Determining Dose-Response Characteristics of Molecular Perturbations in Whole-Organism Assays Using Biological Imaging and Machine Learning

Abstract

Advances in microscopy and high-content imaging now offer a powerful way to profile the phenotypic response of intact systems to molecular perturbation and study the response irrespective of putative target activity and by preserving the physiological context in the living systems. An emerging challenge in bioinformatics and drug discovery is constituted by data generated from such studies that involve analyzing the effect of specific molecules at the system-wide organism level. In this paper we propose a novel automated approach that combines techniques from biological imaging and machine learning to automatically quantify a fundamental measure of molecular perturbation in an intact biological system, namely, its dose-response characteristics. We validate our results using phenotypic assay data involving post-infective larvae (schistosomula) of the parasitic Schistosoma mansoni flatworm. This parasite is one of the etiological agents of schistosomiasis -a significant neglected tropical disease, which puts at-risk nearly two billion people.