Machine Learning-Driven\nData Valuation for Optimizing\nHigh-Throughput Screening Pipelines

In the rapidly evolving field of drug discovery, high-throughput\nscreening (HTS) is essential for identifying bioactive compounds.\nThis study introduces a novel application of data valuation, a concept\nfor evaluating the importance of data points based on their impact,\nto enhance drug discovery pipelines. Our approach improves active\nlearning for compound library screening, robustly identifies true\nand false positives in HTS data, and identifies important inactive\nsamples in an imbalanced HTS training, all while accounting for computational\nefficiency. We demonstrate that importance-based methods enable more\neffective batch screening, reducing the need for extensive HTS. Machine\nlearning models accurately differentiate true biological activity\nfrom assay artifacts, streamlining the drug discovery process. Additionally,\nimportance undersampling aids in HTS data set balancing, improving\nmachine learning performance without omitting crucial inactive samples.\nThese advancements could significantly enhance the efficiency and\naccuracy of drug development.