Back

Combining Scalable Organ Chip Platform with Deep Learning-Based Imaging Analysis for Cancer Therapeutic Screening

Yuan, Y.-C.; Xu, B.; McCormack, J.; Huang, X.; Ma, J.; Marshall, T.; Sun, Y.; Singh, H.; Fanelli, A.; Gilbride, P.; Qian, X.; Xie, Z.; Si, L.; Xie, X.; Bai, H.

2024-06-12 bioengineering
10.1101/2024.06.10.598272 bioRxiv
Show abstract

Functional precision oncology represents an emerging approach to genomic approaches by testing treatment options directly on patient-derived models. Current assays, including the use of patient-derived xenograft (PDX) and patient-derived organoid (PDOs), faces major barriers in clinical use due to technical challenges, such as standardization, cost, assay time, scalability, and faithful mimicry of patient tumor microenvironment (TME). Here, we introduce an Organ Chip (OC) device constructed entirely from thermoplastic materials, free of porous membrane or other barrier structures, and optimized for high-content imaging (HCI). This automation-compatible device supports tissue-specific extracellular matrices and coculture for a wide spectrum of organ and disease types, including the TME. As a proof-of-concept, we demonstrate the growth of pancreatic, lung, and colon cancer cell lines and primary lung cancer cells and the testing of cancer drugs in these models. HCI-based phenotypic profiling enabled accurate quantification of drug response, with better performance than traditional biochemical assays. Moreover, we developed a deep-learning method for assessing drug responses using bright field images. The integration of a low-cost, scalable, and faithful OC models with automatic high-content image analysis represents a significant stride towards functional precision oncology and cancer drug discovery.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.