The invention is a platform that combines a screening system and machine learning algorithms to investigate and report the cardio-activity related information of a certain compound. Through screening cardiac tissue strips, the platform determines whether a compound is cardio-active or not, as well as the associated cardio-active mechanism based on a drug library that is automatically developed. Such information is crucial for the drug development process, especially for evidence based decisions.
While drug screening platforms may vary in their acquisition methodology, the readout of such systems are predominantly composed of an array of parameters that describe the behavior or shape of individual contractile events. Combining this with the number of experimental conditions (e.g. various pacing frequencies or drug concentrations) can yield high-dimensional datasets that make it difficult to draw comprehensive conclusions. Such requirements indicate that traditional methods of pre-selecting one or a few parameters for statistical analysis may not be adequate. By selectively examining a few parameters independently of one another, there is a risk of not detecting important information that differentiates the behavior of normal human pluripotent stem cells derived cardiomyocytes, from those exposed to cardioactive compounds.
Inventors at UCI developed a novel and accurate platform that combines drug screening methods and machine learning algorithms to screen cardiac tissues and determine if a compound is cardio toxic/cardio-active and if yes, to what degree. Observing several force tracing relevant parameters, the platform accurately predicts the compound’s mechanism of cardio-activity and reports drug response relationships between compounds. The proposed platform can be a major cornerstone in the drug development industry.
-Drug development industry
-Compounds’ cardio-activity monitoring
· Able to report the cardio-activity related information of a compound
· Utilizes machine learning algorithms for examining data simultaneously
· Observes a large number of parameters, leading to better sensitivity and specificity, relating to cardio-activity
· Can be applied to platforms with various tissue geometries (e.g. monolayer or 3D tissue constructs) and different monitoring methodologies
· The drug classification library is established in an efficient and automated manner.
· The software is flexible and can be adapted to multiple types of readouts
|United States Of America||Published Application||20180372724||12/27/2018||2017-812|
A 4 drug class library has been established. It successfully predicted 4 ‘unknown’ compounds to its respective class.