AURA: Accelerating Drug Discovery with Accuracy, Utility, and Rank-Order Assessment for Data-Driven Decision Making.

Abstract

Biopharmaceutical companies generate a wealth of data, ranging from in silico physicochemical properties and machine learning models to both low and high-throughput in vitro assays and in vivo studies. To effectively harnesses this extensive data, we introduce a statistical methodology facilitated by Accuracy, Utility, and Rank Order Assessment (AURA), which combines basic statistical analyses with dynamic data visualizations to evaluate endpoint effectiveness in predicting intestinal absorption. We demonstrated that various physicochemical properties uniquely influence intestinal absorption on a project-specific basis, considering factors like intestinal efflux, passive permeability, and clearance. Projects within both the "Rule of 5" (Ro5) and beyond "Rule of 5" (bRo5) space present unique absorption challenges, emphasizing the need for tailored optimization strategies over one-size-fits-all approaches. This is corroborated by the improved accuracy of project-specific correlations over global models. The differences in correlations between and within project teams-due to their unique chemical spaces-highlight how complex and nuanced the prediction of intestinal absorption can be. Here, we implement a standardized methodology, AURA, that any organization can incorporate into their workflow to enhance early-stage drug optimization. By automating analytics, integrating diverse data types, and offering flexible visualizations, AURA enables cross-functional teams to make data-driven decisions, optimize workflows, and enhance research efficiency.