Differences Between Government, Consortium, and Private Database Stewards Impacting the Genomic Data Market: A Survey of U.S. Academic Genetic Researchers

Introduction

Given the costly and time-consuming nature of primary genomic data collection, the market for established databases for secondary analyses is a foundation of contemporary genetic research (Spector-Bagdady 2025). Most of these databases are held by three main types of data stewards: government (generally publicly funded repositories with government oversight and open or controlled access for researchers, e.g., UK Biobank, All of Us, dbGaP), consortia (generally collective repositories that involve multiple institutes across government and academia that have shared governance, e.g., 1000 Genomes Project, ENCODE), or private (generally repositories that are commercially owned by biotechnology or direct-to-consumer genetic testing companies, e.g., TTAM (formerly 23andMe), Ambry, Ancestry DNA) (Jaffe et al. 2024). Each of these types of data stewards have different ownership, access restrictions, funding sources, and priorities – some overlapping, some not – which can influence their accessibility to researchers, the kind of data management features they support, and requirements placed on researchers as a condition of use.

Government stewards in particular have invested time and effort in ensuring that data which are generated with federal funding are shared with publication or by the end of an award period (The National Institutes of Health 2020). For example, the National Institutes of Health’s (NIH) new data sharing policy aims to integrate “effective data management and sharing practices into the process of scientific discovery” itself (Jorgenson et al. 2021). The Director of the NIH also announced in 2021 that the NIH would revise the 2014 Genomic Data Sharing (GDS) Policy (National Institutes of Health 2014), particularly with regards to data harmonization, de-identification, and aggregation (Jorgenson et al. 2021).

The research impact of these different kinds of policies and requirements across data stewards is underexplored. While our recent qualitative work examined the factors driving researchers’ choice of genomic database (Trinidad et al. 2023; Spector-Bagdady et al. 2024), additional quantitative exploration is needed to assess the generalizability of these findings. It is also important to understand genomic researchers’ perspectives on the barriers and facilitators to data sharing, given the promised upcoming revisions to the GDS policy.

Here we present data from a broad survey of U.S.-based genomic academic researchers exploring the current landscape of genomic database usage, including how researchers are choosing between databases; priorities in selection; perceived quality; impact of database selection on research output and its intersection with federal funding; and feedback on the new NIH data sharing requirements such as barriers and facilitators to sharing data and challenges in use. Findings from the same dataset regarding whether researchers are interested in working with diverse ancestral groups, and how to encourage that use, are previously published (Jaffe et al. 2024). Findings offered in this article provide a much-needed overview of current genomic database usage and data sharing in the U.S., establishing a foundation for the future revisions to the NIH GDS policy.

Materials and methods

Results

Discussion

Our survey of 294 U.S. genetic researchers offers insights into current trends in genomic database usage as well as the factors driving the market for data stewards. We found that researchers generally have a choice in databases, reinforcing the importance of developing a deeper understanding of what motivates decisions about which database(s) to use and with which data stewards to work. It is notable that most respondents work with multiple types of data stewards, and that most report research with government and consortium databases. Researchers’ use of privately held data has not been increasing recently; the trend was toward use of consortium-held databases.

We also found that a lack of knowledge about different kinds of data stewards was the biggest barrier to use across all types. This suggests that data stewards could invest more in disseminating information to increase usage as opposed to relying on informal knowledge networks and mentorship. Given that mentorship is a barrier for researchers from underrepresented demographics in particular (Bonifacino et al. 2021; Ramanan et al. 2006; Nivet et al. 2008), formal investments in information sharing about different data stewards may also help enhance equity for genetic researchers. We also found that senior respondents had significantly higher odds of working with consortium data stewards, compared to more junior respondents, which may be an artifact of this knowledge gap. Future research should look more extensively into the ways in which mentorship and knowledge shape database usage across diverse research groups.

In a world of big data where the refrain that “bigger is better” is common, we found that it does not matter how big a database is if it is not efficient to access, does not have the right phenotypic information, or does not offer the right genomic sampling technique. As government data stewards focus on developing databases and attracting researchers, they must carefully address prioritized quality concerns and avoid assuming that merely collecting large amounts of data will guarantee database usage.

The low prioritization of seeking out historically underrepresented populations or the diversity of ancestral populations in databases relative to other features in selecting a database is consistent with our previous work in this area (Trinidad et al. 2023). Increasing data diversity alone is not enough to attract researchers, even if they wish to work with data from historically underrepresented ancestries (Jaffe et al. 2024). It must be paired with other motivating features and quality guarantees. Interestingly, while our sample of researchers were overwhelmingly positive about the ability of the new NIH data sharing requirements to increase the demographic diversity of available data, other recent research found that Black research participants were less likely to be willing to participate in research due to the change (Rivera et al. 2024). Therefore, the ultimate impact of the new data sharing requirements on data diversity will need to be carefully tracked.

Some of these trends toward use of government and consortium databases may be explained by different numbers of databases available, perceived differences in the quality of these databases, as well as the different requirements data stewards place on researchers. Notably, in terms of quality, consortium databases were rated more highly than private databases in each of the three domains we assessed: ease of access, database features, and management features. Government databases were also rated more highly than private ones in terms of ease of access.

Additionally, there were clear differences among the kind and number of requirements different data stewards place on respondents. Government data stewards had the fewest restrictions and private data stewards had the most. These two factors are interesting in contrast: respondents reported increasingly using consortium databases over the past 5 years in comparison to government and private databases, which may be related to the finding that consortium databases were rated more highly across access and data and management features. Past research has also found that, while some international genomic consortia are required to follow the NIH or other funders’ data sharing policies and others are voluntarily adopting standards set forth by the Global Alliance for Genomics and Health, there still exists a broad and opaque range of governance approaches (Budin-Ljøsne et al. 2014) and “consortia tend to adopt access arrangements that best suit the interests of members” (Nicol et al. 2024). Yet, government data stewards had the fewest publication restrictions. These findings indicate that upstream benefits (i.e., access, database features) are more impactful on researchers’ choice in databases than downstream publication limitations (e.g., co-authorship, limited data release). In contrast, minimizing or eliminating downstream publication and data sharing limitations are a strategic goal of government funders.

Our findings about the different requirements data stewards place on researchers are not only important for understanding researcher choices around databases but also the policy implications related to use of these different database types (Spector-Bagdady 2025). Crucially, private data stewards were especially likely to require the release of only an aggregated or limited dataset post-publication, likely because of the business value of those assets, which goes counter to current policy efforts that are moving toward open science (Ross et al. 2023). To nudge behavior toward data sharing, government actors may have to emphasize upstream, rather than downstream, requirements. For example, federal funders may have to prevent or limit the use of federal funding to access or analyze data from private stewards (creating an upstream access barrier), rather than the current practice of requiring funded researchers to share data back with the scientific community after publication or the grant end (a downstream requirement which might be contractually impossible by the time researchers have already signed intellectual property agreements with corporate data stewards). The fact that researchers are already trending away from use of private databases compared to government and consortia suggests that the imposition of these regulations might not stifle innovation but instead simply hasten that trend. It might also impact the private data market, acting as motivation for corporate stewards to change practice as databases which cannot be used for prestigious federally funded research grants, and data that cannot be published in prestigious journals, might decrease the value of those databases for resale purposes.

Additionally, private data stewards were more likely than government data stewards to require co-authorship. This is another area that requires policy attention given that co-authorship may be a way that data stewards control what kind of research is allowed to be published and discourage open science. Also, while private databases appear less stringent than government or consortia in terms of requiring citations, this is likely because they are already being credited in a way that has a higher market value for them through these authorship practices.

Consistent with our previous research (Spector-Bagdady et al. 2019), we again found that about 50% of research involving privately held genomic data were funded in part or whole by the NIH. This is consistent with the NIH’s support of other health research sectors. For example, the NIH financially contributes toward 99.4% of drugs approved by the Food and Drug Administration (Galkina Cleary et al. 2023).

Past work has established that researchers cite costs, lack of data repository infrastructure, questions related to data ownership, and time as barriers to data sharing (Jorgenson et al. 2021). They also report concerns about perceived career risks, including being “scooped” when forced to share unpublished data (Ali-Khan et al. 2017; Trinidad et al. 2023; Spector-Bagdady et al. 2024). Additionally, researchers have reported worries about mandated data sharing dissuading industry collaborations (Ali-Khan et al. 2017). Assessing these metrics more in depth, we found that time was seen as the most important barrier and that misuse of data was more likely to be a concern for junior, as compared to senior, researchers. This may be because junior researchers are more likely to be put in charge of procedural and compliance requirements of data sharing (Smart and Tan 2022) and/or have a lower risk-tolerance for error given the vulnerability of their career (Fischhoff et al. 2021). That said, funding agency policy has been found to influence data sharing behaviors (Kim and Stanton 2016), indicating opportunities for change. Researchers also cite career advancement and increased visibility as motivators for data sharing (Kim and Stanton 2016).

Strengths of this study include its novelty given the lack of previous research examining patterns in the genetic database market and researcher priorities around database selection. The exploratory sequential mixed methods design of building our survey instrument from previous qualitative work offered a rigorous way of approaching our research questions. This study also has limitations. First, given that it is challenging to establish the total number of researchers using human genetic data with an academic affiliation in the United States, the generalizability of our sample size is difficult to assess. That said, we recruited as broadly as possible, including through diverse published literature and the primary U.S. genetic research organization. Second, there may have been confusion between how we scoped our three data stewards of interest (government, private, and consortium). To address this, we eliminated partially completed surveys as this might have been an indication of confusion. Third, there was large variation in the number of respondents who had experience with government or consortium databases versus private databases, which complicates more nuanced statistical comparisons across these database types. Fourth, respondents took this 2022–23 survey based on existing experience, which might not include more recent government initiatives to support diverse genetic databases.

Conclusions

Genetic researchers make choices about what type of data stewards to work with based on several key priorities, particularly efficiency in access and availability of phenotypic information. In general, researchers are selecting government and consortium databases more frequently than private ones, which may reflect access challenges as well as the requirements that private data stewards place on researchers who use their data. Despite the increasing market trend toward the use of government and consortium databases in the last 5 years, knowledge of existing resources remains a barrier for some researchers, suggesting a need for more government investment in outreach and education to foster equity in the field. The federal government can leverage these findings about researcher priorities to continue attracting researchers, to push forward goals related to open science and enabling advances for patients underrepresented in genetic research, and to found new genomic data sharing policy moving forward.