OptiRanker: an Advanced Tool for Simulation and Comparative Analysis of Drug Prioritization Algorithms in in-Vivo Trials

Abstract

A critical challenge in personalized medicine is identifying the optimal drug treatment for individual patients based on their unique biological profiles. While recent advancements have led to a surge in drug prioritization algorithms utilizing patient omics data, including regression, classification, multiple kernel learning, deep learning, and AI-based methods, objective validation remains essential to discern their effectiveness. This study introduces" OptiRanker," an innovative statistical framework designed to simulate and optimize the in vivo validation of these drug prioritization algorithms by ranking their predictive accuracy across varied conditions. OptiRanker not only mimics algorithmic predictions but also integrates them into ranked datasets, enabling the optimal selection of test subjects and drugs for trial setups. By providing a rigorous evaluation subset for ranking and comparison of high- and low-performing algorithms, OptiRanker facilitates the design of cost-effective in vivo trials tailored to efficiently assess predictive capability. This addresses a key bottleneck in personalized medicine: achieving a robust, scalable validation of computational models that can accelerate the clinical translation of promising algorithms. The efficient ranking provided by OptiRanker has the potential to advance personalized treatment strategies, ultimately leading to more effective, individualized patient care. The Python code for OptiRanker is available at the following GitHub repository: https://github.com/OhadLandau/OptiRanker.