Single-Cell Transcriptomics from Flow Cytometry via Machine Learning

The patent creates a computational copy of scRNAseq data by training a machine learning model on paired scRNAseq and flow cytometry datasets. The model learns to predict transcriptomic profiles from flow cytometry measurements, generating synthetic scRNAseq-like data without performing actual sequencing. This copying approach preserves measurement precision while eliminating the complexity and cost of physical sequencing infrastructure.

The patent replaces the physical mechanical process of RNA sequencing with a computational machine learning prediction system. Instead of using complex sequencing hardware and wet-lab processing, the system uses trained neural networks to transform flow cytometry data into transcriptomic profiles, substituting physical measurement with computational inference.

The patent performs preliminary action by pre-training the machine learning model on a comprehensive dataset of paired scRNAseq and flow cytometry measurements before actual use. This training phase captures the relationship between flow cytometry features and transcriptomic profiles, enabling rapid prediction during deployment without requiring time-consuming sequencing processes for each new sample.

The system generates transcriptomic profiles as computational copies predicted from flow cytometry data, eliminating the need for time-consuming physical sequencing. The ML model instantiates transcriptomic information from surface protein measurements, achieving rapid profile generation while maintaining scientific accuracy.

The patent introduces a machine learning model as an intermediary that bridges flow cytometry measurements and transcriptomic information. The model processes flow cytometry data (surface protein expressions) and transforms it into predicted transcriptomic profiles, mediating between the two measurement modalities to recover lost information while maintaining flow cytometry's productivity advantages.

The system changes the parameter representation by transforming flow cytometry measurements (fluorescence intensity of surface proteins) into transcriptomic parameters (gene expression levels). The ML model learns the mapping between these different parameter spaces, enabling the system to work with flow cytometry's high-throughput data while outputting information in transcriptomic format.

The patent generates transcriptomic profiles as computational copies predicted from flow cytometry data, eliminating the need for expensive physical sequencing infrastructure. The ML model instantiates transcriptomic information from surface protein measurements, achieving scalable deployment across hundreds or thousands of samples without proportionally increasing cost.

The system replaces expensive wet-lab sequencing operations with computational prediction, substituting physical infrastructure costs with software processing. This substitution enables scaling to large sample sizes while maintaining cost-effectiveness, as computational resources are significantly cheaper than sequencing infrastructure.