Lessons learned from the legacy of bioinformatic resources at SIB Swiss Institute of Bioinformatics

Lessons learned from the legacy of bioinformatic resources at SIB Swiss Institute of Bioinformatics Monique Zahn-Zabal1, Sebastien Moretti2, Philipp Bucher2 and Severine Duvaud1* 1Biodata Resources Group, SIB Swiss Institute of Bioinformatics, Lausanne, Switzerland. 2Vital-IT Group, SIB Swiss Institute of Bioinformatics, Lausanne, Switzerland. *Corresponding author. Abstract The life cycle of a bioinformatics resource has been described as consisting of the proof-ofconcept phase, the emerging phase, the mature phase, and the legacy phase. At a time when the FAIR principles (Findable, Accessible, Interoperable, Reusable) are being applied to bioinformatic resources, questions around reusability and archiving of resource components— especially ensuring they are interoperable— are unavoidable. Since 2000, the Swiss State Secretariat for Education, Research and Innovation (SERI) has mandated and funded the SIB Swiss Institute of Bioinformatics to identify, support, and develop key open bioinformatics resources. SIB portfolio comprises both emerging and mature resources. When a resource ends—due to insufficient funding, retirement or move of the Principal Investigator outside of Switzerland —SIB ensures that the resources remain available if still relevant to the scientific community. Here, we present use cases that provide insight into current and past practices, as well as lessons learned. Introduction Bioinformatics is a field in which biological data is managed, analysed, and visualized using computational tools and techniques. In its infancy, bioinformatic data dealt with DNA sequences, protein structures, and gene expression data; however, present day bioinformatics involves omics data, such as genomics, proteomics, transcriptomics, and metabolomics, and the use of machine learning or artificial intelligence to deal with the large volume of biological data. In this context, bioinformatics resources include databases, knowledgebases, data sets, software tools, workflows or pipelines, machine learning models. Bioinformatic resources thus play a crucial role in providing the scientific and medical community. Given the considerable time, expertise, and financial investment required for their development and maintenance, extending their lifespan has become a priority. Archiving a resource allows the scientific community to continue to benefit from it, rather than duplicating efforts, thereby optimizing funding and extending the knowhow found in these resources. The life cycle of a bioinformatics resource consists of four phases (Gabella et al., 2022). The first phase is the proof-of-concept phase, in which a research project develops a resource for research purposes. Further development on the resource takes place in the emerging phase. These first two phases form the research stage of a resource. With time, the resource matures and, if successful, may be considered as essential infrastructure. If the decision is taken to archive the resource, it enters the legacy state. The international life science community has been using the freely available databases provided by SIB since its foundation in 1998. These, as well as other resources, are listed in Expasy, the Swiss bioinformatics resource portal (Duvaud et al., 2021). The resources found in Expasy consisted of emerging and mature resources until recently and include five Global Core Biodata Resources (GCBR) considered valuable infrastructure. While guidelines exist for the development (Schultheiss, 2011; Helmy et al., 2016) and management of bioinformatics resources (Gabella et al., 2022), little attention has been given to the transition of an emerging or mature state resource to a legacy state, termed "sunsetting". Factors such as loss of funding, key personnel, users, or technological changes can lead to such a transition and the resulting disappearance of the resource is not an uncommon occurrence. A study by Atwood et al. (2015) showed that the 18-year survival rate of 326 bioinformatics databases was only around 25%, and there is no evidence to suggest an improvement in the last decade (Imker et al., 2023). The sudden disappearance of a bioinformatics database could have significant impact, leading to the loss of valuable research data, especially if the database held curated information essential for various research projects. Bioinformatics tools are similarly vital for analysing biological data. The disappearance of a widely used tool can disrupt ongoing research projects that rely on it for data analysis, leading to decreased productivity and efficiency. Researchers may spend valuable time searching for alternative tools or adapting their workflows to compensate for the missing tool. In the absence of guidelines for sunsetting resources and conscious of the impact this may have on the scientific community, we explored the legacy of bioinformatic resources developed when the SIB was in its infancy. Faced with resources developed by SIB groups whose heads are now retiring, we had to set up a process to minimize the impact this has on the long-term use of the resource. In this paper, we compare the life cycle of three resources developed at SIB - the CleanEx database, the EPD web resource, and the neXtProt knowledgebase – and contrast their legacy. The CleanEx Database The CleanEx (Praz et al., 2004) database of heterogeneous gene expression data was developed by Viviane Praz while doing her thesis in the group of Philipp Bucher at SIB. Further development took place during Viviane's post-doc (Praz and Bucher, 2009) but work on the resource stopped when she left the group, a frequent scenario. At the time, it was rare for a resource to be archived so, although the CleanEx flat files and tools were publicly available online, this is no longer the case. In the case of CleanEx, the resource went from the emerging phase to the legacy state because of the loss of key personnel. The legacy of the project consists of three publications (including Viviane's thesis), of which only two are currently accessible. The metadata for the resource is in two registries, Database Commons (CNCB-NGDC Members and Partners, 2022) and bio.tools (Ison et al., 2019). However, the entries do not show the legacy state of the resource. In terms of FAIRness, the CleanEx data and software tools themselves are no longer FAIR; however, the metadata for CleanEx is. What was the impact on the scientific community? Concomitant to the development of CleanEx, the community had access to the GEO (Edgar et al., 2022) and ArrayExpress (Parkinson et al., 2007) repositories. Today it has access to a number of mature stage resources of which two, Bgee, a database users can use to retrieve and compare gene expression patterns in multiple animal species (Bastian et al., 2025) and Gene Expression Database (GXD), a resource of mouse developmental gene expression information (Baldarelli et al., 2021), are GCBR. The EPD Web Resource The Eukaryotic Promoter Database (EPD) was developed by Philipp Bucher and his group at SIB. Originally a manually curated compilation of promoter sequences which primarily focusing on transcription start sites (TSS), a new section was added called EPDnew (Dreos et al., 2013). At time of writing this web resource which catalogues experimentally determined RNA polymerase II promoters in eukaryotes is available online at https://epd.expasy.org/epd/ and a total of nine publications describing the resource are listed in PubMed, the first in 1997 and most recent in 2020. Philipp's retirement, the loss of funding to further develop EPD, and the unsuccessful search for a successor to continue to develop the resource or another resource to integrate the data, meant that this mature resource needed to be prepared for the legacy state. As the resource is still online, widely used and no equivalent resource exists, the decision was taken in 2022 to archive all valuable parts. To determine which elements of the resource should be archived, our decision was guided by what could be of use to the scientific community in the future - the data which may be integrated in another resource, the code as the method can be applied to other species, and the website in case a provider wishes to revive the resource. In this way, the scientific community will be able to build on the Philipp's legacy. Philipp inventoried the data and code (applications) which needed to be made open and FAIR to enable future use. The data publicly available for download on the resource was deposited in Zenodo (Bucher and Moretti, 2025) as no domain-specific repository was found. Individual compressed files for each of the fifteen model organism genomes, reformatted following the EPD standard, are now available under a Creative Commons Zero v1.0 Universal license permitting reuse. The copyright has been waived, and it is now in the world-wide public domain. The private EPD code was cleaned, revised, documented, and containerized using Docker before being moved to GitHub and made public at https://github.com/sib-swiss/EPD under a GNU General Public License v3.0 to allow reuse. The repository has the code to run the EPD website; there are web and management scripts, as well as the EPD database dumps. Finally, an end-of-service or "tombstone" page with a description of the resource and links to the archived data, code and website was prepared such that, when EPD is no longer available online, it can be put online for users of the resource to know which elements are available and where. The neXtProt Knowledgebase Established in 2010, the neXtProt knowledge platform for human proteins was built on the knowledge on human proteins found in UniProt (Lane et al., 2012). It employed innovative semantic technologies to integrate data from genomics, transcriptomics, and proteomics, and proposed advanced tools to help the exploration and querying of this data. Between 2011 and 2023, neXtProt was the primary reference resource for the Human Proteome Project (HPP). Seven publications describing the resource are listed in PubMed. The loss of funding and the absence of resources willing to integrate part or the entirety of the data led to transitioning this mature resource to the legacy state. Before neXtProt went offline, the data for all the releases were available for download. The data from three annotation projects were in individual portals, as well the data for the functional proteome project. All the data were exported into the TSV interoperable format and README files prepared which describe the content. Again, the different data were deposited in Zenodo in the absence of a suitable domain-specific repository under the Creative Commons Attribution 4.0 International which allows reuse. The code for the Feature-Viewer (Paladin et al., 2020), a visualization tool for positional annotations on a sequence, which was on GitHub was already available in Zenodo. Additional code, albeit not all, for the neXtProt projects and resources was also publicly available on GitHub at https://github.com/calipho-sib. Registries having neXtProt entries were contacted, and the entries updated to show its legacy state. The resource URL was redirected to the end-of-service page on Expasy at https://www.expasy.org/archives/nextprot. In this first archive page on Expasy, links to the data sets, the Feature-Viewer code, as well as the public neXtProt project code are provided. The Legacy of CleanEx, EPD and neXtProt The transition to legacy state of the three SIB resources presented was due to the loss of key personnel and/or funding. While the CleanEx data and code are no longer accessible, the metadata and publications attest to this early effort. This meagre legacy had no impact on the scientific community. For EPD and neXtProt, no successor was found to take over the running of the resource, and no other resource was found to integrate the data. The legacy for these two resources is FAIR (meta)data and code (at least in part for neXtProt), as well as the publications describing the resources. The value and impact of archiving these bioinformatic resources will be seen in the coming years. The accompanying documentation and interoperability of the formats used in the archives will be critical to whether future bioinformaticians can re-use them or not. Lessons Learned Bioinformatic data resources have survived despite precarious funding and repeated calls to fund them as critical scientific infrastructures, rather than as research projects (Poveda et al., 2025). UniProt, the world’s leading high-quality, comprehensive and freely accessible resource of protein sequence and functional information (UniProt Consortium, 2025), has faced two funding crises, several changes in the team leader and the transition from being a resource provided by a single institute to that provided by a consortium – yet it continues to be a vital resource for the life science community. However, given that the probability that a resource will not survive is high, mature resource providers need to take suitable measures. While resource providers have the responsibility to leave their legacy in order, their institutions also play a role in ensuring they do so. What lessons have we learned in our process of transitioning resources to the legacy state? 1. Use file formats suitable for long-term archiving from the beginning. Interoperability is key to sharing and reusing. 2. Start sunsetting preparations early. Funding or personnel may no longer be available to archive your resource if you wait until the last minute. 3. Try to find a successor. As the resource provider, you are the best person to find a suitable colleague to adopt your resource or a resource which will include your data. 4. Keep only what is useful. Archives have a size limit, and long-term storage comes at a price, so sort through and decide what elements may be useful to the community. 5. Provide the source code. Consider the reason for not making your source code publicly available at a time when this has become the norm and ensures reproducibility. 6. Document data, code, and processes extensively. This is usually a pain point, but it also makes it easier to onboard new team members and for others to reuse and extend your work. 7. Ensure licenses are as permissive as possible. Your legacy must come with a license defining its reuse, whether for its original purpose (conventional reuse) or to fulfil a different function (creative reuse or repurposing). 8. Provide metadata to ensure findability. If your resource cannot be found, it is as though it never existed. 9. Make your legacy publicly available for the long-term. Choose the repository wisely so that the scientific community will have access. 10. Update the information in registries. Let your future users know where to find your legacy by providing an end-of-service page and updating the information on your resource in registries. Acknowledgments We thank Christine Durinx for her valuable suggestions and Christophe Dessimoz for his perspective on resource management, as well as Amos Bairoch for reviewing the manuscript. Funding This work received funding from the Swiss State Secretariat for Education, Research and Innovation (SERI). Conflict of interest statement. None declared. References Gabella C, Duvaud S, Durinx C. Managing the life cycle of a portfolio of open data resources at the SIB Swiss Institute of Bioinformatics. Brief Bioinform. 2022 Jan 17;23(1):bbab478. doi: 10.1093/bib/bbab478. PMID: 34850820; PMCID: PMC8769900. Duvaud S, Gabella C, Lisacek F, Stockinger H, Ioannidis V, Durinx C. Expasy, the Swiss Bioinformatics Resource Portal, as designed by its users. Nucleic Acids Res. 2021 Jul 2;49(W1):W216-W227. doi: 10.1093/nar/gkab225. PMID: 33849055; PMCID: PMC8265094. Schultheiss SJ. Ten simple rules for providing a scientific Web resource. PLoS Comput Biol. 2011 May;7(5):e1001126. doi: 10.1371/journal.pcbi.1001126. Epub 2011 May 26. PMID: 21637800; PMCID: PMC3102757. Helmy M, Crits-Christoph A, Bader GD. Ten Simple Rules for Developing Public Biological Databases. PLoS Comput Biol. 2016 Nov 10;12(11):e1005128. doi: 10.1371/journal.pcbi.1005128. PMID: 27832061; PMCID: PMC5104318. Attwood TK, Agit B, Ellis BM. Longevity of Biological Databases. EMBnet J. 2015;21:e803. doi: 10.14806/ej.21.0.803. Imker HJ, Schackart KE 3rd, Istrate AM, Cook CE. A machine learning-enabled open biodata resource inventory from the scientific literature. PLoS One. 2023 Nov 28;18(11):e0294812. doi: 10.1371/journal.pone.0294812. PMID: 38015968; PMCID: PMC10684096. Praz V, Jagannathan V, Bucher P. CleanEx: a database of heterogeneous gene expression data based on a consistent gene nomenclature. Nucleic Acids Res. 2004 Jan 1;32(Database issue):D542-7. doi: 10.1093/nar/gkh107. PMID: 14681477; PMCID: PMC308841. Praz V, Bucher P. CleanEx: new data extraction and merging tools based on MeSH term annotation. Nucleic Acids Res. 2009 Jan;37(Database issue):D880-4. doi: 10.1093/nar/gkn878. PMID: 19073704; PMCID: PMC2686468. CNCB-NGDC Members and Partners. Database Resources of the National Genomics Data Center, China National Center for Bioinformation in 2022. Nucleic Acids Res. 2022 Jan 7;50(D1):D27-D38. doi: 10.1093/nar/gkab951. PMID: 34718731; PMCID: PMC8728233. Ison J, Ienasescu H, Chmura P, Rydza E, Ménager H, Kalaš M, Schwämmle V, Grüning B, Beard N, Lopez R, Duvaud S, Stockinger H, Persson B, Vařeková RS, Raček T, Vondrášek J, Peterson H, Salumets A, Jonassen I, Hooft R, Nyrönen T, Valencia A, Capella S, Gelpí J, Zambelli F, Savakis B, Leskošek B, Rapacki K, Blanchet C, Jimenez R, Oliveira A, Vriend G, Collin O, van Helden J, Løngreen P, Brunak S. The bio.tools registry of software tools and data resources for the life sciences. Genome Biol. 2019 Aug 12;20(1):164. doi: 10.1186/s13059-019-1772-6. PMID: 31405382; PMCID: PMC6691543. Edgar R, Domrachev M, Lash AE. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res. 2002 Jan 1;30(1):207-10. doi: 10.1093/nar/30.1.207. PMID: 11752295; PMCID: PMC99122. Parkinson H, Kapushesky M, Shojatalab M, Abeygunawardena N, Coulson R, Farne A, Holloway E, Kolesnykov N, Lilja P, Lukk M, Mani R, Rayner T, Sharma A, William E, Sarkans U, Brazma A. ArrayExpress--a public database of microarray experiments and gene expression profiles. Nucleic Acids Res. 2007 Jan;35(Database issue):D747-50. doi: 10.1093/nar/gkl995. Epub 2006 Nov 28. PMID: 17132828; PMCID: PMC1716725. Bastian FB, Cammarata AB, Carsanaro S, Detering H, Huang WT, Joye S, Niknejad A, Nyamari M, Mendes de Farias T, Moretti S, Tzivanopoulou M, Wollbrett J, Robinson-Rechavi M. Bgee in 2024: focus on curated single-cell RNA-seq datasets, and query tools. Nucleic Acids Res. 2025 Jan 6;53(D1):D878-D885. doi: 10.1093/nar/gkae1118. PMID: 39656924; PMCID: PMC11701651. Baldarelli RM, Smith CM, Finger JH, Hayamizu TF, McCright IJ, Xu J, Shaw DR, Beal JS, Blodgett O, Campbell J, Corbani LE, Frost PJ, Giannatto SC, Miers DB, Kadin JA, Richardson JE, Ringwald M. The mouse Gene Expression Database (GXD): 2021 update. Nucleic Acids Res. 2021 Jan 8;49(D1):D924-D931. doi: 10.1093/nar/gkaa914. PMID: 33104772; PMCID: PMC7778941. Baldarelli RM, Smith CM, Finger JH, Hayamizu TF, McCright IJ, Xu J, Shaw DR, Beal JS, Blodgett O, Campbell J, Corbani LE, Frost PJ, Giannatto SC, Miers DB, Kadin JA, Richardson JE, Ringwald M. The mouse Gene Expression Database (GXD): 2021 update. Nucleic Acids Res. 2021 Jan 8;49(D1):D924-D931. doi: 10.1093/nar/gkaa914. PMID: 33104772; PMCID: PMC7778941. Dreos R, Ambrosini G, Cavin Périer R, Bucher P. EPD and EPDnew, high-quality promoter resources in the next-generation sequencing era. Nucleic Acids Res. 2013 Jan;41(Database issue):D157-64. doi: 10.1093/nar/gks1233. Epub 2012 Nov 27. PMID: 23193273; PMCID: PMC3531148. Bucher P, Moretti S. EPDnew genomes [Data set]. In Nucleic Acids Research (Vol. 41, Number D1, pp. D157–164). Zenodo. 2025 doi: 10.5281/zenodo.15706069. Lane L, Argoud-Puy G, Britan A, Cusin I, Duek PD, Evalet O, Gateau A, Gaudet P, Gleizes A, Masselot A, Zwahlen C, Bairoch A. neXtProt: a knowledge platform for human proteins. Nucleic Acids Res. 2012 Jan;40(Database issue):D76-83. doi: 10.1093/nar/gkr1179. Epub 2011 Dec 1. PMID: 22139911; PMCID: PMC3245017. Paladin L, Schaeffer M, Gaudet P, Zahn-Zabal M, Michel PA, Piovesan D, Tosatto SCE, Bairoch A. The Feature-Viewer: a visualization tool for positional annotations on a sequence. Bioinformatics. 2020 May 1;36(10):3244-3245. doi: 10.1093/bioinformatics/btaa055. PMID: 31985787. Poveda L, Farrell G, Tosatto S, Zahn M, Ruch P, Gobeill J, Waterhouse R, Dessimoz C. The missing link in FAIR data policy: data resources. Zenodo. 2025 doi: 10.5281/zenodo.15724104. UniProt Consortium. UniProt: the Universal Protein Knowledgebase in 2025. Nucleic Acids Res. 2025 Jan 6;53(D1):D609-D617. doi: 10.1093/nar/gkae1010. PMID: 39552041; PMCID: PMC11701636.