Robust Analysis of High Throughput Screening (HTS) Assay Data.

Area of Science:

• Toxicology
• Computational Biology
• Statistical Modeling

Background:

• Quantitative high throughput screening (qHTS) assays are crucial for rapid compound evaluation.
• Statistical analysis of qHTS data often relies on nonlinear regression models, but variance structures (homoscedasticity vs. heteroscedasticity) and outliers complicate accurate toxicity assessments.
• Manual residual analysis for variance structure is infeasible for thousands of compounds screened simultaneously.

Purpose of the Study:

• To develop and evaluate a novel preliminary test estimation (PTE) based methodology for statistical inference in qHTS assays.
• To create a method that is robust to varying variance structures and potential outliers/influential observations.
• To provide a practically useful and easy-to-interpret approach for toxicity decisions in qHTS.

Main Methods:

• A preliminary test estimation (PTE) based statistical methodology was developed.
• The performance of the PTE method was evaluated using simulation studies mimicking real qHTS data.
• Key performance metrics included false discovery rate (FDR) and statistical power.
• The proposed method was compared against two existing methods used in qHTS data analysis.
• The methodology was illustrated using a dataset from the National Toxicology Program (NTP).

Main Results:

• The proposed PTE-based methodology demonstrated robust control of the false discovery rate (FDR) while maintaining good statistical power.
• Simulation studies indicated that one existing method was overly conservative (low power), while the other was too liberal.
• The PTE method offered a superior balance of FDR control and power compared to existing approaches.
• The method proved effective in handling potential outliers and influential observations common in qHTS data.

Conclusions:

• The developed PTE-based methodology provides a reliable and robust approach for statistical analysis in qHTS assays.
• This method improves upon existing techniques by offering better control over FDR and power, even with complex data structures.
• The PTE method is a valuable tool for accurate toxicity assessment and decision-making in large-scale compound screening.
• The supplementary materials provide code and data for reproducibility and further application.