Decoding coccidioides: an LLM AI approach to genomic, phenotypic, and clinical characterization of 200 clinical isolates - introduction: valley fever is a fungal disease estimated to cause > 200,000…

The National Institute of Allergy and Infectious Diseases awarded $220,796 to Mayo Clinic Arizona on September 1, 2026, under the Allergy and Infectious Diseases Research program (CFDA 93.855) to develop diagnostic and prognostic markers for coccidioidomycosis using genomic analysis and large language model-driven clinical data abstraction.

The award funds three specific aims: sequencing 200 clinical isolates of Coccidioides species and conducting genome-wide association studies in the context of clinical disease features; abstracting longitudinal patient disease course data (symptoms, laboratory results, radiology) using a Complex Longitudinal Abstraction Tool; and analyzing the integrated genomic and clinical data using large language models to identify links between genomic differences and clinical pathological categories (pulmonary, extrapulmonary, meningitis), disease outcomes (resolution, severe/chronic disease, death), and clinical features (erythema nodosum, fatigue). The research directly supports the NIH Notice of Special Interest (NOT-AI-23-031) on advancing research needed to develop a coccidioidomycosis vaccine by providing broader understanding of the Coccidioides genomic space and improved prognostic markers.

Performance occurs at Mayo Clinic Arizona in Scottsdale, Arizona, with a period of performance from September 1, 2026, through August 31, 2028. This is a project grant, the standard assistance type under the Allergy and Infectious Diseases Research program.

Decoding coccidioides: an LLM AI approach to genomic, phenotypic, and clinical characterization of 200 clinical isolates - introduction: valley fever is a fungal disease estimated to cause > 200,000 infections yearly. The impact of genomic features on clinical disease is largely unexplored. During antifungal therapy, the disease only has one specific marker, antibody titers, which is only semi- quantitative and slowly responsive to clinical changes. Better prognostic markers are needed. Long-term objective: the goal of the laboratory is to develop better diagnostic and prognostic markers of coccidioidomycosis. Data from this study will contribute significant information to the field of valley fever and open new areas for research that will improve human health. Not-AI-23-031 notice of special interest (NOSI): advancing research needed to develop a coccidioidomycosis (valley fever) vaccine. This project directly supports the NIH NOSI for coccidioides by increasing our understanding of clinical data and develop better prognostic markers using our complex longitudinal abstraction tool (CLAT) and our genomic data will inform the potential robustness of vaccine candidates by providing a broader understanding of the coccidioides genomic space. Hypotheses: 1 genomic differences in coccidioides SPP. Are linked to clinical pathological categories (e.g., pulmonary, extrapulmonary, meningitis) and outcome (resolution, severe/chronic disease, death). 2. Deep annotation of longitudinal patient data (symptoms, lab results, and radiology, etc.) may reveal previously unknown links of clinical features (e.g., erythema nodosum, fatigue, etc.) to fungal strain. Specific aims: aim 1: sequence the 200 clinical isolates of coccidioides SPP. And perform genome-wide association studies in context of clinical disease features. Aim 2: using the CLAT, abstract the course of disease for the patients from which the 200 isolates grew. Aim 3: analyze the data using a large language model (LLM) with a model context protocol (MCP). Expected outcomes and impact: we will complete preliminary analyses of the genomes and associations of certain regions/features to clinical disease. The genomes will be deposited into a public database. The association of genomes to highly detailed patient data may reveal how some genomic regions are associated with specific disease presentations. These associations may provide both diagnostic and prognostic insights that could inform pathogenic mechanisms and treatment approaches. The analysis using the LLM and MCP models will provide for deep and iterative inquiry into our complex dataset, maximizing the conclusions that can be drawn from the data.