Parameter Optimization for Feature and Hit Generation in a General Unknown Screening Method—Proof of Concept Study Using a Design of Experiment Approach for a High Resolution Mass Spectrometry Procedure after Data Independent Acquisition

Abstract

High resolution mass spectrometry and modern data independent acquisition (DIA) methods enable the creation of general unknown screening (GUS) procedures. However, even when DIA is used, its potential is far from being exploited, because often, the untargeted acquisition is followed by a targeted search. Applying an actual GUS (including untargeted screening) produces an immense amount of data that must be dealt with. An optimization of the parameters regulating the feature detection and hit generation algorithms of the data processing software could significantly reduce the amount of unnecessary data and thereby the workload. Design of experiment (DoE) approaches allow a simultaneous optimization of multiple parameters. In a first step, parameters are evaluated (crucial or noncrucial). Second, crucial parameters are optimized. The aim in this study was to reduce the number of hits, without missing analytes. The obtained parameter settings from the optimization were compared to the standard settings by analyzing a test set of blood samples spiked with 22 relevant analytes as well as 62 authentic forensic cases. The optimization lead to a marked reduction of workload (12.3 to 1.1% and 3.8 to 1.1% hits for the test set and the authentic cases, respectively) while simultaneously increasing the identification rate (68.2 to 86.4% and 68.8 to 88.1%, respectively). This proof of concept study emphasizes the great potential of DoE approaches to master the data overload resulting from modern data independent acquisition methods used for general unknown screening procedures by optimizing software parameters.

Comparison of data acquisition modes with Orbitrap high‐resolution mass spectrometry for targeted and non‐targeted residue screening in aquacultured eel

“…However, when using a slower model Orbitrap mass analyzer, the methods may only generate 3–5 points across narrow chromatographic peaks that can make it difficult to even determine whether an analyte is present. Several publications have demonstrated the equivalent capabilities using Orbitrap high‐resolution mass spectrometers for quantitation compared with traditional triple quadrupole methods . The number of data points across an elution peak will be important if more rigorous quantitation is required.…”

“…Data acquisition methods were initially evaluated by determining how well targeted compounds could be detected and identified in eel samples fortified with a mix of analytes (Table ). Collecting data using non‐targeted acquisition (AIF or DIA) and then analyzing the results against a list of targeted analytes is a common strategy for HRMS methods and can be referred to as “non‐targeted data acquisition targeted analyses or “nDATA.” In order for a method to be applicable to regulatory monitoring, it was important to determine that critical analytes, including those listed in the current compliance program, could be identified in eel at the established target testing levels.…”

“…Several publications have demonstrated the equivalent capabilities using Orbitrap high-resolution mass spectrometers for quantitation compared with traditional triple quadrupole methods. [39][40][41][42] The number of data points across an elution peak will be important if more rigorous quantitation is required. With PRM, MS2 scans for precursor ions are automatically performed based on the inclusion list and loop count parameters before periodically returning to a MS1 scan.…”

Liquid chromatography–tandem mass spectrometry screening method using information‐dependent acquisition of enhanced product ion mass spectra for synthetic cannabinoids including metabolites in urine

“…Still, unknown peaks remain with untargeted acquisition which must be laboriously identified using in silico and literature data. Attempts are being made to tackle the big data problem of untargeted acquisition by optimizing software parameters but still, targeted acquisition with its high specificity and sensitivity remains a routine friendly and valid option.…”

A machine learning approach for handling big data produced by high resolution mass spectrometry after data independent acquisition of small molecules – Proof of concept study using an artificial neural network for sample classification

“…Using statistical software such as ProGenesis QI enables it to cope with the huge amount of data. Recently, optimization of software parameters using a Design of Experiment (DoE) approach has been proposed in a proof of concept study to reduce the amount of data …”