Improving detection thresholds and robust event filtering in single-particle and single-cell ICP-MS analysis

Abstract

In this work, a modular data processing workflow for single-particle (sp) and single-cell (sc) inductively coupled plasma-mass spectrometry (ICP-MS) is presented. To achieve more reliable detection thresholds, a special focus is placed on the parameter estimation of Gaussian and Poisson distributions that describe the background (BG) signal. For Gaussian models, the widely used iterative outlier test was improved by an algorithm that adjusts the test for different BG levels by incrementing the test factor. Through careful evaluation, the standard deviation of the experimental sc- and sp-ICP-MS data was applied as a robust measure of the convergence quality of the test. In addition, the outlier analysis was separated from the subsequent event detection more strictly than it is often reported. Importantly, a data-dependent decision criterion based on Gaussian and Poisson modeling was developed to effectively address extra-Poisson variance in experimental data. In the second part, a gate filter was developed to reduce the amount of excess false-positive events in sc-ICP-MS. To that end, a secondary filter based on the signal peak height is used to remove rare false-positive events without affecting the signal intensity of the events that are detected correctly. Two approaches, based on a numerical approximation via the detection limit, and critical values of the Gaussian and Poisson distribution are presented to calculate the gate filter level. Possible sources of false-positive events, some of which are specific to sc-ICP-MS, are discussed. The combined processing workflow was applied to analyze the distribution of six endogenous elements in Chlamydomonas reinhardtii cell populations. The gate filter corrected the cell number concentration by up to 44% (22% on average), and mass per cell by up to 30% (17% on average).