Comparison of untargeted gas chromatography-mass spectrometry analysis algorithms with implications to the interpretation and putative identification of volatile aroma compositions.

• 5 untargeted GCMS workflows benchmarked on essential oil aroma characterization. • Workflows putatively identified >90% peak areas; but only 50–60% similarity. • Trace compounds more susceptible to differences in algorithmic interpretations. • We demonstrate application of benchmark metrics to improve confidence of results. The aroma profiling process requires the identification of the volatile compounds in a sample or its headspace. Typically, the identification of compounds relies on automated feature finding and matching algorithms to (putatively) identify and report compounds based on retention index and mass spectra matching against a compound library. We investigated the use of five different workflows and proposed three metrics (target accuracy A , identification percentage I , uniqueness U) to quantify their impact on generated aroma profiles of a mixture of fragrance standards and a commercial grade essential oil. All workflows accurately identified target compounds (100% in standards, >90% in samples) and reported similar compound identities for major GC–MS features, but beyond that could differ by up to 40–50%. Despite the variances, different workflows did not report conflicting compound identities. Aroma compositions primarily contained unreported or extra (putatively) identified compounds due to variations in mass spectral elucidations within the various workflows. Considering these differences, we show how the proposed metrics, I and U , could be modified to help the analyst interpret and evaluate reported volatile aroma compositions of unknown materials. [ABSTRACT FROM AUTHOR]