PROJECT SUMMARY Some cell signaling systems operate by a mechanism of promiscuous signaling, where multiple ligands can bind to a single receptor before starting a downstream cascade of signaling that results in gene expression. Promiscuous signaling systems present in cells are prevalent in many different types of biological processes from development and maintenance, to disease, including cancer. The bone morphogenetic pathway (BMP) is an ideal promiscuous signaling pathway to study because, of the 10 distinct BMP ligands that act as growth factors, each competitively binds with a type I or type II receptor of the pathway. Recent work created mathematical models of the promiscuous interactions within the BMP pathway that were able to replicate experimental observations of BMP pathway signaling by dosing a BMP-responsive cell line, which expressed YFP when the BMP gene expression was activated, to a 6-fold BMP ligand titration series. However, previous results relied on matrix combination screening of BMP pathway to examine responses and fit a small subset of the parameters of the mathematical models replicating experimental results. Continuing to screen combinations of ligands results in this manner results in an exponential increase in the number of ligand screens required. Better hardware and mathematical tools are needed to screen the BMP pathway to better understand promiscuous signaling phenomena. This project aims to develop a droplet microfluidic device, the DropShop platform, that can screen BMP ligand combinations in a high throughput manner. To do this, an adherent epithelial mammary gland murine BMP-responsive cell line will be adapted to screening by droplet microfluidics through a novel method of cell culture using microcarriers. The droplet microfluidics of the DropShop platform will be optimized to work with the novel cell culture method. Proof of principle of screening of BMP ligands in a certain cell type will be demonstrated in this system by use of a fluorescent measurement system typically used in high throughput droplet microfluidic screening. Finally, machine learning methods will be developed to optimize screening of ligands to reduce the time to determine parameter of the BMP mathematical model, as well as help in selecting the correct model that characterizes experimental results. The resulting system will demonstrate a proof of concept for a droplet microfluidic device capable of automatically determining mechanistic models and their parameters in promiscuous signaling pathways.