Abstract:
Samples of biogenic VOCs are varied and complex, presenting a significant challenge to analytical scrutiny. This dual study investigates the applicability of comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry (GC×GC-TOFMS), in combination with machine learning, in identifying chemical markers — in the form of biogenic volatile organic compounds (VOCs) — as a tool of classification and prediction of discrete biological states.
The first study (Identifying predictive volatile markers of genus for southern African Plectranthus and Coleus using GC×GC-TOFMS and machine learning) investigates foliar VOCs as markers of genus for southern African Plectranthus and Coleus species. The second study (Identifying predictive volatile markers of malaria infection from human skin using GC×GC-TOFMS and machine learning) investigates cutaneous VOCs from the human epidermis as markers of malaria-infection. GC×GC-TOFMS was used to analyse the relevant VOC analytes, and three machine learning algorithms (an elastic-net regression, a random forest and a support-vector machine) were used to construct models of the acquired data from a training set, and to make predictions — of genus, in the case of the first study, and on malaria-infection status, in the case of the second study — on samples from a testing set.
For the first study (N=45 samples), a predictive accuracy as high as 90% was obtained (with a sensitivity of up to 100%), and a suite of sesquiterpenes (including α- and β-cubebene, β-ylangene, β-copaene, γ-cadinene and isogermacrene D) were identified as putative markers of genus Coleus. Though predictive models were not obtained in the case of the second study (N=52 samples), certain compounds were identified as being potential markers of a participant’s malaria-status. These include alcohols (such as (E)-2-octen-1-ol), sulphur species (such as isoamyl cyanide and isothiazole), and short- to long-chain aliphatic carboxylic acids (such as n-decanoic acid and 9-hexadecenoic acid).