Ranking of multidimensional drug profiling data by fractional-adjusted bi-partitional scores

Area of Science:

• Computational Biology
• Bioinformatics
• Drug Discovery

Background:

• High-throughput drug profiling (HCS) generates vast multidimensional data, posing challenges for drug effectiveness ranking.
• Traditional methods for ranking drug performance are often inadequate for large-scale HCS data analysis.

Purpose of the Study:

• Introduce a novel framework, fractional adjusted bi-partitional score (FABS), for automated drug performance ordering.
• Address the limitations of existing methods in ranking drug effectiveness from complex HCS datasets.

Main Methods:

• Developed the FABS framework utilizing graph-based formulations.
• Implemented FABS with a normalized cut variant (FABS-NC(')), designed for polynomial time complexity and scalability.
• Compared FABS-NC(') against FABS-SVM, FABS-Spectral, Center Ranking, PCA Ranking, and GTEM.

Main Results:

• FABS-NC(') demonstrated superior performance in ranking drug effectiveness compared to all five alternative methods.
• FABS-SVM achieved the second-best performance but significantly lagged behind FABS-NC(').
• FABS-NC(') showed a substantial improvement in correctly predicted ranking trials over FABS-SVM.

Conclusions:

• The FABS framework, particularly FABS-NC('), offers a robust and scalable solution for automated drug ranking in high-throughput screening.
• This novel approach enhances the efficiency and accuracy of drug discovery pipelines.
• The proposed method provides a significant advancement over conventional drug ranking techniques.