Stoichiometric Encoding of Amplification Kinetics Enables High-Level Multiplexing in Single-Channel TaqMan Real-Time PCR

Multiplexing in TaqMan real-time PCR (qPCR) is limited to four-six targets per reaction by fluorophore spectral overlap. Amplification curve analysis (ACA) can discriminate several targets within one fluorescence channel from amplification curves (AC) kinetic features, but has been confined mainly to digital qPCR (dqPCR) and is sensitive to domain shifts between instruments and specimen types. Here, we co-design biochemical encoding of target-specific information into ACs and their computational decoding to achieve single-channel 8-plex TaqMan qPCR on a standard thermocycler. Using asymmetric PCR, we show that forward primer, reverse primer and probe concentrations act as a programmable encoding mechanism, generating a 27-condition design space of reproducible non-canonical ACs, conserved from digital to bulk qPCR. Applying this stoichiometric encoding to an 8-plex central nervous system pathogens panel increased inter-target separability, raising the mean silhouette score from 0.329 to 0.711. To decode ACs without labelled target-domain bulk qPCR data, we pretrained a transformer encoder by self-supervised learning on more than 170,000 unlabelled dqPCR curves and aligned source and target domains using conditional domain adversarial training. On contrived clinical specimens, the framework classified a canonical 7-plex respiratory virus panel with 83.6% accuracy (versus 67.7% for a supervised baseline) and the encoded 8-plex panel with 89.8% accuracy. Furthermore, stoichiometric encoding enabled resolution of double- and triple-target co-amplification events with 98.7% accuracy in controlled dqPCR experiments. This approach expands the information capacity of existing qPCR instrumentation without new hardware, fluorophores or oligonucleotides, offering a low-cost route to syndromic molecular diagnostics.