Pattern Recognition-Driven Detection of Circadian-Disruptive Compounds from Gene Expressions: High-Throughput Screening and Experimental Verification

Abstract

Circadian rhythms regulate the timing of numerous biological functions in organisms. Besides well-known external stimuli like the light–dark cycle and temperature, circadian rhythms can also be modulated by environmental substances. However, this area remains largely underexplored. Here, we developed a robust Pattern Recognition-Driven Prediction Approach (PRD–PA) that enables the identification of circadian-disruptive compounds from large-scale zebrafish transcriptomic profiling. The approach utilizes a circadian gene panel consisting of over 270 Circadian-Indicating Genes (CIGs) with stable and robust periodicity and combines it with a predictive model, known as the Differential Gene Expression Values of an Individual Comparison Model (DGVICM), that can effectively predict internal circadian phases from transcriptomic samples. By analyzing 692 aggregated gene expression profiles across 40 environmental substances, several were identified as having significant circadian-disruptive potential. These include glucocorticoids (e.g., prednisone (PRE) and triamcinolone (TRI)), the antithyroid agent propylthiouracil (PTU), and the widely used UV filter benzophenone-3 (BP-3). Both glucocorticoids and PTU are well-documented disruptors of circadian rhythms, and BP-3’s circadian-disrupting properties were validated through experimental exposures. Moreover, BP-3 analogs, including 2,4-dihydroxybenzophenone (BP-1) and 2,2’-dihydroxy-4-methoxybenzophenone (BP-8), were also found to exhibit similar circadian-disruptive effects. Overall, the present findings demonstrated the reliability of the PRD–PA approach for circadian disruption screening and highlighted the presence of diverse circadian-disruptive substances in our environment.