Banerjee, S.; Kumar, D.; Deshpande, P.; Kimbahune, S.; Panwar, A. S.

Cortisol is a primary biomarker of stress, released in sweat at concentrations that directly correlate with physiological stress levels. Detecting cortisol non-invasively offers significant potential for real-time stress monitoring and healthcare applications. Biosensors capable of binding cortisol can thus enable the development of novel diagnostic platforms for personalised health management. In our earlier work, a 38-mer peptide fragment derived from the protein 2V95 was identified as a functional binder to cortisol. In the present study, we applied generative artificial intelligence (AI) approaches to expand the sequence space and identify superior candidate peptides with improved binding affinity. By integrating sequence-based and structure-based AI models, we generated and screened a peptide library of nearly 10,000 sequences against cortisol, leading to the identification of high-affinity candidates for further evaluation. Keywords: cortisol, biosensors, stress detection, generative AI, peptide design