机器学习辅助单管 LAMP 熔解分析用于单核苷酸多态性（SNP）基因分型

Abstract

DNA breathing dynamics enable probe-based SNP melting analysis in LAMP amplicons. • Loop-independent probe design generates reproducible melting fingerprints for SNPs. • Machine learning suppresses pseudo-peaks and improves melting curve classification. • High-sensitivity genotyping achieved for human genomic DNA and viral RNA targets. • Portable, lyophilized, one-pot workflow enables point-of-care SNP detection. Rapid and accurate detection of single-nucleotide polymorphisms (SNPs) is essential for personalized medicine and infectious disease surveillance. Although loop-mediated isothermal amplification (LAMP) provides outstanding speed and sensitivity, the highly concatenated and heterogeneous structure of LAMP amplicons has historically limited reliable probe-based melting curve analysis for SNPS discrimination. Here, we present a robust LAMP–Melting platform that exploits the intrinsic “DNA breathing” dynamics of LAMP amplicons to enable allele-specific genotyping without reliance on predefined loop regions or complex strand-displacement probe designs. After isothermal amplification, transient dissociation–rehybridization fluctuations allow fluorescent probes to intermittently access nominally double-stranded target regions, generating distinct and reproducible melting fingerprints for single-base variants. Using this mechanism-driven strategy, we successfully genotyped the human MTHFR C677T polymorphism and the SARS-CoV-2 D614G mutation with high analytical sensitivity, achieving limits of detection of 1 ng/μL for gDNA and 10 RNA copies per reaction. To address the biochemical complexity and signal heterogeneity inherent to isothermal amplification products, a XGBoost–based machine-learning framework was integrated to suppress artifacts and enhance melting curve classification. Moreover, the assay exhibits excellent compatibility with lyophilized reagents and portable fluorescence detection instruments, enabling a one-pot, closed-tube workflow suitable for de