Microfluidic devices and methods of designing and using microfluidic devices

A method, comprising:

• receiving a molecular compound in a microfluidic device comprising a plurality of organ constructs and a fluid loop configured to recirculate a fluid media through the plurality of organ constructs;
• recirculating the molecular compound through the microfluidic device at a specified flow rate to cause the molecular compound to interact with the plurality of organ constructs, the flow rate based on a type of at least one of the organ constructs;
• measuring one or more concentration values of the molecular compound at one or more of the organ constructs based on the interacting;
• determining, by a computing device associated with the microfluidic device, based at least partially on the measuring, a plurality of concentration profiles for the molecular compound, each concentration profile of the plurality being associated with a respective organ construct of the microfluidic device;
• determining, by the computing device and based at least partially on the plurality of concentration profiles, a plurality of first pharmacokinetic (PK) parameters comprising a first type;
• translating, by the computing device and based on the type of at least one of the organ constructs, the plurality of first PK parameters to a plurality of second PK parameters comprising a second type, the translating comprising executing a quantitative systems pharmacology (QSP) model that includes an empirical scaling factor (ESF) to scale in vitro parameters to in vivo parameters, the ESF being determined based on a machine learning model trained with a data set representing molecular properties of molecular compounds, the molecular properties including a chemical structure or physiochemical properties of the molecular compounds, in vitro PK parameter profiles corresponding to the molecular compounds in the microfluidic device, and in vivo PK parameter profiles of the molecular compounds, and wherein the machine learning model is selected based on known outcomes of previous circulations in the microfluidic device of a molecular compound of a similar molecular structure;
• generating, by the computing device, data comprising a profile comprising the second PK parameters, the profile comprising PK parameters simulating in vivo data;
• and
• outputting, by the computing device, the generated data comprising the profile to a data store.