compare it with a 2D density plot. This area is tiny; the magnitude plot is inefficient; therefore, it can be represented well using a colour scale from white to dark green, where dark green represents the most significant area. We used the matplotlib widget module for⁸ interactivity so that the 3D plot could be rotated to observe the distribution of the vagal and area innervation in more detail.

Figure 5. The projection of the spatial locations of neurites on the surface of the rat stomach onto a 3D scaffold of the organ.

The result of our tutorial is the visualization of the neurons in an interactive figure in Jupyter Notebook. The user can select which type of neurons to visualise on the surface of the stomach from three categories: vagal afferent (IGLE), vagal afferent (IMA), and vagal efferent.

Conclusion

Our tutorials differ from the current tutorials on the SPARC Portal in a substantial way because they guide the users through understanding and using multiple SPARC tools to fulfil data manipulation and visualisation needs. Instead of the dense documentation and tutorials on the SPARC portal which mostly targets a single tool and is appropriate for users already familiar with the SPARC program, we have developed tutorials, along with examples, for a broader audience. We demonstrated the feasibility of using SPARC resources to drive our proposed workflow using a rat stomach scaffold as an example. A central benefit of this approach is that it leverages the FAIR principles that underlie these SPARC resources to ensure that the integrated models that are produced also follow the FAIR Data Principles. Our approach is creative because it encapsulates the workflow within a Jupyter Notebook online platform, which is designed to be readily accessible to new users of SPARC. Having tutorials that can be easily understood will not only accelerate the growth of the SPARC community but also increase the uptake and usage of tools by members of the community.