ACA BYOL qPCR datasets and model checkpoints

<p dir="ltr">Associated manuscript</p><p dir="ltr">Stoichiometric encoding of amplification kinetics enables high-level multiplexing in single-channel TaqMan real-time PCR</p><p dir="ltr">Louis Kreitmann, Kenny Malpartida-Cardenas, Ye Mao, Zexuan Zhao, San Chun Hin, Anirudhha Hazarika, Ke Xu, Luca Miglietta, Zara Breese, Alison H. Holmes, Karen Brengel-Pesce, Laurent Drazek, and Jesus Rodriguez-Manzano.</p><p dir="ltr">This repository contains the processed amplification curves (ACs) datasets and trained model checkpoints supporting the analyses reported in the manuscript above.</p><p dir="ltr">The accompanying analysis and model-training code is available at:</p><p dir="ltr">github.com/lkreitmann-bmx/ACA_BYOL_qPCR</p><p>---</p><p dir="ltr">Overview</p><p dir="ltr">The study investigates amplification curve analysis (ACA) as a strategy for increasing the multiplexing capacity of TaqMan real-time PCR (qPCR).

It combines:</p><p dir="ltr">Stoichiometric feature encoding, in which forward primer, reverse primer, and probe concentrations are varied to generate target-specific AC morphologies.</p><p dir="ltr">Self-supervised representation learning using Bootstrap Your Own Latent (BYOL) on digital qPCR (dqPCR) amplification curves.</p><p dir="ltr">Conditional domain adversarial training using a transformer-based CDAN (T-CDAN) to improve transfer across experimental domains.</p><p dir="ltr">Co-amplification analysis to determine whether double- and triple-target amplification events remain distinguishable.</p><p dir="ltr">The deposit is intended to provide the datasets required to reproduce the main analyses and the trained neural-network checkpoints used in the manuscript.