A Modular and Programmable Cas13d Platform for RNA Single Nucleotide Variant Detection

CRISPR-based nucleic acid diagnostics have shown broad potential, yet reliable single-nucleotide variant (SNV) discrimination remains limited by flanking sequence requirements that constrain targetability, and an inherent specificity-sensitivity trade-off where mismatch designs used to suppress wild type recognition often penalize enzymatic activity. Here we develop a scenario-guided Cas13d framework that supports pre-defined operating modes tailored to distinct analytical goals. Leveraging the minimal protospacer flanking site constraints of Cas13d, we first map mismatch-sensitive windows to derive rule-based crRNA designs that improve allelic discrimination. We then restore assay performance through structure-guided engineering of a miniaturized Cas13d scaffold by internally inserting auxiliary RNA binding domains (RBDs). Systematic benchmarking across representative oncology hotspots delineates two practical regimes comprising an ultra-sensitive, amplification-free mode in which a dual-RBD variant paired with optimized mismatched crRNAs achieves ∼0.6% variant allele fraction (VAF) detection, and a robust amplified mode incorporating optional loop-mediated isothermal amplification coupling that favors simpler architectures to balance performance and background across broader low-VAF ranges. In an evaluation of 45 clinical tumor RNA specimens spanning pancreatic, cholangiocarcinoma, and colorectal cancers, the assay correctly classified mutation status with full concordance for KRAS G12D, IDH1 R132C and BRAF V600E, with a subset of positive cases corroborated by orthogonal RT-ddPCR. A prospective IDH1 R132C clinical-matrix spike-in further supported sub-1% detection without pre-amplification. Collectively, this work establishes a configurable Cas13d toolkit and a rule-guided strategy for deploying CRISPR-based RNA SNV diagnostics with application-specific performance objectives.