Work

Discrimination of ranges of closely related RNA targets using CRISPR-based detection assay developed using machine learning-based optimization

Year: 2026
Type: article
Abstract: Early detection and identification of pathogen strains are crucial for mitigating infectious disease threats, including those posed by closely related zoonotic variants. A key limitation of current molecular detection approaches is their narrow focus on individual targets, which can delay recognition of outbreaks caused by closely related but non-identical pathogens. Here we report a two-channel CRISPR/Cas13a detection assay capable of discriminating between SARS-CoV-2 and a panel of twelve closely related zoonotic sarbecoviruses using a compact duplex assay format. Guide RNA sets with tunable specificity were designed using a two-stage machine learning pipeline: initial candidate filtering with a pre-existing RuleFit model, followed by retraining on 382 sarbecovirus-specific in vitro data points to refine the mismatch threshold from 4 to 6 mismatches for target exclusion. The optimized assay achieved proof-of-concept three-way classification of synthetic RNA targets into: (1) SARS-CoV-2, (2) the closest known near-neighbor BANAL-236, and (3) remaining zoonotic near-neighbors, with a limit of detection of 15–150 fM for synthetic RNA targets. This approach provides a practical framework for surveillance-oriented detection that identifies not only a specific pathogen but also its closest relatives, provided that system-specific validation and model retraining are included for new target systems.
Language: English
Field: Medicine
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W4413124828
FWCI:
Cites: 35
Cited by:
Related to: 9