Last March, US government-funded National Center for Biotechnology Information (NCBI) websites went dark for about a day, leaving researchers all over the world unable to access information in these repositories. Gaurav Sharma, a microbial genomics and data analytics researcher at the Indian Institute of Technology Hyderabad, realized the repercussions of this blackout.
“NCBI is [the] bread and butter for all the people in the whole world,” said Sharma. Researchers, healthcare professionals, and students across the globe heavily rely on NCBI’s PubMed database, which indexes nearly 40 million articles in biomedical literature. The NCBI also houses several other databases in genomics and biomedical fields.
Gaurav Sharma, a microbial genomics and data analytics researcher at the Indian Institute of Technology Hyderabad, and a graduate student in his lab, Utkarsha Mahanta, authored a commentary proposing a framework to decentralize databases.
Gaurav Sharma
“Modern biology runs on databases,” said Sharma. To him, the outage of NCBI databases highlighted the vulnerability of relying on one agency to manage such massive scale resources. Some of his colleagues who use the databases for their work shared his concerns.
Recently, Sharma and his colleagues called for decentralizing databases such that data can be stored and managed across a network of independent agencies without relying on a central authority.1 In their commentary, published in Nature Genetics, they proposed a framework to ensure that scientists all over the world can access databases, encouraging open science practices to sustain the progress of global research.
“[This] is the right direction,” said Friederike Kleinfercher, a computer scientist and Head of the Max Planck Decentralized Science Initiative, who was not associated with the commentary.
“[The commentary] does a very good job of highlighting issues which are incredibly important for all of us,” said Katy Wolstencroft, a bioinformatician at Amsterdam University Medical Center, who was not involved in writing the commentary. “It also does a good job of explaining that this is quite a complex issue, and it requires really long-term thinking,” she added.
The Different Systems of Biological Data Sharing
While centralized systems—such as the NCBI databases—ensure uniform standards, they also entrust decision-making authority within a single institution. This can render such structures vulnerable by creating one single point of technical or administrative failure.
By contrast, ELIXIR, the European life sciences infrastructure for biological information, operates under a federated model as a network of 25 independently run organizations, or nodes, instead of a single centralized authority.2 At each node, experts autonomously maintain their specialized databases, while sharing metadata and accessing policies through agreed-upon standards. This ensures distributed risk.
Decentralized systems constitute a third model wherein they distribute storage and governance across many independent nodes, reducing the risk of data censorship. But this could have its own problems.
“If it's completely decentralized, then some initiatives will fail over the longer term because their funding will run out,” said Wolstencroft. “So, there needs to be enough redundancy in the network that others will take over when things do fail.”
Friederike Kleinfercher is the Head of the Max Planck Decentralized Science Initiative.
Picture credit https://patrickart.es/
Based on this principle, Sharma and his colleagues proposed a framework that integrates federated and decentralized models of data sharing and management. This would connect independent national hubs—each of which retain full ownership and control over their own data—into one network that follows the same standards.
But by sharing both governance and funding across the network, this system builds a safety net in that disruption of one hub would not lead to loss of knowledge since data would remain safely accessible through the rest of the network.
Practical Challenges to Decentralized Databases and the Way Forward
Despite its importance, Sharma noted that decentralization is not easy to attain. “You need resources, you need funding, you need manpower,” he said. “The most important challenge is an arrangement of thought process and financial concerns.”
There are also logistic challenges to decentralized databases, according to Wolstencroft. To maintain a good level of service, scientists must be well-versed with technologies for data sharing, and have institutional support. “To make this work properly, you need shared standards and protocols,” Wolstencroft added. “If you don't have that, then sharing between nodes in the network is not very easy.”
Kleinfercher has also faced another difficulty with getting some scientists to share their data. “They say, ‘I'm not sure if I [will] get credit for it. Someone can take it and say this was his or her idea,’” said Kleinfercher.
Blockchain technology that tracks and timestamps each step to ensure a clear record could help circumvent this problem. But Kleinfercher has sometimes found it difficult to convince researchers to add their data on the blockchain-based project she runs.
“What I found very useful [in the commentary, was]…this federated idea to take already existing sources and to integrate that in a decentralized architecture,” said Kleinfercher. “That's a very smart way to go.”
However, she believes additional direct practical examples of decentralized science initiatives could have helped in the commentary. In her experience, this method seems abstract to people and having examples helps them understand how to participate in decentralized science themselves. Going forward, she hopes to see stakeholders increasingly adopt decentralized practices.
Sharma said this proposal is a first step and they still have to iron out some details. “The discussion has started,” he said. “That was the whole point behind this [commentary].”
- Sharma G, et al. Towards a decentralized future for open-science databases. Nat Genet. 2026.
- Drysdale R, et al. The ELIXIR core data resources: Fundamental infrastructure for the life sciences. Bioinformatics. 2020;36(8):2636-2642.
