Matrix Complexity Evaluation and its effect on Nontargeted detection of Small-Molecule Toxicants by LC-HRMS

Lilach Yishai Aviram, Analytical Chemistry Dep, IIBR, Ness Ziona, Israel (LILACHY@iibr.gov.il)
Dana Marder, Analytical Chemistry Dep, Iibr, Ness Ziona, Israel
Nitzan Tzanani, Analytical Chemistry Dep, Iibr, Ness Ziona, Israel
Eyal Drug, Analytical Chemistry Dep, Iibr, Ness Ziona, Israel
Hagit Prihed, Analytical Chemistry Dep, Iibr, Ness Ziona, Israel
Shai Dagan, Analytical Chemistry Dep, Iibr, Ness Ziona, Israel

Complex mixtures, characterized by high density of compounds, pose a challenge on trace detection and identification especially in "Tox-Screen" analysis. This is further exacerbated under nontargeted analysis, where a compound of interest may be well hidden under thousands of matrix compounds, that are all detectable by modern LC-MS(/MS) technologies.

Here we present our study on the effect of matrix complexity on nontargeted detection (peak picking) following LC–MS/MS (Orbitrap) analysis. A series of ∼20 drugs, V-type chemical warfare agents (CWAs) and pesticides, representing toxic “unknowns”, were spiked (at various concentrations) in several complex matrices including rosemary leaves and soil extracts as well as diluted urine. The Orbitrap “TraceFinder” software was incorporated to explore their peak intensities, in relation to the matrix (peak location in an intensity-sorted list). Average practical detection limits of nontargets were elucidated: While detection among the first 10,000 peaks was achieved at 0.3–1 ng/mL levels in the extract, for the more realistic “top 1000” list, much higher concentrations were required, approaching 10–30 ng/mL. A negative power law functional dependence between the peak tabular location in an intensity-sorted suspect list and the nontarget concentration, is proposed. Controlled complexity was also explored with a series of urine dilutions, resulting in an excellent correlation between the power law coefficient and dilution factor. The intensity distribution of matrix peaks was found to spread (unevenly) on a broad range, fitting well the Weibull distribution function with all matrices and extracts. The above empirically-based quantitative approach gives a measure of the actual capabilities and limitations of LC–MS in the analysis of nontargets in complex matrices. It may be used to estimate and compare the complexity of matrices and predict the typical detection limits of unknowns by LC-MS.