Large Language Model Agents for Evidence Based Genetic Disease Severity Classification
Abstract
Disease severity classification for genetic conditions is subjective and labor-intensive, creating bottlenecks in genomic screening, where commercial panels vary widely in size and overlap. We developed an autonomous AI agent integrating Reasoning and Acting (ReAct) with Retrieval-Augmented Generation (RAG) to classify 10,211 Human Phenotype Ontology terms. It uses American College of Medical Genetics (ACMG)-endorsed severity guidelines and American College of Obstetricians and Gynecologists (ACOG) quality-of-life criteria to retrieve PubMed literature, generate interpretable reasoning chains, and independently verify claims. At the phenotype level, using expert-curated cohorts, the agent achieved 93.55% accuracy (MCC 0.9237) with 82.6% to 91.4% of claims supported by direct evidence or valid inferences. Gene-level severity was aggregated across 8,738 pairs, identifying 3,283 autosomal recessive pairs with severe or profound presentations. External validation showed 95.2% concordance with Mackenzie's Mission gene list. This system enables standardized panel design by providing reliable, automated classification supported by direct evidence.
Related
- Explainable plant disease classification and severity assessment using MobileNetV3-large for crop health monitoring
- MEMOIR-VLM—a multimodal vision-language model for Alzheimer's disease classification and question answering
- Leakage-Aware Ablation of Data Augmentation for Sugarcane Leaf Disease Classification: Operator- and Severity-Specific Effects
- Mapping human genetic associations to drug targets in valvular heart disease: an integrative evidence synthesis
- SeverYX: An ensemble framework for robust variant severity classification
BibTeX
@misc{ghasemnejad2026large,
title = {{Large Language Model Agents for Evidence Based Genetic Disease Severity Classification}},
author = {Tohid Ghasemnejad and Ahmadreza Argha and Mark Grosser and John Wang and Min Yang and Thantrira Porntaveetus and Tony Roscioli and Nigel H. Lovell and Mahmoud Aarabi and Hamid Alinejad-Rokny},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.19569},
eprint = {2609.19569},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.19569},
note = {CC BY-SA 4.0},
}
RIS
TY - UNPB
TI - Large Language Model Agents for Evidence Based Genetic Disease Severity Classification
AU - Tohid Ghasemnejad
AU - Ahmadreza Argha
AU - Mark Grosser
AU - John Wang
AU - Min Yang
AU - Thantrira Porntaveetus
AU - Tony Roscioli
AU - Nigel H. Lovell
AU - Mahmoud Aarabi
AU - Hamid Alinejad-Rokny
PY - 2026
DA - 2026/09/18
PB - arXiv
DO - 10.48550/arxiv.2609.19569
UR - https://arxiv.org/abs/2609.19569
AB - Disease severity classification for genetic conditions is subjective and labor-intensive, creating bottlenecks in genomic screening, where commercial panels vary widely in size and overlap. We developed an autonomous AI agent integrating Reasoning and Acting (ReAct) with Retrieval-Augmented Generation (RAG) to classify 10,211 Human Phenotype Ontology terms. It uses American College of Medical Genetics (ACMG)-endorsed severity guidelines and American College of Obstetricians and Gynecologists (ACOG) quality-of-life criteria to retrieve PubMed literature, generate interpretable reasoning chains, and independently verify claims. At the phenotype level, using expert-curated cohorts, the agent achieved 93.55% accuracy (MCC 0.9237) with 82.6% to 91.4% of claims supported by direct evidence or valid inferences. Gene-level severity was aggregated across 8,738 pairs, identifying 3,283 autosomal recessive pairs with severe or profound presentations. External validation showed 95.2% concordance with Mackenzie's Mission gene list. This system enables standardized panel design by providing reliable, automated classification supported by direct evidence.
ER -