The FAIR Data Point Populator: collaborative FAIRification and population of FAIR Data Points

14 15

16 Use of the FAIR principles (Findable, Accessible, Interoperable and Reusable) allows the rapidly 17 growing number of biomedical datasets to be optimally (re)used. An important aspect of the FAIR 18 principles is metadata. The FAIR Data Point specifications and reference implementation have been 19 designed as an example on how to publish metadata according to the FAIR principles. Various tools to 20 create metadata have been created, but many of these have limitations, such as interfaces that are 21 not intuitive, metadata that does not adhere to a common metadata schema, limited scalability, and 22 inefficient collaboration. We aim to address these limitations in the FAIR Data Point Populator. 23 24 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 2

25 The FAIR Data Point Populator consists of a GitHub workflow together with Excel templates that have 26 tooltips, validation and documentation. The Excel templates are targeted towards non-technical users, 27 and can be used collaboratively in online spreadsheet software. A more technical user then uses the 28 GitHub workflow to read multiple entries in the Excel sheets, and transform it into machine readable 29 metadata. This metadata is then automatically uploaded to a connected FAIR Data Point. We applied 30 the FAIR Data Point Populator on the metadata of two datasets, and a patient registry. We were then 31 able to run a query on the FAIR Data Point Index, in order to retrieve one of the datasets. 32 33

34 The FAIR Data Point Populator addresses several limitations of other tools. It makes creating 35 metadata easier, ensures adherence to a common metadata schema, allows bulk creation of 36 metadata entries and increases collaboration. As a result of this, the barrier of entry for FAIRification is 37 lower, which enables the creation of FAIR data by more people. 38 39

40 FAIR, Metadata, FAIR Data Point, RDF, Semantic web, data sharing 41

42 The rate of data acquisition in biomedical research is rapidly increasing [1]. In order to enable optimal 43 (re)use of this growing number of datasets, it is important that they are Findable, Accessible, 44 Interoperable and Reusable (FAIR) for both humans and machines [2]. An important aspect of increasing 45 the degree by which data is FAIR (FAIRification), is the provisioning of rich metadata (Principle R1). 46 47 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 3 The FAIR Data Point (FDP) reference application [3] has been designed to serve as an example of how 48 to publish metadata according to the FAIR principles. The FDP specification consists of a metadata 49 schema structure based on the Data Catalog Vocabulary (DCAT), which has been designed to share 50 metadata of different types of digital objects in an interoperable manner [4]. The FDP reference 51 application serves as a reference implementation of the FDP specification [5]. It has interfaces 52 (webpages, documents, API and SPARQL endpoint) that present the metadata following this schema, 53 and allows for creation and modification of entries both for humans and machines. The machine 54 readable metadata for the FDP is represented as RDF (Resource Description Framework) documents 55 [6] in order to adhere to the FAIR principle I1, of formal knowledge representation of metadata. Finally, 56 there is service for FDPs that indexes metadata from other FAIR Data Points, and allows queries to be 57 executed on all this metadata simultaneously. 58 59 Various tools have been developed for creating and publishing metadata. These include tools such as 60 CEDAR [7], RightField [8], the FAIR data station [9], the OpenRefine metadata extension [10] and the 61 FAIR metadata editor [11]. These tools all have their own design philosophy, each with their own 62 advantages and disadvantages. For example, CEDAR, OpenRefine and the FAIR metadata editor work 63 primarily with web interfaces, while RightField and the FAIR data station are focused on Excel. Some of 64 the tools are more adaptable, while some of the tools are more easy to use. 65 66 There are some limitations to existing tools. Most importantly, many tools are not easy to use for people 67 who are not familiar with FAIR. They also lack a convenient way to make sure that the metadata is 68 published and indexed, which is one of the requirements for the FAIR principles. Further, many tools 69 require the need to add every entry manually, are not easy to use in collaborations and do not offer 70 version control of the metadata. On the side of availability, many tools require local installations, or are 71 only available online, and thus rely on either local installations, or the continued hosting of the provider 72 of the tool. Finally, some tools create metadata that do not adhere to a common metadata schema. 73 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 4 74 In this work, we implement a FAIRification tool called the “FAIR Data Point Populator”, that aims to 75 address these limitations, and hereby make the provisioning of FAIR metadata as easy as possible. The 76 tool is written in Python, and is accessible as either a GitHub workflow, or a Jupyter notebook. It reads 77 an Excel template, and automatically publishes the contained metadata in a FDP application as RDF. 78 Implementation 79 The FDP Populator consists of an Excel template together with a python tool. The Excel template 80 contains the metadata that is needed to populate the FDP, which is based on the DCAT schema. There 81 are two sheets in the Excel template that can be filled in, the dataset sheet, and the distribution sheet. 82 The column names in these sheets are shown in table 1. In order to aid in filling in the template, all of 83 the columns are described in the readme file in the same GitHub repository as the template. In addition, 84 the template contains tooltips to describe what information should be filled in every column, and what 85 the requirements are. There is also validation applied on the Excel fields to give warning when metadata 86 does not meet the requirements. For example, when a link is not actually a link. Some extra metadata, 87 like the date of publication and modification, are automatically added by the FDP. 88 89 This Excel template is then used by the FDP Populator tool, which performs the actual publication 90 process. A diagram of its function is shown in figure 1. The tool reads the Excel file, and turns the 91 information into RDF which follows the FDP metadata schema. Finally, the RDF is automatically sent to 92 a FDP using the API of the FDP. URIs of the newly made resources are used in the RDF of the 93 succeeding resources. Both the conversion and the publishing scripts are written in python. We chose 94 to implement these steps as a GitHub workflow in order to make use of GitHubs features. A Jupyter 95 notebook that runs the FDP Populator has also been created as an alternative that does not depend on 96 GitHub. 97 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 5 98 In order to make things as easy as possible for data owners, the tool has been developed with a 99 collaboration with a data steward in mind. In this case, the data owners have the role of filling in the 100 Excel template, with metadata, while the data steward has the role of uploading this template to GitHub, 101 and starting the workflow that runs the FDP Populator. A diagram of this workflow is shown in figure 2. 102 103 The European Joint Programme on Rare Diseases (EJP RD) has created a metadata schema to 104 describe rare disease resources [12]. This schema is focused on making it possible to find resources 105 such as biobanks, patient registries and guidelines. Like the FDP specification, The EJP RD metadata 106 schema is an extension of the DCAT schema. We extended the FDP Populator to use the Excel template 107 from the EJP RD and thereby help disease researchers to describe and publish their resources in a FDP 108 with the help of the FDP populator. 109

110 Collaborative FAIRification of datasets using automated publication 111 To show the application of the FDP Populator, we applied it on a number of resources in the context of 112 the EJP RD consortium. First, we applied it on a collection of Inclusion Body Myositis data. These 113 datasets included raw and normalized transcriptomics, raw and normalized miRNA transcriptomics, and 114 whole exome sequencing together with sample information. All of the metadata for these entries were 115 filled in the Excel template in collaboration with the data owners. The dataset sheet with these entries is 116 shown in figure 3. We then ran the FDP Populator on this Excel file in order to generate the 117 corresponding RDF, which can be seen in listing 1 and 2. Finally, immediately after the last step, the 118 tool also automatically populated the FDP with this RDF, which can be seen in figure 4. This made it 119 unnecessary to upload all the metadata manually. Also note that the datasets were all published in one 120 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 6 run of the tool, without the need to repeat the process for every dataset. The input Excel file and output 121 RDF can be found at: https://github.com/jdwijnbergen/fdp-populator-usecases, along with a link of the 122 resulting entries in the FDP. 123 124 We also show the application of the FDP Populator in two other use cases. The first one is three datasets 125 for CAKUT (Congenital Anomalies of the Kidney and Urinary Tract). This included a mirnome, peptidome 126 and proteome dataset. In this use case, we took advantage of online spreadsheet software to fill in the 127 Excel template. We found that users were able to work on different sections of the Excel template 128 simultaneously, which greatly sped up the process. The resulting Excel file and RDF can be found in the 129 same GitHub repository, along with the link to the entries on the FDP. Finally, showing the 130 implementation of the EJP RD metadata schema in our tool, we also applied the tool on the ERKNet. 131 ERKNet is a patient registry that focuses on rare kidney diseases. In this case, we filled in the EJP RD 132 Excel template, which is adapted to the EJP RD metadata schema. The Excel file and the corresponding 133 output can be found in the same GitHub repository. 134 The FAIRified resources are findable through RDF queries 135 In order to find datasets related to CAKUT, and show that retrieving these datasets is possible for 136 humans and machines, we created a SPARQL (SPARQL Protocol and RDF Query Language) query 137 to run on RDF. This query retrieves entities that are an instance of the dataset class in the DCAT 138 schema, and have CAKUT (in ontologized form) as one of their themes. The query is shown in figure 139 5. This query was executed on the FAIR Data Point index on the 7th of August 2023. At this time, 140 there were 46 active FAIR Data Points, including the FDP with the CAKUT metadata, on which the 141 index applied the query. We were able to find the three CAKUT datasets that we applied the FDP 142 Populator on (figure 5). It is important to note that this query is correct not only for metadata that is 143 created by our tool, but for any metadata that conforms to the DCAT schema. 144 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 7

145 In this work, we created the FDP Populator, which addresses some limitations of previous tools to make 146 it easy to use, enable collaboration, and ensure metadata adhering to a common metadata schema 147 among other advantages. We applied the FDP Populator on a number of use cases, and show that the 148 created metadata allows the data to be found by humans and machines. 149 150 The FDP Populator has four main advantages. Firstly, the FPD Populator primary objective is to simplify 151 the provisioning of FAIR metadata, and hereby lower the barrier of entry. This is achieved by letting the 152 data owners use the Excel template to describe their resources, which is likely done in software that 153 they are familiar with. Secondly, the FDP Populator allows for the bulk upload of multiple entries to a 154 catalog. This makes the creation of a large number of entries in a catalog feasible. Thirdly, the template 155 can be filled in simultaneously by multiple people, which reduces the time needed for FAIRification. 156 Fourthly, the tool can be run as a GitHub workflow, or as a Python notebook, which reduces the reliance 157 on a local or hosted tool. Additionally, as a GitHub workflow, it keeps the version history to ensure no 158 metadata is lost. 159 160 The FAIR Data Populator also improves the interoperability of the created metadata. Many tools are 161 very flexible, and allow any type of triples to be made. This however can result in RDF that is not easy 162 to use and interpret. The FAIR Data Populator instead adheres to the commonly used DCAT metadata 163 schema as it is implemented in the FDP, and in the EJP RD extension hereof, which ensures 164 interoperability with other resources. 165 166 The FDP Populator also has some limitations. It is relatively inflexible in regards to data schemas due 167 to the focus on the FDP and EJP RD metadata schemas. This is a tradeoff made to ensure it outputs 168 proper metadata, though it can be limiting for some use cases. The FDP Populator could be extended 169 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 8 to create Excel templates based on FDP metadata schemas, in order to work with rich domain specific 170 metadata schemas. This would however also decrease the user friendliness, as documentation, tooltips 171 and validation would have to be created by the administrator, and a schema would have to be available 172 on a FDP. Further, the FDP Populator also relies on the access to a FAIR Data Point, since it does not 173 host the metadata itself, so a FAIR Data Point needs to be created. Finally, the tool is not hosted online 174 as a webtool, which would lower the barrier of entry even more. 175

176 In this work we created the FAIR Data Populator with the goal of addressing limitations of current tools 177 in the field. Our tool creates RDF from information filled in an Excel template, and automatically uploads 178 this to a connected FDP. It improves the ease of use through using an Excel template with 179 documentation, automatic publication of results, allowing the bulk creation of FDP entries within a 180 catalog, and allowing the use of collaborative software such as online spreadsheet software and GitHub, 181 and. It is less reliant on hosting by external parties, and if used as a GitHub tool, keeps track of the 182 version history of the input metadata. Finally, it also improves the operability of the metadata by adhering 183 to commonly used specifications like the FDP specification, and the EJP RD metadata schema which 184 are based on the DCAT schema. 185 186 With these advantages, the barrier of entry of FAIRification will be lower. Together with the use of 187 commonly used data schemas, this ensures that more data will be FAIR. 188 Availability and requirements 189 Project name: FAIR Data Point Populator 190 Project home page: https://github.com/jdwijnbergen/fdp-populator 191 Operating system(s): Platform independent 192 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 9 Programming language: Python 193 Other requirements: GitHub or Google Colab. For local use, Python3 with required packages. 194 License: MIT 195 Any restrictions to use by non-academics: no 196 List of abbreviations 197 CAKUT - Congenital Anomalies of the Kidney and Urinary Tract 198 DCAT - Data Catalog Vocabulary 199 EJP RD - European Joint Programme on Rare Diseases 200 FAIR - Findable, Accessible, Interoperable and Reusable 201 FDP - FAIR Data Point 202 RDF - Resource Description Framework 203 SPARQL - SPARQL Protocol and RDF Query Language 204 Availability of data and materials 205 The FDP Template is available at https://github.com/LUMC-BioSemantics/EJP-RD-WP13-FDP-206 template, The EJP RD template is available at https://github.com/ejp-rd-vp/resource-metadata-207 schema/tree/master/template and the repository with use cases is available at 208 https://github.com/jdwijnbergen/fdp-populator-usecases. 209 Competing interests 210 The authors declare that they have no competing interest. 211 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 10 Funding 212 This initiative has received funding from the European 213 Union's Horizon 2020 research and innovation programme 214 under grant agreement N°825575. 215 Authors’ contribution 216 D.W. and R.K.: Conceptualization, development and application on use cases. K.B.: Infrastructure 217 support. D.W. and E.M.: Manuscript preparation. E.M. and M.R.: Supervision. D.W., R.K., K.B. L.S., 218 B.M., M.R. and E.M., contributed to the manuscript, reviewed the manuscript, and gave approval for 219 publication. 220

221 We would like to thank Henriette for the creation of the EJP RD metadata schema, Mirdul Johari for 222 the collaboration on the Inclusion Body Myositis use case, Friederike Ehrhart for the collaboration on 223 the CAKUT use case, and Jose Ramírez García for the collaboration on the ERKNet use case. 224

225 [1] J. D. Van Horn and A. W. Toga, “Human neuroimaging as a ‘Big Data’ science,” Brain Imaging 226 Behav., vol. 8, no. 2, pp. 323–331, Jun. 2014, doi: 10.1007/s11682-013-9255-y. 227 [2] M. D. Wilkinson et al., “Comment: The FAIR Guiding Principles for scientific data management 228 and stewardship,” Sci. Data, vol. 3, pp. 1–9, 2016, doi: 10.1038/sdata.2016.18. 229 [3] L. O. B. da Silva Santos, K. Burger, R. Kaliyaperumal, and M. D. Wilkinson, “FAIR Data Point: A 230 FAIR-Oriented Approach for Metadata Publication,” Data Intell., vol. 5, no. 1, pp. 163–183, Mar. 231 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 11 2023, doi: 10.1162/dint_a_00160. 232 [4] R. Albertoni et al., “Data Catalog Vocabulary (DCAT) - Version 2.” https://www.w3.org/TR/vocab-233 dcat-2/ (accessed Feb. 10, 2023). 234 [5] L. O. Bonino, K. Burger, and R. Kaliyaperumal, “FAIR Data Point specifications.” 235 https://specs.fairdatapoint.org/ (accessed Aug. 10, 2023). 236 [6] R. Cyganiak, D. Wood, M. Lanthaler, G. Klyne, J. J. Carroll, and B. McBride, “RDF 1.1 Concepts 237 and Abstract Syntax.” https://www.w3.org/TR/rdf11-concepts/ (accessed Jul. 04, 2023). 238 [7] M. A. Musen et al., “The center for expanded data annotation and retrieval,” J. Am. Med. Inform. 239 Assoc., vol. 22, no. 6, pp. 1148–1152, Nov. 2015, doi: 10.1093/jamia/ocv048. 240 [8] K. Wolstencroft et al., “RightField: embedding ontology annotation in spreadsheets,” 241 Bioinformatics, vol. 27, no. 14, pp. 2021–2022, Jul. 2011, doi: 10.1093/bioinformatics/btr312. 242 [9] B. Nijsse, P. J. Schaap, and J. J. Koehorst, “FAIR data station for lightweight metadata 243 management and validation of omics studies,” GigaScience, vol. 12, p. giad014, Jan. 2023, doi: 244 10.1093/gigascience/giad014. 245 [10] M. Suchánek, “OpenRefine-metadata-extension.” FAIR Data Team, Nov. 18, 2022. Accessed: 246 Jul. 05, 2023. [Online]. Available: https://github.com/FAIRDataTeam/OpenRefine-metadata-247 extension 248 [11] K. Burger, “FAIR-metadata-editor.” FAIR Data Team, Sep. 22, 2016. Accessed: Jul. 05, 2023. 249 [Online]. Available: https://github.com/FAIRDataTeam/FAIR-metadata-editor 250 [12] R. Kaliyaperumal, S. Jupp, H. Harmse, O. Ajigboye, R. Cornet, and M. Wilkinson, “Metadata for 251 EJP rare disease patient registries, biobanks and catalogs.” EJP RD Pillar 2: Coordinated Access 252 to Data and Services, Jun. 19, 2023. Accessed: Aug. 10, 2023. [Online]. Available: 253 https://github.com/ejp-rd-vp/resource-metadata-schema 254 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 12 Table 1: The column names in the Excel template for the “dataset” sheet, and the “distribution” sheets. Required columns are marked with an asterisk. Dataset Distribution Title* Title* Publisher* Dataset title* Themes* Publisher* Description AccessURL* Language Download URL* License MediaType* ContactPoint Description LandingPage Language

License CompressionFormat Format ByteSize (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 13 Figure 1: The workflow of the FAIR Data Point Populator. A spreadsheet is read into the FPD Populator, then converted to RDF. The RDF is sent to the FDP after which the FDP returns a URI to this new entry. The FDP Populator can then link to this URI from the following resources. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 14 Figure 2: The workflow for users, split into data owner and data steward. The data owner creates the metadata spreadsheet for their data with help from documentation tooltips and validation. The data steward uploads this to GitHub, and runs the GitHub workflow. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 15 Figure 3: Part of the dataset sheet of the Excel template filled in with metadata for the Inclusion Body myositis datasets. The tooltip for the publisher column is shown. The template also contains validation. Required fields are marked with an asterisk, and have been filled. Some of the optional fields have been left empty. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 16 Listing 1: The RDF metadata created for the raw expression IBM dataset expressed in the Turtle RDF format. Note that the FDP adds additional triples to this. @prefix dcat: . @prefix dct: . @prefix foaf: . @prefix language: . @prefix loc: . loc:new a dcat:Dataset, dcat:Resource ; dct:description "Gene expression for Inclusion Body Myositis" ; dct:hasVersion "1" ; dct:isPartOf ; dct:language language:en ; dct:publisher ; dct:title "IBM Gene Expression Raw" ; dcat:keyword "Gene expression", "Homo Sapiens", "Inclusion Body Myositis", "Person with amputation", "RNA-Seq", "RNA-seq read count analysis", "Skeletal muscle tissue" ; dcat:theme , , , , , , . a foaf:Agent ; foaf:name "https://orcid.org/0000-0002-7449-6657" . (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 17 Listing 2: The RDF metadata created for the raw expression IBM distribution expressed in the Turtle RDF format.. Note that the FDP adds additional triples to this. @prefix dcat: . @prefix dct: . @prefix foaf: . @prefix language: . @prefix loc: . loc:new a dcat:Distribution, dcat:Resource ; dct:description "Distribution of IBM raw gene expression" ; dct:hasVersion "1" ; dct:isPartOf ; dct:language language:en ; dct:publisher ; dct:title "Downloadable distribution of : IBM Gene Expression Raw" ; dcat:byteSize "7586125" ; dcat:downloadURL ; dcat:mediaType "text/tab-separated-values" . a foaf:Agent ; foaf:name "https://orcid.org/0000-0002-7449-6657" . (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 18 Figure 4: The metadata after it is automatically published on the FAIR Data Point. The dataset metadata can be browsed further, and data downloaded where it is available. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint 19 Figure 5: The results of a SPARQL query that looks for any datasets that have CAKUT as their theme. The FAIRified CAKUT datasets are found in the results. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted September 10, 2024. ; https://doi.org/10.1101/2024.09.06.611505doi: bioRxiv preprint