Examining the effects of analytical replication on data quality in a non-targeted analysis experiment

Area of Science:

• Environmental chemistry and analytical science.
• Development of computational methods for chemical analysis.

Background:

• Non-targeted analysis (NTA) expands chemical detection in environmental monitoring but generates massive datasets.
• Manual review of NTA data is infeasible; computational tools aid processing but struggle with signal vs. artifact.
• Replicate analysis improves data reliability but is costly and becomes untenable in large NTA studies.

Purpose of the Study:

• To investigate the information penalties associated with reduced analytical replication in non-targeted analysis.
• To quantify the impact of replication levels on false discovery rates (FDR) and false negative rates (FNR).
• To explore differences in information penalties across various chemical feature groups, including per- and polyfluoroalkyl substances (PFAS).

Main Methods:

• Conducted over 70,000 simulations using an existing NTA dataset with variable replication designs.
• Calculated FDR and FNR for NTA features and occurrences under different replication scenarios.
• Employed regression models to analyze the relationship between replication percentage and FDR/FNR, and to assess feature attribute effects.

Main Results:

• Observed inverse relationships between replication percentage and FDR/FNR, indicating higher information penalties with reduced replication.
• Demonstrated significant increases in FDR/FNR for suspected per- and polyfluoroalkyl substances (PFAS) compared to non-PFAS compounds.
• Highlighted that quantitative information penalties are study-specific, influenced by sample type and workflow.

Conclusions:

• Reduced analytical replication in NTA studies leads to quantifiable information penalties, increasing error rates.
• Specific chemical classes, such as PFAS, may experience disproportionately higher information penalties.
• The presented methods can guide pilot studies to determine optimal replication strategies for large-scale NTA environmental monitoring.