The scenario of scientific production has undergone significant changes over the last 20 years with the introduction of the Open Science paradigm. This movement is guided by the understanding that scientific knowledge should be widely and openly shared for the benefit of all people (United Nations Educational, Scientific and Cultural Organization, 2022). Notably, scientific knowledge is considered a driving force for economic growth and for solving complex problems in an interconnected world (Silva; Silveira, 2019; United Nations Educational, Scientific and Cultural Organization, 2022). Hence, the more open and global its production is, the faster these problems can be solved.

The Open Science model is a contemporary approach that aims to democratize access to scientific knowledge, strengthen the transparency of research processes, and expand the social impact of science. Among its main advantages are free access to publications and data, which reduce economic barriers and promotes greater equity in the dissemination of knowledge (United Nations Educational, Scientific and Cultural Organization, 2021). Furthermore, interdisciplinary collaboration and information sharing foster innovation and scientific efficiency, avoiding duplication of efforts and optimizing public resources (European Commission, 2016; Fecher; Friesike, 2014). Methodological transparency, supported by the openness of protocols, codes, and databases, contributes to the reproducibility and reliability of results (Nosek et al., 2015). From a social perspective, the model encourages knowledge transfer and the practical application of discoveries, driving public policies and evidence-driven technological solutions (Vicente-Saez; Martinez-Fuentes, 2018). Finally, by favoring the participation of historically underrepresented countries and groups, Open Science promotes greater inclusion and cognitive justice in the global scientific system (Chan et al., 2020). Thus, it has consolidated itself as a paradigm that combines scientific rigor, social responsibility, and equity in access to knowledge.

Despite the undeniable gains of opening the production of scientific knowledge, Open Science faces a series of challenges, including the need to prepare scientists for a necessary change in behavior regarding their activity. Incentives and technical training for data sharing are on the agenda (Walters, 2020). Additionally, an adequate information infrastructure is required and involves practices from library science, archival science, and data management (Sayão; Sales, 2023). There is also a demand for technological resources, namely servers, hardware, and software. In this sense, open data repositories are foundational elements of the Open Science ecosystem. Pampel et al. (2023) counted more than 3,000 data repositories registered in re3data, signaling a growing increase in the ecosystem that meets re3data’s quality and openness requirements. Moreover, along with data repositories, data journals have been listed as possible sources of open data and are seen as a response to the need for recognition of data authors (Kindling; Strecker, 2022; Walters, 2020).

The study was conceived as a quantitative-qualitative study, based on the context of scientific production on data repositories and open science indexed in the Web of Science (WoS) database. The Methodi Ordinatio, proposed by Pagani, Kovaleski and Resende (2015) was applied and adapted to achieve the proposed objectives. This method proposes a strategy for building a portfolio of relevant studies to answer the research question, following a systematized review proposal. It adopts a multi-criteria model for evaluating and ranking publications (InOrdinatio) based on the impact factor, year of publication, and number of citations to determine the relevance of studies to the field (Pagani; Kovaleski; Resende, 2015).

Methodi Ordinatio is composed of nine phases, as follows: Phase 1) − Establishment of research intention; Phase 2) − Preliminary exploratory research with keywords in the databases; Phase 3) - Definition and combination of keywords and databases; Phase 4) − Final search in databases; Phase 5) − Filtering procedures; Phase 6) − Identification of the impact factor, year of publication and number of citations; Phase 7) − Ranking of articles using InOrdinatio; Phase 8) - Search for full articles; Phase 9) − Final reading and systematic analysis of articles.

Next, the way in which the nine phases were applied in this study is described.

Phase 1) − Establishment of research intention: the initial research theme was defined as the requirements for building a data repository.

Phase 2) − Preliminary exploratory research with keywords in databases: the Web of Science (WoS) database was chosen because it provides the largest number of studies on the topic in an exploratory and comparative search with Scopus. Also, it was chosen because it provides one of the metrics for ranking studies, the impact factor, in addition to the number of citations. At this stage, a preliminary, broader search on the topic was carried out using the following search strategy:

The search was carried out on November 8th 2023. Documents in press (with early access) were removed from the search strategy itself because the publication date was not established.

Phase 3) − Definition and combination of keywords and databases: Based on the analysis of indexed keywords and authors’ keywords, a thematic focus was defined, considering the frequency of keywords and thematic growth trends. Therefore, it was decided to address the association between the theme “Open Science” and its derivations with data repositories. The question to be answered was: “What is the impact of the Open Science model on the construction of data repositories?”

Phase 4) − Final search in databases: The final search strategy was structured based on words representing ‘data repositories’ and words that represented the context of Open Science, according to the taxonomy defined in Pontika et al. (2015). Finally, early access studies were removed, as were conference papers, data articles, and editorial material. Therefore, only published original articles and review articles made up the portfolio.

Phase 5) − Filtering procedures: the complete WoS records database was downloaded in Excel format (.xlsx) and studies were selected considering whether the title represented the topic under study. A single researcher selected the studies, and, in case of doubt, the researcher kept the study among those eligible.

Phase 6) − Identification of the impact factor, year of publication and number of citations: the metric adopted was the WoS Impact Factor (IF). When the journal did not present an IF, the Journal Citation Indicator (JCI) − also provided in WoS − was considered. The year of publication and the number of citations across all databases were also collected from the database.

Phase 7) − Ranking of articles using InOrdinatio: the selected articles were ranked using the following formula:

Where FI is the impact factor, α is a weighting factor for the weight of the year of publication, which can vary from 1 to 10; the year of research is the year in which the study is being carried out, the year of publication is the year in which the article was published, and ΣCi is the sum of citations. Regarding the weighting factor of the year of publication, it is considered that topics in which more recent studies are more relevant can adopt weighting factors closer to or equal to 10, while topics in which the date of publication is irrelevant or which tends to consider older studies as more relevant, brings the weighting factor closer to 1. In this study, the weighting factor adopted was 10, considering the current relevance of the studies in the context analyzed.

Phase 8) − Search for full articles: the 15 studies with the highest InOrdinato index were included in the study.

Phase 9) − Final reading and systematic analysis of the articles: at this stage, the reading, interpretation, and synthesis of the texts were conducted following the content analysis method proposed by Bardin (2020). First, an exploratory reading of the studies was carried out. Then, excerpts (units of analysis) relevant to answering the research question were selected and represented by thematic categories and subcategories.

The results of the final search (Phase 4) indicated a total of 545 documents. Filtering these documents by title led to the exclusion of 358 records, resulting in 152 remaining documents. The InOrdinatio index was applied to all 152 documents and Table 1 presents the 15 studies with the highest InOrdinatio index included in the study.

Ten out of the 15 studies included (66.7%) described open data repositories, eight of which (53.3%) were from the biological field. The United States was the country with the highest number of studies (n=5), followed by Japan and China (n=2). Eight of the repositories belonged to the biological field, one of chemistry and one of health. The other five studies addressed data reuse (Piwowar; Vision, 2013), sharing data on ecology (Michener, 2015), merging tables of open data from different sources (Nargesian et al., 2018), reuse of qualitative data (Bishop; Kuula-Luumi, 2017) and data selection and classification algorithms (Gad et al., 2022).

The content analysis of the 15 studies resulted in 57 units of analysis representing four distinct categories: “open access, information technologies, data processing, and information retrieval” (Figure 3).

Open data repositories enable open access to data, promoting transparency and accessibility. Twelve units of analysis highlighted this relationship:

Furthermore, repositories provide information technologies such as systems, metadata and the data itself. Twelve units of analysis referred to systems including relational data management systems written in open source, graphical and programmatic interfaces (data sets, filters and attributes), databases and data analysis systems written in different open codes and machine learning algorithms aimed at combining data:

Another four units of analysis referred to metadata standards, as follows:

Besides, another nineteen units of analysis identified the data itself as a constituent of information technologies. These units present elements related to data analysis, such as open-source technologies and programs, data management tools, incentives and barriers to data sharing, citations and metrics related to data:

The data processing category was represented by four units of analysis, and addressed data generation and data management, including its curation. The unit below expresses this context:

Finally, six units of analysis expressed the information retrieval category. The category involved the concepts search, semantic web and ontologies, all provided by open repositories to provide accessibility to research data:

The quantitative-qualitative study method was based on Pagani, Kovaleski, and Resende’s (2015) Methodi Ordinatio, adapted to answer the proposed question. The adaptations involved the application of the bibliometric method to analyze the scenario of scientific production regarding data repositories in Phases 2 and 4. The Methodi Ordinatio generated a portfolio of studies, among which the 15 best ranked were analyzed. The analysis of the studies attributed to data repositories the collaboration with the openness of scientific data and its transparency (Bhattacharya et al., 2018; Kearnes et al., 2021; Zheng et al., 2021), as proposed by United Nations Educational, Scientific and Cultural Organization (2020). Furthermore, our results indicated repositories proposing strategies for processing data, helping to organizing them for sharing and reuse since their generation by researchers, in addition to their curation and management (Bhattacharya et al., 2018; Michener, 2015; Zheng et al., 2021). In addition, repositories provide the necessary information technology structure, such as systems, metadata and the data itself (Bhattacharya et al., 2018; Guo et al., 2020; Smedley et al., 2015; Zheng et al., 2021), as well as the structure for information retrieval that involves ontologies, semantic web, and the search processes themselves (Bhattacharya et al., 2018; Gad et al., 2022; Moriya et al., 2019; Okuda et al., 2017). Most studies were not necessarily related to the area of information science but rather described active data repositories in the biological, chemical, and health areas.

The method adopted in the study, involving InOrdinatio to highlight the most relevant studies, considered the criteria involved in calculating the index (Pagani; Kovaleski; Resende, 2015). The decision to adopt a weighting factor (α) equal to 10 together with the year of publication was based on the understanding of the relevance of the recent aspect of the study, in contrast to the total number of citations, known to be lower in more recent studies. On the other hand, the adoption of InOrdinatio may have presented studies that did not necessarily associate open data repositories with information science, which may have generated a selection bias due to the impact factor of the journals.

The studies analyzed revealed the commitment of data repositories to open access to data, in line with what is recommended in the UNESCO Recommendations (United Nations Educational, Scientific and Cultural Organization, 2021). The Recommendations advocate for “open access to scientific publications, research data, metadata, open educational resources, software, source codes and hardware available in the public domain or under copyright and open licenses”. In this sense, in addition to the openly expressed commitment to open access in some studies (Bhattacharya et al., 2018; Kearnes et al., 2021; Zheng et al., 2021), others observed the construction of repository systems in open source (Bhattacharya et al., 2018; Smedley et al., 2015); and the use of open data analysis resources (Sud et al., 2016; Zheng et al., 2021).

Importantly, the concept “system”, in the context of information retrieval presupposes − beyond the hard aspect of software or technology − “the set of processes, human skills, techniques and technologies, sources of information united in a systemic way with the purpose of effectively providing Information Retrieval” (Lancaster, 1979). Thus, it includes all processes and technologies required to guarantee the findability of information and its use or reuse. Regarding data repositories, the FAIR principles (Findability, Accessibility, Interoperability and Reusability) guide this particular information system, including metadata standards (United Nations Educational, Scientific and Cultural Organization, 2021; Wilkinson et al., 2016). Based on this, models of good metadata practices for data repositories have focused on components of dataset description in order to align with the semantic web proposal and the linked data perspective, enhancing the automatic synthesis of data and its reuse (Greenberg et al., 2009). Interestingly, none of the studies analyzed mentioned FAIR principles. Despite this, some of the studies that described the repositories mentioned the use of systems built on open-source software (Sud et al., 2016), metadata standards based on machine-readable XML and RDF schemas (Okuda et al., 2017) and open-source data analysis tools (Kearnes et al., 2021).

Based on the proposed method, and considering the limitations of the study, it was possible to conclude that the most impactful production has occurred in the biological, biomedical or interdisciplinary fields and highlights the structure involved in existing repositories. Data repositories provide systems for depositing, managing, searching, accessing and reusing data based on processes and technologies, most of which are developed using open-source software, in alignment with the proposed Open Science model.