DeepScreen: An Accurate, Rapid, and Anti-Interference Screening Approach for Nanoformulated Medication by Deep Learning

Abstract

Accuracy of current efficacy judgment methods for nanoformulated drug remains unstable due to the interference of nanocarriers. Herein, DeepScreen, a drug screening system utilizing convolutional neural network based on flow cytomerty single-cell images, is introduced. Compared to existing experimental approaches, the high-throughput system has superior precision, rapidity, and anti-interference, and is cost-cutting with high accuracy. First, it can resist most disturbances from manual factors of complicated evaluation progress. In addition, class activation maps generated from DeepScreen indicate that it may identify and locate the tiny variation from cell apoptosis and slight changes of cellular period caused by drug or even nanoformulated drug action at very early stages. More importantly, the excellent performance of assessment on two types of nanoformulations and fluorescent drug proves the fine generality and anti-interference of this novel system. All these privileged performances make DeepScreen a very smart and promising system for drug detection.