AgBioData-LIST October 2026

agbiodata-list@bioinfo.wsu.edu

1 participants
2 discussions

AgBioData Webinar November 2026 on AI-Assisted Literature Curation by the NLP for Biocuration WG

by MARCELA KAREY TELLO-RUIZ

*How do we move from promising AI tools to practical, scalable GGB literature curation?* Join us on *Wednesday, November 4* for a webinar with Olivia Haley (USDA), Kapeel Chougule (CSHL), and Laura Tibbs-Cortes (USDA) from the AgBioData NLP for Biocuration Working Group on the *lessons learned from working with Textpresso, plant papers and genes*. Since its formation in March 2025, the NLP for Biocuration Working Group has been evaluating AI-assisted tools and approaches to help scale literature curation across agricultural genomics databases. The group has explored tools including Textpresso, BioBERT, Elicit, and OntoGPT, developed shared benchmarking practices, and coordinated efforts toward a Textpresso proof of concept tailored to agricultural species. As part of the monthly AgBioData Webinar Series <https://www.agbiodata.org/webinars>, Olivia, Kapeel, and Laura will share lessons learned from evaluating these tools, benchmarking AI-assisted literature workflows, and building shared infrastructure to support literature curation across the agricultural genomics community. The presentation will also introduce the AgBioData implementation of Textpresso, including its graphical user interface (GUI), API, and command-line tools. Wednesday, November 4 | 1P ET | 12P CT | 11A MT | 10A PT | Find your local time here <https://www.timeanddate.com/worldclock/fixedtime.html?msg=AgBioData+Nov+202…>. Note that it will be Standard Time in the United States (no longer daylight saving time). Join Zoom Meeting https://us06web.zoom.us/j/82038356125?pwd=YVFMRElMdEpHZmtObXFvZlA4QVFXQT09 Meeting ID: 820 3835 6125 Passcode: 160683 *N**LP4Biocuration: Advancing FAIR, AI-Assisted Literature Curation Across the AgBioData Consortium The AgBioData* The NLP4Biocuration Working Group was formed in March 2025 to evaluate, prototype, and coordinate the integration of natural-language-processing (NLP) and large language model (LLM) tools to support scalable, ontology-driven literature curation across agricultural genomics databases. This effort is conducted under the umbrella of the AgBioData Consortium and is supported by the *NSF Research Coordination Network award #2126334* <https://www.agbiodata.org/projects/nsf-rcn>, a shared community resource initiative that promotes interoperable informatics solutions. Over the past eight months, working group members performed a coordinated assessment of AI-assisted literature tools, developed shared benchmarking practices, and explored infrastructure options for cross-database deployment. Tools evaluated include BioBERT, Elicit, Textpresso, OntoGPT, and early agentic-AI frameworks, with attention to their utility for named entity recognition, relationship extraction, summarization, and paper prioritization for plant and crop species. Several members implemented pilot pipelines comparing AI-generated literature sets with manually curated datasets, providing practical insights into recall, precision, and model behavior within agriculture-specific contexts. To facilitate shared learning and reproducibility, the group established a structured "tool review" framework and a community repository of test cases, datasets, and documented workflows. A major milestone is the ongoing organization of a Textpresso proof-of-concept deployment tailored for agricultural species. Participating AgBioData resources, including SorghumBase, MaizeGDB, GDR, SoyBase, GrainGenes, and others, are assembling FAIR-compliant gene lists, entity vocabularies, and open-access corpora, and are coordinating with the Alliance of Genome Resources team on compute requirements for a shared AWS-based test environment. In parallel, the group distributed a survey across AgBioData member databases to capture literature volume, workflows, AI adoption, and technical barriers; survey findings are guiding both a manuscript currently in development and continued infrastructure planning. Aligned with FAIR principles, the AgBioData consortium advances sustainable informatics infrastructure. By aligning NLP4Biocuration activities with these principles, the working group is helping establish sustainable informatics infrastructure and analytics-driven methods that benefit the entire agricultural genomics community. This presentation will cover the AgBioData implementation of Textpresso, including the GUI and programmatic access by API and command line tools. See also this news blog <https://www.agbiodata.org/node/654>. Marcela -- Marcela Karey Tello-Ruiz, PhD AgBioData Program Manager https://www.agbiodata.org/ Phoenix Bioinformatics

1 week, 6 days

Reminder Tomorrow: PlantGraph Webinar, Wed 6 October 2026 -- AgBioData

by MARCELA KAREY TELLO-RUIZ

[image: :seedling:] *What if the answer isn’t in a database—but in the connections between databases?* Join us on *Wednesday, October 7* for a webinar with Ehsan Estaji of the Umeå Plant Science Centre, Sweden: [image: :link:] Trapped knowledge in the AI-agent era: how PlantGraph turns multi-database questions into one traversal you can check PlantGraph connects plant data across the transcriptome, epigenome, regulome, and proteome, helping researchers turn complex, multi-database questions into one connected, traceable query. As part of the monthly AgBioData Webinar Series <https://www.agbiodata.org/webinars>, Ehsan will show how this approach can help AI agents retrieve evidence more reliably—and how we can check whether AI-generated answers actually stay grounded in that evidence. Wednesday, October 7 | 1P ET | 12P CT | 11A MT | 10A PT | Find your local time here <https://www.timeanddate.com/worldclock/fixedtime.html?msg=AgBioData+Oct+202…> . Join Zoom Meeting https://us06web.zoom.us/j/82038356125?pwd=YVFMRElMdEpHZmtObXFvZlA4QVFXQT09 Meeting ID: 820 3835 6125 Passcode: 160683 Title: Trapped knowledge in the AI-agent era: how PlantGraph turns multi-database questions into one traversal you can check Abstract: A real biological question is rarely a record — it is a path. “Which salt-responsive genes also carry a promoter methylation change, a transcription-factor binding site, and a known phosphosite?” has no home database: the answer exists only as a traversal across four of them. This is why so much of what we collectively know sits trapped. Not because the facts are missing — they are curated, somewhere — but because the connections between them were never stored anywhere. Researchers rebuild those connections by hand, one question at a time, and lose them the moment they close the tabs. Capable AI agents raise the cost of that fragmentation rather than resolving it. An agent handed a set of endpoints has to plan the join itself, guessing identifiers and directions at every hop, and every hop is an opportunity to improvise. In our own benchmarking, giving a strong model tools and API access over fragmented sources did not reliably fix retrieval; connecting the data first, and retrieving over it deterministically, did. PlantGraph (plantgraph.se) integrates community plant resources into one connected knowledge graph and separates two jobs that agents usually conflate: the graph retrieves the evidence, the language model only writes over it. A multi-database question becomes a single traversal — and because that traversal actually ran, every claim can be followed back to the query that produced it and the source record that curated it. I will walk through questions spanning transcriptome, epigenome, regulome and proteome resolving in one step, how we measure when the generated narration drifts from the retrieved evidence, and where the approach still fails. PlantGraph <https://plantgraph.se/> is not a replacement for the databases it draws on — it stands on them, and every resource that connects makes the rest worth more. See also this news blog <https://www.agbiodata.org/node/652>. Marcela -- Marcela Karey Tello-Ruiz, PhD AgBioData Program Manager https://www.agbiodata.org/ Phoenix Bioinformatics

1 month

2026

2025

2024

2023

2022

2021

2020

2019

AgBioData-LIST October 2026