Use of recursion forests in the sequential screening process: consensus selection by multiple recursion trees

Area of Science:

• * Cheminformatics
• * Computational Chemistry
• * Bioinformatics

Background:

• * High-Throughput Screening (HTS) generates large datasets requiring robust modeling.
• * Recursive Partitioning models accommodate data imperfections but exhibit high false positives in prediction.
• * Sparse HTS datasets exacerbate prediction challenges.

Purpose of the Study:

• * Introduce Consensus Selection to mitigate false positives in Recursive Partitioning models.
• * Enhance the reliability and efficiency of HTS data analysis.
• * Improve the hit rate and predictive accuracy for drug discovery pipelines.

Main Methods:

• * Developed Consensus Selection, a novel procedure for Recursive Partitioning model enhancement.
• * Employed Multiple Recursion Trees within the Consensus Selection framework.
• * Validated the method on sparse High-Throughput Screening datasets.

Main Results:

• * Consensus Selection significantly reduced the false positive rate compared to single Recursion Tree models.
• * Achieved over 30-fold increase in hit rate for High-Throughput Screens.
• * Demonstrated improved explanatory and predictive power for sparse HTS data.

Conclusions:

• * Consensus Selection offers a robust solution for analyzing sparse HTS data.
• * This method enhances the efficiency and accuracy of identifying potential drug candidates.
• * The approach effectively balances model interpretability with predictive performance.