Progress 09/01/24 to 08/31/25
Outputs
Target Audience:Our target audiences include veterinarians, producers, and researchers across multiple disciplines. This initiative also supports students at all levels, who will receive comprehensive training and mentorship throughout the project. To promote effective collaboration and knowledge-sharing, the K-State Olathe (KSO) team holdsweekly meetings and participates in quarterly gatherings with other FARAD units, providing a medium to present and discuss new findings and science-based insights. These efforts foster a dynamic, inclusive environment that empowers all participants to contribute actively to advancing food safety and residue management. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?Here is the estimated FTE breakdown for the team: Two graduate students: 0.5 FTE each (Total: 1.0 FTE) One postdoc: 1 FTE Overall, the team consists of 2Full-Time Equivalents (FTE). How have the results been disseminated to communities of interest?The FARAD call center and online platforms serve as primary channels for disseminating information, providing timely and relevant data to veterinarians, producers, and researchers. Completed research projects are published in peer-reviewed journals, with a comprehensive list available in the Products section. In addition, FARAD actively shares findings at local and national conferences, ensuring broad access to research outcomes. Through these dissemination strategies, FARAD supports knowledge-sharingand advances food safety and residue management across the scientific community. What do you plan to do during the next reporting period to accomplish the goals?During the next reporting period, we plan to advance several key activities to accomplish our goals: Enhance AI and Data Extraction Tools - Continue refining hybrid approaches that combine LLMswith rule-based methods for PKand residue data extraction, building on our prior work using Knowledge in Triples and Semantic Table QA. Expand Database Access and Functionality - Further develop interfaces and navigation systems to facilitate global access to FARAD data, including Global MRLsand WDPs, using HTML, XML, and PDF formats. Full-Text Retrieval and Metadata Analysis - Implement and test the full-text retrieval system for veterinary and FARAD bibliographic records, performing metadata analysis to validate and enhance the database. Collaboration and Training - Maintain weekly meetings within the KSO team and quarterly interactions with other FARAD units. Provide ongoing mentorship and training tostudents, ensuring effective knowledge transfer. Dissemination of Findings - Prepare manuscripts and conference presentations to share new research results. Continue publishing in peer-reviewed journals and presenting at local and national meetings to reach a broad audience of stakeholders. Integration of Hybrid Models for Machine Learning - Develop and validate hybrid models that combine content-based and bibliometric features for efficient extraction, cleaning, and analysis of global PK and residue data to support next-generation FARAD applications. These activities are designed to ensure that we meet FARAD's objectives of providing safe, animal-derived food products and supporting evidence-based decision-making in food safety and residue management.
Impacts
What was accomplished under these goals?
The national FARAD program aims to protect the American public by ensuring that animal-derived food products (milk, eggs, meat, honey, etc.) are free from violative or potentially unsafe chemical residues, including drugs, pesticides, environmental contaminants, natural toxins, and other harmful substances. Kansas State University-Olathe (KSUO) will develop interfaces enabling FARAD responders to access KSUO's comparative medicine databases, computational tools, and relevant global data. The KSUO team is advancing practical AI technologies for food-animal production, focused on collecting, organizing, merging, and cleaning data, and facilitating easy query submission for global maximum residue limits (MRLs) and withdrawal periods (WDPs). In support of these efforts, we developped models and have published work on "Enhancing Pharmacokinetic Data Extraction with LLMs and Rule-Based Methods: A Hybrid Approach Using Machine Learning and Regex," where we employed Knowledge in Triples for LLMs and Semantic Extraction for Table QA to improve accuracy in extracting global pharmacokinetic (PK) data. Building on this, the team is designing a full-text retrieval system to create a queryable information retrieval database, performing metadata analysis on veterinary bibliographic records, and validating against the current FARAD database. This system underpins hybrid models combining content-based and bibliometric features for machine learning applications, enabling efficient extraction and cleaning of data from diverse resources and forming a core component of the next-generation FARAD program.
Publications
- Type:
Peer Reviewed Journal Articles
Status:
Published
Year Published:
2025
Citation:
Wu X, Chen Q, Chou WC, Maunsell FP, Tell LA, Baynes RE, Davis JL, Jaberi-Douraki M, Riviere JE, Lin Z. Development of a physiologically based pharmacokinetic model for flunixin in cattle and swine following dermal exposure. Toxicological Sciences. 2025 Feb;203(2):181-94.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2025
Citation:
Remya Ampadi Ramachandran, Hossein Sholehrasa, Lisa Tell, Doina Caragea, Majid Jaberi-Douraki, Enhancing Pharmacokinetic Data Extraction with LLMs and Rule-Based Methods: A Hybrid Approach Using Machine Learning and Regex, AAVPT 23rd Biennial Symposium, Raleigh, NC, May 18-21, 2025, Pages 42-47, https://cdn.ymaws.com/www.aavpt.org/resource/resmgr/biennial_2025/proceedings_for_the_23rd_aav.pdf
- Type:
Other Journal Articles
Status:
Published
Year Published:
2024
Citation:
Sholehrasa H, Norouzi SS, Hitzler P, Jaberi-Douraki M. Knowledge in Triples for LLMs: Enhancing Table QA Accuracy with Semantic Extraction. arXiv preprint arXiv:2409.14192. 2024 Sep 21.
https://arxiv.org/pdf/2409.14192v1
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2025
Citation:
Avoiding Drug Residues: A Multivariate Approach to Estimating Withdrawal Interval in Goat Edible Tissues Following Extralabel Administration of Flunixin Meglumine. Authors: Sheela FF, Baynes RE, Hughes-Oliver J, Riviere J, Jaberi-Douraki M. The 23rd Biennial Meeting of the American Association of Veterinary Pharmacology and Therapeutics, Raleigh, NC. Proceedings available at https://www.aavpt.org/page/SymposiumWorkshops. (May 18-21, 2025)
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2024
Citation:
Parrish BL, Holliday AJ, Hulbert LE, Goering MJ, Forster SL, Vieregger AE, Jaberi-Douraki M, Mote BE, Schmidt TB. 269 Evaluation of changes in the activity of group-housed nursery pigs exposed to an endotoxin challenge using the NUtrack Livestock Monitoring System. Journal of Animal Science. 2024 Sep 1;102(Supplement_3):5-6.
- Type:
Conference Papers and Presentations
Status:
Other
Year Published:
2025
Citation:
Enhancing Pharmacokinetic Data Extraction with LLMs and Rule-Based Methods: A Hybrid Approach Using Machine Learning and Regex. Authors: Remya Ampadi Ramachandran, Hossein Sholehrasa, Lisa Tell, Doina Caragea, Majid Jaberi-Douraki. The 23rd Biennial Meeting of the American Association of Veterinary Pharmacology and Therapeutics, Raleigh, NC. Proceedings available at https://www.aavpt.org/page/SymposiumWorkshops. (May 18-21, 2025)
|