IVA, a locally deployable Intelligent Variant Analyser based on a small language model for complete case-level and trio-based genomic interpretation.
- 1CHUV
Abstract
BackgroundRecent work has brought large language models (LLMs) into clinical variant interpretation. In most existing approaches, the model is confined to text-level operations, retrieving, extracting, compacting, and summarizing heterogeneous evidence, while the genuine genetic evaluation is delegated to a separate deterministic engine. Other tools employ LLMs to mine functional evidence for a single ACMG criterion, to rerank a classifiers output, or to answer isolated variant questions. Integrative genetic reasoning, by contrast, demands multi-source evidence weighing for each variant, reconciliation of inheritance with observed zygosity and parental genotypes, joint interpretation of variants in the same gene, and the derivation of a case-level conclusion. Moreover, to protect patient privacy, this must be done on hospital premises.
Methodso build an end-to-end variant interpreter that runs on local infrastructure, we benchmarked 13 small language models (SLMs) on 150 multiple-choice genetics questions and selected Qwen3.5-9B. In IVA (Intelligent Variant Analyser), a four-stage pipeline, the model assigns ACMG/AMP criteria, applies de novo, dominant, recessive, and X-linked inheritance logic to trios, and writes a structured report, while deterministic code handles retrieval, phasing, point summation, citation checking, and formatting. We evaluated IVA on synthetic scenarios, 16 solved trios, and 327 solved singleton cases of visual impairment, on a single GPU.
ResultsOn the benchmark, Qwen3.5-9B reached a mean accuracy of 0.81, compared with 0.89 for a 1.6-trillion-parameter cloud model. In the 16 trios, IVA reported every previously identified pathogenic, likely pathogenic, or uncertain-significance variant of interest (sensitivity 100%; precision 39.5%). In the 327 singleton cases, it recovered 93% of the expert-reported causative variants and classified 82% of them as pathogenic or likely pathogenic (precision 43%). Failures fell into four recurrent classes: omission, unverified mechanism, mismatch between reasoning and label, and arithmetic errors.
ConclusionsA small language model running on local hardware can serve as the reasoning layer for case-level and trio-based variant interpretation when it is surrounded by deterministic safeguards. Because its outputs are not deterministic, analyst review and prospective multicenter validation remain necessary.
DescriptionIVA is a locally deployable pipeline in which a sub-10B small language model assigns ACMG/AMP criteria, reasons over free text phenotype descriptions, familial inheritance, and writes the clinical report for a complete candidate-variant table, with all model inference running on hospital hardware.
Related
- Open-source offline-deployable retrieval-augmented large language model for assisting pancreatic cancer staging
- Automated Interpretation of EEG Reports Using a Large Language Model with Structured Confidence Outputs
- A clinician aligned vision language framework for stepwise interpretation in fundus fluorescein angiography
- Bridging trust and performance in intelligent systems: Hybrid explainable AI approaches for interpreting large language models
- Toward trustworthy diagnostic AI: debiasing and grounding large language models for seizure semiology interpretation
BibTeX
@misc{santoni2026locally,
title = {{IVA, a locally deployable Intelligent Variant Analyser based on a small language model for complete case-level and trio-based genomic interpretation.}},
author = {F. Santoni and E. Ducret and S. Oreshkov and R. Abdur and H. Frederiksen and M. Ullah and a. laverre and r. forey and c. Le caignec and M. Ansar},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.07.26364945},
url = {https://www.medrxiv.org/content/10.64898/2026.10.07.26364945v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - IVA, a locally deployable Intelligent Variant Analyser based on a small language model for complete case-level and trio-based genomic interpretation.
AU - F. Santoni
AU - E. Ducret
AU - S. Oreshkov
AU - R. Abdur
AU - H. Frederiksen
AU - M. Ullah
AU - a. laverre
AU - r. forey
AU - c. Le caignec
AU - M. Ansar
PY - 2026
DA - 2026/10/07
PB - medRxiv
DO - 10.64898/2026.10.07.26364945
UR - https://www.medrxiv.org/content/10.64898/2026.10.07.26364945v1
AB - BackgroundRecent work has brought large language models (LLMs) into clinical variant interpretation. In most existing approaches, the model is confined to text-level operations, retrieving, extracting, compacting, and summarizing heterogeneous evidence, while the genuine genetic evaluation is delegated to a separate deterministic engine. Other tools employ LLMs to mine functional evidence for a single ACMG criterion, to rerank a classifiers output, or to answer isolated variant questions. Integrative genetic reasoning, by contrast, demands multi-source evidence weighing for each variant, reconciliation of inheritance with observed zygosity and parental genotypes, joint interpretation of variants in the same gene, and the derivation of a case-level conclusion. Moreover, to protect patient privacy, this must be done on hospital premises. Methodso build an end-to-end variant interpreter that runs on local infrastructure, we benchmarked 13 small language models (SLMs) on 150 multiple-choice genetics questions and selected Qwen3.5-9B. In IVA (Intelligent Variant Analyser), a four-stage pipeline, the model assigns ACMG/AMP criteria, applies de novo, dominant, recessive, and X-linked inheritance logic to trios, and writes a structured report, while deterministic code handles retrieval, phasing, point summation, citation checking, and formatting. We evaluated IVA on synthetic scenarios, 16 solved trios, and 327 solved singleton cases of visual impairment, on a single GPU. ResultsOn the benchmark, Qwen3.5-9B reached a mean accuracy of 0.81, compared with 0.89 for a 1.6-trillion-parameter cloud model. In the 16 trios, IVA reported every previously identified pathogenic, likely pathogenic, or uncertain-significance variant of interest (sensitivity 100%; precision 39.5%). In the 327 singleton cases, it recovered 93% of the expert-reported causative variants and classified 82% of them as pathogenic or likely pathogenic (precision 43%). Failures fell into four recurrent classes: omission, unverified mechanism, mismatch between reasoning and label, and arithmetic errors. ConclusionsA small language model running on local hardware can serve as the reasoning layer for case-level and trio-based variant interpretation when it is surrounded by deterministic safeguards. Because its outputs are not deterministic, analyst review and prospective multicenter validation remain necessary. DescriptionIVA is a locally deployable pipeline in which a sub-10B small language model assigns ACMG/AMP criteria, reasons over free text phenotype descriptions, familial inheritance, and writes the clinical report for a complete candidate-variant table, with all model inference running on hospital hardware.
ER -