June 5, 2005, ASMS 2005, San Antonio, TX, USA

Automated Algorithms for Componentization of LC/MS Datasets: Current Challenges and Goals for the Future

Mark Bayliss

Abstract

This presentation is aimed at Mass Spectrometrists, Informatics and MS Driven Automation Professionals

Currently all mass chromatographic software systems use extracted MS1 mass chromatograms with peak retention time and area detection to provide confirmation of the presence or absence of a chemical entity of interest. This technique is used in a large variety of application areas including but not limited to; metabolite extraction, impurity extraction, synthetic confirmation, forensics, food and flavours, biomarkers, de-replication studies and so forth.

For those applications using software driven automated analysis systems, that have low intervention by the spectrometrists, the fundamental assumption that is made; is that an extracted chromatographic peak at a defined mass value signifies the presence of a compound. In practice it is common to expect that upwards of 10% of extracted peaks in these automated systems will in fact be false positives. This is due directly to the presence of an isotope from another molecular ion species such as an adduct ion, ¹³C isotope, A+2 isotope, a fragment ion and so forth which is used as confirmation of ion presence. The problem with a projected 10% false positive response for a spectrometrist is that the confirmation process requires manual review of 100% of extracted peaks for satisfactory approval. This is clearly both time consuming and clearly not within the expectations of an automated system.

A software system that is able to overcome the complexities of ion analysis to ensure that the ¹²C molecular ion is used for confirmation must consider at least the following:

• The accurate retention time of each chromatographic peak.
• The presence of closely eluting chemical species in chromatographic space.
• The presence of significant noise due to random nature of ionization, spikes and dropouts throughout the data set.
• Extremely wide signal dynamic range (>10⁶).
• Isotopic contributions
• Adduct ions and their isotopic contributions
• Fragment ions and their isotopic contributions
• Differing chromatographic peak shapes as a function of intensity and purity.
• Large numbers of peaks.
• Drifting baselines due to gradient solvents.
• The presence of chromatographic artefacts not related to the compound(s) of interest.
• Keeping the processing time within acceptable limits to ensure high throughout of data processing tasks.

Clearly any software algorithm will require extensive algorithmic development to be able to overcome these hurdles in order to reduce the potential for false positive identification of individual molecular ion masses.

This presentation aims to provide a detailed overview of the advances that have been made in this field by ACD/Labs.