Using Glittr.org to find, compare and re-use online materials for training and education

12 A wealth of excellent training materials on computational life sciences is scattered around the 13 internet, but they can be hard to find. Many materials reside in public git repositories that are 14 hosted on platforms such as GitHub and GitLab. Glittr.org aggregates and indexes such git 15 repositories enabling users to find training materials that would otherwise be hard to identify. 16 With the application, users can search and compare training materials based on topic and 17 author, but also on engagement metrics like stargazers (bookmarks) and recency (days since 18 last commit). Glittr.org currently contains 568 entries, which are assigned to six different 19 categories within the field of computational life sciences. By analysing the database, we reveal 20 insights in the availability of materials per topic, collaboration patterns of developers, and 21 licensing practices. This knowledge helps to understand in which areas open training materials 22 are lacking, the importance of git for collaboration on training materials and how licensing can 23 be improved to enhance sharing and reuse. Taken together, we show that Glittr.org contains a 24 wealth of connected and openly available metadata. Therefore, it enhances adherence to the 25 FAIR (Findable, Accessible, Interoperable, Reusable) principles, which benefits the entire life 26 sciences community. 27

28 The life sciences domain is constantly growing and evolving, requiring professionals to be 29 trained to stay up to date with the most relevant skills and knowledge. To facilitate this, various 30 organizations deliver courses on a large range of topics in many different formats. Many course 31 providers dedicate themselves to actively share their materials with the public, resulting in a 32 growing wealth of public training materials available across the web. This greatly enhances the 33 possibility to re-use these training materials, which is invaluable for both learners and lesson 34 developers; learners can use the material on their own time to obtain required skills, while 35 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 3 lesson developers can use the existing material to build their courses further upon existing 36 ones. 37 To help developers make their shared material easier to find and re-use, the FAIR principles [1–38 5] provide guidelines on organizing their materials and metadata. The way training material 39 metadata are represented depends on the platform where they are hosted. Currently, training 40

are hosted in diverse locations, like organizational websites, cloud storage, learning 41 management systems, and Git platforms (e.g., GitHub, GitLab). Therefore, metadata 42 representation often does not conform to standards, making it difficult to parse and utilize it 43 effectively. This issue can be relieved by aggregation and indexation, which encourages 44 metadata standardization and therefore enhances findability, re-usability, and integration of 45 metadata over multiple platforms. 46 Registries that currently aggregate and index training materials in the life sciences include the 47 Open Educational Resources (OER) Commons [6], the training portal of the Global Organisation 48 for Bioinformatics Learning, Education and Training (GOBLET) [7], and ELIXIR’s online training 49 registry Training eSupport System (TeSS) [8]. These registries can accommodate any digital 50 object related to education that are accessible through an URL, making them potentially 51 comprehensive. However, this flexibility also comes with disadvantages. Since not every 52 webpage contains standardized metadata, it requires investments from the training material 53 developer to manually enter metadata at the registry or to automate metadata provision, for 54 example by implementing a standardized metadata markup like Bioschemas [9,10]. 55 Using public git repositories for developing and hosting training materials is gaining popularity, 56 especially for materials related to (bio)informatics and data science. It is generally 57 recommended for reproducibility [11] and adopted by many institutions, communities, and 58 individuals (e.g. the Carpentries [12], Galaxy training network [13], the National Bioinformatics 59 Infrastructure Sweden (NBIS) and the SIB Swiss Institute of Bioinformatics. Having training 60 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 4

in public Git repositories has major advantages. First, using Git encourages 61 community-driven development. Git is version-control software, and as such, can be used to 62 track changes in files and associate these changes to individuals with distinct levels of 63 permissions. This makes it ideal for co-development and re-use of materials. Using git drives 64 community engagement by supporting direct feedback, suggestions, and corrections. 65 Secondly, the infrastructure around the most popular platform for public git repositories, 66 GitHub (https://github.com) and GitLab (https://about.gitlab.com), are standardized and 67 commonly used. Many people have the know-how to use version control, so the threshold to 68 use or learn how to use git for training material is low – especially in communities related to the 69 informatics field. Thirdly, public Git repositories offer standardized retrieval of a wealth of 70 metadata. Authorship, activity logs, and popularity metrics are all recorded and made 71 accessible through a single point of entry. 72 Despite many advantages, it remains difficult to find training materials on public git 73 repositories. Although both GitHub and GitLab have sophisticated search functionality, it 74 remains difficult to identify and compare repositories with training materials on a topic of 75 interest. The findability of repositories with training materials depends on the effort of the 76 author to describe the repository, for example in the repository description, the name, the 77 associated website, and repository tags. Since not all authors annotate their repositories 78 extensively and the way developers annotate repositories is not standardized, many 79 repositories with excellent training materials remain hard to find. 80 This article introduces Glittr.org, a novel web-based resource designed to enhance the 81 discovery and comparison of reusable bioinformatics training materials hosted on public Git 82 repositories. Addressing the needs of both students and trainers, Glittr.org contains intuitive 83 search functionalities based on pre-defined topics and repository metadata. This approach not 84 only facilitates the users’ search based on specific topics but also the refinement of their 85 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 5 search based on factors like the training material's popularity and recency. Moreover, Glittr.org 86 supports the adherence to the FAIR principles by making all repository metadata readily 87 accessible through REST API endpoints, fostering findability and reusability of bioinformatics 88 training materials within the wider research community. 89

90 Data collection and curation 91 Glittr.org is a manually curated list of bioinformatics training materials from anywhere in the 92 world and in any language. The list items are public GitHub or GitLab repositories. The teaching 93 aim of the material should be a general topic in bioinformatics, not tool usage – meaning that 94 inclusion of tutorials on using single software are discouraged. The reason for this is to ensure 95 Glittr.org provides an overview of metadata on training materials, and not on metadata related 96 to software. The quality of the training materials is not evaluated. Repositories that are 97 potentially eligible to be part of Glittr.org are identified from existing collections, internet 98 searches, specific searches on GitHub, (social media) promotions and user contributions 99 through a web form. 100 Repository metadata are retrieved from GitHub and GitLab. Among those are the repository 101 name, URL, repository description, author and contributor names, their profile and web site, 102 the number of stargazers (bookmarks), the date of the last push, and the license. These are all 103 programmatically accessible from the REST APIs of GitHub and GitLab. 104 To enable categorization and improve search functionality in Glittr.org, the training content in 105 the repositories are described with tags. These tags are based on existing ontologies as much 106 as possible. Using ontologies greatly enhances standardization and therefore knowledge 107 sharing. Training material in bioinformatics can cover a very wide range of topics, and using a 108 single ontology to describe them can be problematic. Glittr.org therefore uses multiple 109 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 6 ontologies, aiming to give an as rich as possible topic description for each repository. Currently, 110 the following ontologies are used: EDAM [14], Data Science Education Ontology 111 (http://www.pagestudy.org/DSEO/), Experimental Factor Ontology [15] and FAIR Data Train 112 Ontology (https://bioportal.bioontology.org/ontologies/FDT-O). If a term is considered relevant, 113 but not in any of these ontologies, a term will be used from any other ontology on BioPortal [16], 114 a software on bio.tools [17] or an item at Wikidata [18]. By using multiple ontologies, we have 115 the flexibility to annotate the wide range of training material, but still have good level of 116 standardization in relation to the relevant field. 117 Each repository is associated with one or more tags by a curator, where the order of tags 118 matters, meaning that the tag that describes the content best is first (the ‘main tag’) and is then 119 followed by other relevant tags all with equal importance. For example, for a single cell 120 transcriptomics course in R, the tag order could be ‘Single-cell sequencing’, ‘R’, 121 ‘transcriptomics’ and ‘RNA-seq’. The first and therefore main tag would be ‘Single-cell 122 sequencing’, which is used to categorize the material, while the others have equal importance 123 and are used to further describe the material. Each tag falls within a broader category, which 124 are maintained within Glittr.org. Currently, these are: (1) Scripting and languages, (2) 125 Computational methods and pipelines, (3) Omics analysis, (4) Reproducibility and data 126 management, (5) Statistics and machine learning, and (6) Others. The main tag of the above-127 mentioned single cell course is part of the category ‘Omics analysis’ and so the course material 128 falls within this category. 129 Web application 130 The Glittr.org web application is created using the Laravel (https://laravel.com/) framework and 131 is written in PHP. Repository metadata is actively updated twice daily using Laravel jobs. This 132 approach enables handling of potential issues by implementing delayed retries if the 133 GitHub/GitLab APIs are temporarily unavailable. Each metadata update for a single repository 134 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 7 constitutes a dedicated Laravel job, with the overall management of all repository metadata 135 updates performed with the Laravel Haystack package (https://docs.laravel-haystack.dev/). 136 This package ensures job validation and sends email notifications in case of errors. 137 All the reactivity supporting list functionalities such as filtering, pagination, and sorting is based on 138 the Livewire (https://livewire.laravel.com/) framework. The JSON API endpoints for repositories and 139 Bioschemas are generated with the Laravel-query-builder (https://spatie.be/docs/laravel-query-140 builder/v5/introduction) package developed by Spatie (https://spatie.be/open-source). This ensures 141 the query parameters and output results follow the JSON API specification as closely as possible. The 142 backend to manage data (topics and categories, repositories, submissions) is created with the 143 package Jetstream (https://jetstream.laravel.com/introduction.html). 144 All source code is accessible at https://github.com/sib-swiss/glittr and the website is available 145 through https://glittr.org. The application is a SIB Training resource and is hosted at servers 146 from SIB. As an institute with a long history of working with database life cycles, we are 147 committed to sustaining glittr.org for the time it is serving the community by aligning to the 148 TRUST principles [19]. 149 A snapshot of the front-end website is found in Fig 1. The user can easily access both the 150 repository with course material and the course web site (when applicable). In addition, a link to 151 public information about the author or organization that created the material is provided to 152 retrieve more author information. The topics (tags) covered by the material allows users to 153 quickly find course material relevant to their search. The number of stargazers estimates the 154 popularity of the repository, while the number of days since the last push provides an indication 155 of the recency of the material in the repository. The license which applies to the repository 156 contents indicates to trainers wishing to re-use the material what conditions apply. 157 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 8 158 Fig 1. Snapshot of Glittr.org. Here, two tags are selected ‘Transcriptomics’ and ‘R’, showing the (current) 159 top 2 of 25 repositories with both tags. By default, results are sorted based on number of ‘stargazers’, i.e. 160 bookmarks. 161 The application currently has three REST API endpoints in which data stored by Glittr.org can be 162 programmatically accessed: api/repositories, api/tags and api/bioschemas. The 163 api/repositories endpoint contains all repositories including their metadata as presented on the 164 website. The repositories/tags endpoint contains tag metadata organized by their 165 corresponding category. Tag metadata consists of the ontology, ontology class and description. 166 The api/bioschemas endpoint contains repository and tag metadata according to the 167 Bioschemas [9] TrainingMaterial profile [10] (https://bioschemas.org/profiles/TrainingMaterial). 168 Training material characteristics 169 To give an overview and statistics of all the metadata stored on Glittr.org, we used the R httr2 170 package to access the Glittr.org REST API (see above). Public information on authors, and 171 number of contributors was extracted from the GitHub REST API. Location of authors and 172 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 9 organization was extracted from the ‘location’ field within the users API. This string was 173 translated in a country with the R package geocode, which has wrappers for the Google 174 Geocoding API. The source code for acquiring the statistics can be found at 175 https://github.com/sib-swiss/glittr-stats and the report can be approached through https://sib-176 swiss.github.io/glittr-stats/ (updated weekly). 177

and Discussion 178 There are currently 568 repositories listed on Glittr.org. By combining data from our REST APIs 179 and those of GitHub and GitLab we can give a detailed insight in these repositories. In the first 180 part of this section, we show what kind of repositories with training material are currently 181 available, for which topics training materials exist, what licenses are used, and by whom the 182 repositories are created. In the second part we present several use-cases on how different 183 types of users can interact with Glittr.org to show how trainers, learners and developers can 184 use different functionalities of glittr.org. 185 Training material topics 186 Repositories on glittr.org are categorized in six categories and tagged with 55 different topics. 187 All are connected to bioinformatics, but many are not specific to bioinformatics: for example, 188 computer languages are relevant to informatics in general, and statistics to all fields related to 189 science. Therefore, using glittr.org is relevant to a broader public than just persons working in 190 the field of bioinformatics. In this section we give a brief overview of popular categories and 191 topics. 192 By far most repositories currently on Glittr.org have ‘Scripting and languages’ as main category 193 – 315 in total (Fig 2A). Within this category, R [20] is the most frequent tag with 230 repositories, 194 followed by Python with 81 repositories (Fig 2C). This suggests that the R language is currently 195 the most popular language to teach and therefore probably to learn in bioinformatics, with 196 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 10 Python coming in second. The popularity of R might not be the only reason; the ease with which 197 one can host static websites within the R framework, with e.g. bookdown [21] or Quarto 198 (https://quarto.org/), could also explain this. 199 200 Fig 2. Key statistics by category. Overview of number of repositories by category (A), distribution of 201 number of contributors per repository by category (B), number of repositories per tag, square root 202 transformed (C), per author or organization (D) and per license (E). Colours are according to category as 203 depicted in (A). In C) only the tags with more than 10 repositories are displayed and in D) only the 204 authors/organizations with five or more repositories. At time of writing Glittr.org contained repositories 205 Others Statistics and machine learning Reproducibility and data management Computational methods and pipelines Omics analysis Scripting and languages 0 100 200 300 Number of repositories CategoriesA Others Statistics and machine learning Reproducibility and data management Computational methods and pipelines Omics analysis Scripting and languages 0 100 200 300 400 500 Number of contributors ContributorsB Total number of tags: 55 ChIP−seq Julia Nextflow Docker Metagenomics Shiny Workflows Containerization Version control Variant analysis Data visualization FAIR data General Reproducibility Data management Single−cell sequencing Unix/Linux Genomics Machine learning Data science Next generation sequencing Statistics RNA−seq Transcriptomics Python R 0 50 100 150 200 Number of repositories Tags with > 10 repositoriesC Authors with < 5 repos: 242 cambiotraining carpentries−lab hadley biocorecrg JuliaAcademy semacu swcarpentry RockefellerUniversity learnbyexample rstudio−conf−2022 bioinformatics−ca GTPB bioinformaticsdotca bioinformatics−core−shared−training datacarpentry hbctraining ucdavis−bioinformatics−training NBISweden posit−conf−2023 sib−swiss carpentries−incubator 0 10 20 30 40 Number of repositories Author or organizationD unlicense wtfpl agpl−3.0 artistic−2.0 bsd−3−clause apache−2.0 cc0−1.0 cc−by−4.0 gpl−3.0 cc−by−sa−4.0 mit none other 0 50 100 150 200 Number of repositories License typeE Repos with undefined country: 244 Argentina China Finland India Luxembourg Norway Poland Ukraine Bulgaria Ireland Italy Denmark Spain Belgium France Netherlands Portugal Germany Australia United Kingdom Sweden Canada Switzerland United States 0 50 100 Number of repositories CountryF .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 11 authored by 242 authors/organizations with less than five repositories. For A), D) and E) the plots are 206 coloured according to the category of the main (i.e. first) tag. 207 The category ‘Omics analysis’ is specific to the field of bioinformatics, and many repositories 208 fall within this category. ‘Transcriptomics’ is the most frequent tag (74 repositories), followed by 209 ‘RNA-seq’, suggesting an important place for transcriptomics-related courses in current 210 bioinformatics curricula. Transcriptomics analysis skills are popular to acquire, likely because 211 it is one of the most used applications of next generation sequencing, it is applied in many fields 212 in biology [22] and each experiment requires its own analysis choices. Its demand is well served 213 by the vast amount of material available on transcriptomics. 214 There are 34 repositories on ‘Data management’, representing its growing importance within 215 the scientific community regarding trends in open and FAIR data. Using public git repositories to 216 develop training materials on data management related topics, is likely to be of particular 217 interest because using git supports transparency and community-based development. Efforts 218 that support open and reproducible science like the ELIXIR Research Data Management (RDM) 219 community, European Open Science Cloud (EOSC) and other research infrastructures likely 220 influence this trend towards using git repositories for training materials both on data 221 management and other topics. 222 The number of repositories and their diversity show that there is a wealth of public repositories 223 with bioinformatics training material. For many topics, there are high-quality materials in 224 different formats, giving lesson developers and learners many possibilities to re-use existing 225 training materials. Glittr.org organizes this wealth of materials, making them better findable and 226 therefore available for re-use. 227 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 12 Collaboration 228 An important advantage of using version control for training materials is that it supports 229 collaboration and building further on existing material. This is also represented in the number of 230 collaborators per repository that are on Glittr.org. Most repositories (78.8%) have more than 231 one contributor, meaning the repository is developed by more people than the original author. 232 Many repositories have only a few contributors, between 1 and 5 (59.9%). However, quite a 233 large fraction (25.6%) has more than 10 contributors, showing that these repositories gain from 234 a strong community involvement and therefore increased (re-)usage and quality. 235 Repositories in the category ‘Reproducibility and data management’ and ‘Computational 236

and pipelines’ have the largest fraction of repositories developed by a large 237 community (> 50 contributors; Fig 2B). Repositories with a large contributor base within these 238 categories are e.g. elixir-europe/rdmkit - which is a community-created website with RDM best 239 practices [23], galaxyproject/training-material – the Galaxy training network repository with all 240 training materials related to Galaxy [13] - or are forked from a repository with a large community 241 base (e.g. a template from the carpentries [12]). This shows that large and important initiatives 242 rely on git for the development of training materials and that these thrive by cooperation. 243 Authors and organizations 244 The origin of repositories on Glittr.org is very diverse; currently, the 568 repositories originate 245 from 263 different GitHub/GitLab authors or organizations. The organization with most 246 repositories is the carpentries-incubator [12], followed by two national institutes, NBIS and SIB 247 (Fig 2C). Obviously, the distribution of repository category varies per organization/author. 248 Bioinformatics institutes, like bioinformatics.ca, Instituto Gulbenkian de Ciência (GTPB), CRUK 249 CI (bioinformatics-core-shared-training), Harvard bioinformatics core (hbctraining), UC Davis 250 Bioinformatics Core (ucdavis-bioinformatics-training), SIB and NBIS, are more focused on 251 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 13 computational biology and have more courses in the category ‘Omics analysis’. Other popular 252 namespaces like carpentries incubator [12], the ’23 posit conference (posit-conf-2023) and 253 learn by example (learnbyexample) are focused on computer skills and therefore have mostly 254 repositories on ‘Scripting and languages’ and ‘Computational methods and pipelines’ (Fig 2C). 255 In addition to the number of authors, the number of countries the authors affiliate themselves 256 with is also diverse. Current repositories on Glittr.org originate from 24 different countries. Most 257 repositories with the main author or organization having a country defined in its profile originate 258 from the United States, followed by Switzerland and Canada (Fig 2F). This shows that most 259 training materials originate from the Western and particularly English-speaking parts of the 260 world. English is currently the ‘lingua franca’ of the academic world [24], and most training 261

in bioinformatics reflect that. Connections between the academic ecosystem and 262 training materials not written in English can be literally lost in translation, and therefore hard to 263 identify. However, there are many communities that demand non-English bioinformatics 264 training material, and we are open to serving these communities by listing them on Glittr.org. 265 Organization of the curation of repositories in other languages is more challenging because it 266 relies on reading and understanding the content, but we encourage initiatives in this direction. 267 Licensing 268 Although licensing is an important way to enable and encourage re-use, licensing of public 269 training materials on GitHub and GitLab is not very standardized. Fig 2B gives an overview of 270 licenses per repository that could be recognized as such. For many repositories (172; 30.4%), 271 there is no license detected, and for 211 (37.3%) repositories the license could not be 272 automatically recognized. The most popular recognized license is the MIT license, a software 273 license. The copyright licenses based on Creative Commons (CC) are also quite popular, 75 274 repositories have a license recognized as such. Since training materials generally consists of 275 text, presentations, and exercises, and not software, it is usually more appropriate to use a 276 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 14 copyright license instead of a software license [4,11]. It therefore seems that many developers 277 of training materials might want to (re-)consider their license. 278 Use cases 279 Glittr.org has been developed for use by bioinformatics students, trainers, lesson developers 280 and database developers. To showcase different applications of the resource, we present five 281 different use cases that we have envisaged for Glittr.org. The first four use cases illustrate the 282 ease of use, as only a single click is necessary to obtain an answer, while the last one shows 283 how to retrieve information programmatically. 284 A student looking for training materials 285 Our first use case is that of a student looking for training materials, for instance someone 286 wishing to learn the programming language Julia. This high-level, general-purpose dynamic 287 programming language used for numerical analysis and computational science is usually 288 offered in university and higher order education curricula. Using the ‘Filter by topic’ option to 289 select Julia under the ‘Scripting and Languages’ section, only repositories which have been 290 manually assigned the topic are retrieved. The top result is the most popular as, by default, the 291

are displayed in order of decreasing number of stargazers. It is also possible to sort by 292 increasing number of days since the last push to find material which has been recently 293 updated. The link to the website takes the student directly to the page showing the different 294 options to get started with Julia. Glittr.org thus allows students to quickly access the most 295 popular, public course material on the subject. 296 A trainer looking for material for re-use 297 This use case is that of a trainer who will be teaching a course on Snakemake and is looking for 298

that can be re-used. Snakemake is a workflow management system to create 299 reproducible and scalable data analyses. Selecting Snakemake in the Computational methods 300 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 15 and pipelines section of the ‘Filter by topic’ option retrieves several repositories with materials 301 to teach Snakemake. By looking at the license column, the trainer can easily spot that there are 302 two repositories that contain material that have a re-use permissive license, one under the 303 Creative Commons Attribution-ShareAlike 4.0 International license (CC BY-SA 4.0) and another 304 under the GNU General Public License version 3 license (pl-3.0). For other repositories in the 305 list, the license could not be automatically identified, and therefore the license has been 306 indicated as ‘other’. For these repositories some further investigation is needed before they can 307 be re-used. The trainer has quickly found re-usable material on the topic of interest using the 308 Glittr.org web site. 309 A trainer sharing their own materials 310 The third use case is that of a trainer who has developed a course on a hot topic in 311 bioinformatics on GitHub or GitLab and wishes to share the material. Contributions to the 312 Glittr.org web site from the training community are welcome and encouraged. The Contribute 313 button on the top right of the home page (Fig 1) links to a simple form. By completing the 314 submission form, the trainer's repository will be added (or updated) to the collection provided it 315 meets the inclusion criteria (public GitHub or GitLab repositories containing bioinformatics 316 training material). The trainer’s material is now findable by being included in the Glittr.org web 317 site. Obviously, this does not prevent the trainer from posting the material in other repositories 318 which is, in fact, recommended by the FAIR principles (the more the material is registered, 319 more chances are that they are findable). 320 A database developer wanting to programmatically access Glittr.org 321 A final use case is that of a developer who must programmatically access one or more subsets 322 of training materials repositories on Glittr.org. This can easily be done using the API endpoints 323 (https://github.com/sib-swiss/glittr#api) and applying suitable filters. For example, the data for 324 the graphs in Fig 2 are all extracted from the Glittr.org API with the R packages httr2 [25]. The 325 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 16 Bioschemas endpoint is a special API endpoint in Glittr.org. It provides access to the metadata 326 of all the repositories in Glittr.org in Bioschemas TrainingMaterial markup. This allows portals 327 that can ingest Bioschemas markup to re-use Glittr.org data. For example, ELIXIR TeSS [8], 328 currently ingests this metadata (https://tess.elixir-europe.org/content_providers/glittr-org). 329

330 When we created Glittr.org, our aim was to create a simple-to-use, web-based listing of 331 bioinformatics training materials available on GitHub/GitLab because as training providers we 332 saw the need for such a tool. Within one year, more than 500 repositories with bioinformatics 333 training materials with a diverse set of topics have been included in Glittr.org. By focussing only 334 on git repositories, Glittr.org can automatically display relevant information like popularity, 335 license and recency that would otherwise not be possible. The resource has the potential to 336 help in training material selection, decision-making by both students and trainers and training 337

database integration, as illustrated by the five use cases presented. All the information 338 is available programmatically via the API and Bioschemas markup thereby enabling other 339 training registries to freely update their existing entries or add entries not yet present in their 340 database. This will further increase the coverage of these registries and promote the findability, 341 accessibility, interoperability, and re-usability (FAIRness) of training materials in the life 342 sciences. 343

344 1. Filiposka S, Green D, Mishev A, Kjorveziroski V, Corleto A, Napolitano E, et al. Draft 345 Methodology for FAIR-by-Design Learning Materials. 2023 [cited 11 Apr 2024]. Available: 346 https://zenodo.org/records/7875541 347 2. Haase R, Tischer C, Bankhead P, Miura K, Cimini B. A Call for FAIR and Open-Access 348 Training Materials to advance Bioimage Analysis. Open Science Framework; 2024 Mar. 349 doi:10.31219/osf.io/2zgmc 350 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 17 3. Alexandra Holinski, Alexia Cardona, Alice Matimba, Allegra Via, Anastasios Anastasios 351 Papaioannou, Anna Swan, et al. ELIXIR fair training handbook. 2024 [cited 22 Mar 2024]. 352 doi:10.5281/ZENODO.10839551 353 4. Garcia L, Batut B, Burke ML, Kuzak M, Psomopoulos F, Arcila R, et al. Ten simple rules for 354 making training materials FAIR. PLoS Computational Biology. 2020;16: 1–9. 355 doi:10.1371/journal.pcbi.1007854 356 5. Wilkinson MD, Dumontier M, Aalbersberg IjJ, Appleton G, Axton M, Baak A, et al. The FAIR 357 Guiding Principles for scientific data management and stewardship. Sci Data. 2016;3: 358 160018. doi:10.1038/sdata.2016.18 359 6. Sparks SD. Open Educational Resources (OER): Overview and Definition. Education Week. 360 12 Apr 2017. Available: https://www.edweek.org/teaching-learning/open-educational-361 resources-oer-overview-and-definition/2017/04. Accessed 5 Feb 2024. 362 7. Corpas M, Jimenez RC, Bongcam-Rudloff E, Budd A, Brazas MD, Fernandes PL, et al. The 363 GOBLET training portal: a global repository of bioinformatics training materials, courses 364 and trainers. Bioinformatics. 2015;31: 140–142. doi:10.1093/bioinformatics/btu601 365 8. Beard N, Bacall F, Nenadic A, Thurston M, Goble CA, Sansone S-A, et al. TeSS: a platform 366 for discovering life-science training opportunities. Bioinformatics. 2020;36: 3290–3291. 367 doi:10.1093/bioinformatics/btaa047 368 9. Gray AJG, Goble C, Jimenez RC. Bioschemas: From Potato Salad to Protein Annotation. 369 ISWC 2017 Posters & Demonstrations and Industry Tracks: Proceedings of the ISWC 2017 370 Posters & Demonstrations and Industry Tracks co-located with 16th International 371 Semantic Web Conference (ISWC 2017). RWTH Aachen University; 2017. Available: 372 https://research.manchester.ac.uk/en/publications/bioschemas-from-potato-salad-to-373 protein-annotation 374 10. Castro LJ, Palagi PM, Beard N, Attwood TK, Brazas MD. Bioschemas training profiles: A set 375 of specifications for standardizing training information to facilitate the discovery of 376 training programs and resources. PLOS Computational Biology. 2023;19: e1011120. 377 doi:10.1371/journal.pcbi.1011120 378 11. Dogucu M, Çetinkaya-Rundel M. Tools and Recommendations for Reproducible Teaching. 379 Journal of Statistics and Data Science Education. 2022;30: 251–260. 380 doi:10.1080/26939169.2022.2138645 381 12. Wilson G. Software Carpentry: lessons learned. F1000Research; 2016. 382 doi:10.12688/f1000research.3-62.v2 383 13. Hiltemann S, Rasche H, Gladman S, Hotz H-R, Larivière D, Blankenberg D, et al. Galaxy 384 Training: A powerful framework for teaching! PLOS Computational Biology. 2023;19: 385 e1010752. doi:10.1371/journal.pcbi.1010752 386 14. Ison J, Kalaš M, Jonassen I, Bolser D, Uludag M, McWilliam H, et al. EDAM: an ontology of 387 bioinformatics operations, types of data and identifiers, topics and formats. 388 Bioinformatics. 2013;29: 1325–1332. doi:10.1093/bioinformatics/btt113 389 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint 18 15. Malone J, Holloway E, Adamusiak T, Kapushesky M, Zheng J, Kolesnikov N, et al. Modeling 390 sample variables with an Experimental Factor Ontology. Bioinformatics. 2010;26: 1112–391 1118. doi:10.1093/bioinformatics/btq099 392 16. Noy NF, Shah NH, Whetzel PL, Dai B, Dorf M, Griffith N, et al. BioPortal: ontologies and 393 integrated data resources at the click of a mouse. Nucleic Acids Res. 2009;37: W170–394 W173. doi:10.1093/nar/gkp440 395 17. Ison J, Rapacki K, Ménager H, Kalaš M, Rydza E, Chmura P, et al. Tools and data services 396 registry: a community effort to document bioinformatics resources. Nucleic Acids Res. 397 2016;44: D38–D47. doi:10.1093/nar/gkv1116 398 18. Wikidata. [cited 9 Apr 2024]. Available: https://www.wikidata.org/ 399 19. Lin D, Crabtree J, Dillo I, Downs RR, Edmunds R, Giaretta D, et al. The TRUST Principles for 400 digital repositories. Sci Data. 2020;7: 144. doi:10.1038/s41597-020-0486-7 401 20. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R 402 Foundation for Statistical Computing; 2023. Available: https://www.R-project.org/ 403 21. Xie Y. bookdown: Authoring Books and Technical Documents with R Markdown. Boca 404 Raton, Florida: Chapman and Hall/CRC; 2016. Available: 405 https://bookdown.org/yihui/bookdown 406 22. Stark R, Grzelak M, Hadfield J. RNA sequencing: the teenage years. Nat Rev Genet. 407 2019;20: 631–656. doi:10.1038/s41576-019-0150-2 408 23. ELIXIR. Research Data Management Kit. A deliverable from the EU-funded ELIXIR-409 CONVERGE project (grant agreement 871075). Available: https://rdmkit.elixir-europe.org 410 24. Jenkins J. English as a Lingua Franca in the International University: The Politics of 411 Academic English Language Policy. Routledge; 2013. 412 25. Wickham H. httr2: Perform HTTP Requests and Process the Responses. 2023. Available: 413 https://httr2.r-lib.org 414 415 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 22, 2024. ; https://doi.org/10.1101/2024.08.20.608021doi: bioRxiv preprint