Monte Carlo Wavelet Analysis for Objective Peak Detection in SRM LC-MS/MS Analysis

Detection of low-level analytes in complex chromatographic-mass spectrometric data requires a criterion to discern apparent peaks from the background. Conventional signal-to-noise criteria rely on simple, constant-variance noise models and overlook spurious peaks generated by chemical noise and coeluting interferences. We introduce a wavelet-based Monte Carlo technique for determining the statistical significance of SRM LC-MS/MS peaks in the presence of structured chemical noise. The method empirically characterizes chemical noise peaks in samples and builds a generative noise-only null model. Monte Carlo resampling of the noise model assigns p values that are controlled for the family wise type I error rate (FWER). We validated the method with SRMs from a dilution series of drug compounds in plasma with known ground truth concentrations. Triplicate technical replicates were used, spanning concentrations from far above the limit of detection to far below it. Peaks with adjusted p values of less than 0.05 matched the expectation for true positives above the detection limit. Peaks below the limit of detection matched matrix blanks as true negatives, and intermittent detection in the transition region was observed. An independent external validation using a clinical pain panel confirmed that the method detects ketamine in confirmed positive samples with signal intensity below the lowest calibration standard while correctly classifying matrix blanks and biological negatives. As a demonstration, we applied our method to a recently published lipid mediator data set. By replacing subjective noise region selection with a formal hypothesis test against an empirical null model, the method provides an objective and reproducible criterion for deciding whether peak integration is warranted.