Cancer Drug Screening Scale-up: Tumor-microenvironment-mimicking Microfluidics and Deep Learning Image Analysis

Abstract:

Tumor is a vastly heterogeneous population of cancer cells with diverse chemo-drug resistance. In order to evaluate drug resistance, drug screening assays are widely used to test a collection of candidate compounds in preclinical trials. In addition, accumulating evidence suggests that cancer drug response can be regulated by tumor microenvironment (TME). Due to the complexity in cancer cell heterogeneity and the growing drug library, there is a need to scale-up the cancer drug screening system recapitulating local TME. Here I present two microfluidic platforms for high-throughput drug screening: Firstly, I developed a 3D gel-island chip containing 1,500 isolated cell chambers with extracellular matrix. I demonstrated that TME is favorable for cancer cells to maintain stemness and develop drug resistance. In addition, drug combination screening chip provided scalable screening solution for all pairwise drug combinations from 8 different drugs, using a 3D spheroid model. In order to process the massive image data generated from scale-up experimental tools, I also developed an augmented deep learning model to automatically assess drug efficacy at high accuracy (94.7%), and estimate half-maximal inhibitory concentration with <8% error. In summary, the integration of TME-mimicking microfluidic chip and smart data analysis facilitates the development of high-throughput cancer drug screening.

Chair: Professor Euisik Yoon

Remote Access: https://bluejeans.com/566743594