Multiplex single-cell droplet PCR with machine learning for detection of high-risk human papillomaviruses

High-risk human papillomavirus (HPV) testing can significantly decline the incidence and mortality of cervical cancer. Microfluidic technology provides an effective method for accurate detection of high-risk HPV by utilizing multiplex single-cell droplet polymerase chain reaction (PCR). However, current strategies are limited by low-integration microfluidic chip, complex reagent system, expensive detection equipment and time-consuming droplet identification. Here, we developed a novel multiplex droplet PCR method that directly detected high-risk HPV sequences in single cells. A multiplex microfluidic chip integrating four flow-focusing structures was designed for one-step and parallel droplet preparation. Using single-cell droplet PCR, multi-target sequences were detected simultaneously based on a monochromatic fluorescence signal. We applied machine learning to automatically identify the large populations of single-cell droplets with 97% accuracy. HPV16, 18 and 45 sequences were sensitively detected without cross-contamination in mixed CaSki and Hela cells. The approach enables rapid and reliable detection of multi-target sequences in single cells, making it powerful for investigating cellular heterogeneity related to cancer diagnosis and treatment.