A time-resolved axillary volatile fingerprint of human fear

Abstract

Human axillary odour can convey information about emotional state, but the molecular composition of fear related volatile emissions remains undefined. Here, we combine low background dual chamber axillary sampling, synchronous real time and offline mass spectrometry, chemical domain group independent component analysis and hierarchical Bayesian modelling to identify a molecular finger-print of acute fear during immersive virtual reality fear induction. In 37 healthy adults, this framework recovered five axillary volatile components associated with a continuous physiology derived fear index. The fingerprint comprised increased emissions of acetic, butyric, caproic and caprylic acids, octanal and acetone, together with decreased sulcatone, decanal, geranylacetone and citraconic anhydride. This coordinated pattern was reproducible across individuals and emerged from unsupervised chemical decomposition followed by regularised Bayesian selection, without prior biochemical constraints. Posterior predictive checks and leave one participant out refitting supported its robustness within the cohort. Its composition and directionality are consistent with sympathetic autonomic metabolic mobilisation, suggesting contributions from fatty acid handling, acetone related metabolism and axillary gland output. These findings define a candidate molecular cue of acute human fear and establish a replicable methodological template for decoding the chemical language of human emotion.