The Lineberger Bioinformatics Core partners with investigators to apply AI, machine learning, and large language models to their research. We help investigators analyze complex datasets, develop image analysis workflows, extract information from documents, and build custom tools. Our team provides support from project planning and method selection through development and deployment.
AI and Machine Learning Capabilities
We use a range of methods and technologies to match the needs of each project.
Generative AI & Large Language Models
Use language models for information extraction, summarization, question answering, semantic search, and conversational research tools.
Machine Learning & Predictive Modeling
Develop models for classification, prediction, biomarker discovery, and analysis across different types of biomedical data.
Fine-Tuning & Model Customization
Adapt existing models through retrieval, fine-tuning, and other approaches when a general-purpose model is not enough.
Image Analysis & Computer Vision
Analyze biomedical images through classification, segmentation, feature extraction, object detection, and related methods.
AI Agents & Research Applications
Connect language models with databases, APIs, computational tools, and research systems to support multi-step tasks.
Secure Local AI Infrastructure
Run language and embedding models on local GPU infrastructure when privacy, security, performance, or control requires it.
AI in Action
Current projects show how these approaches are being used in practice:
Clinical Trial Discovery & Search
A locally hosted search platform combines trial metadata with information extracted from study protocols, supporting semantic search, trial summaries, questions about individual studies, and eligibility review.
Automated Protocol & Lab Manual Extraction
Language models extract time points, specimen requirements, and collection details from trial laboratory manuals. Staff review the structured output before it is imported into LIMS.
AI-Assisted Clinical Trial Matching
A proof-of-concept agent evaluates trial eligibility criteria against patient demographics, laboratory results, procedures, and pathology data using controlled access to the information it needs.
Responsible and Appropriate Use
Not every research question calls for generative AI or a large language model. In some cases, a traditional machine learning, statistical, or other computational approach may be more appropriate. We work with investigators to choose a method that fits the research question and available data, determine how results will be evaluated, and keep human review in the workflow.
Project planning also considers reproducibility, data sensitivity, and whether a locally hosted model or an approved third-party service is appropriate. These decisions are guided by the project’s scientific goals, data requirements, and privacy and security needs.
Have an AI Project in Mind?
Whether you have a specific idea or a research problem that may benefit from AI or machine learning, we can help evaluate the possibilities and identify an appropriate path forward.