Predicting cell properties with AI from 3D imaging flow cytometer data

Predicting the properties of tissues or organisms from the genomics data is widely accepted by the medical community. Here we ask a question: can we predict the properties of each individual cell? Single-cell genomics does not work because the RNA sequencing process destroys the cell, not allowing us to verify our predictions. To test the hypothesis, we investigate the approach of using AI to analyze single-cell images obtained from a 3D imaging flow cytometer. We analyze the cell image at day zero and make the AI-assisted cell property prediction. The prediction is then examined later when the cells continue to live and develop. Our preliminary results are promising, showing 88% accuracy in predicting cells that will have a high protein expression level. The technique can have strong ramifications and impact on preventive medicine, drug development, cell therapy, and fundamental biomedical research.