Novel Approach for Decoding Olfactory Receptors Interactions With Molecules Based on Multimodal Feature and Deep Learning Network.

Abstract

The olfaction transduction process commences when odorant molecules bind to specific olfactory receptors (ORs) located in the nasal cavity. Recognizing the interactions between odorant molecules and ORs remains a significant challenge due to the complex and nonlinear nature of the molecule-receptor relationship. Addressing this challenge is pivotal for advancing our understanding of human olfaction mechanisms and aiding in the development of novel synthetic pharmaceuticals. The primary difficulty arises from the intricate interactions between odorant molecules and ORs, where diverse molecule with varying physical and chemical properties can activate specific receptors, and conversely, individual receptors exhibit the ability to bind with multiple distinct molecules. In this study, we present a novel approach for predicting molecule-receptor interactions by leveraging multimodal deep learning networks to precisely identify specific ORs for given molecules. Our method demonstrates significant advancements and achieves an impressive accuracy of 95.1% when evaluated on a newly curated dataset. Notably, the proposed method achieves substantial improvements in performance metrics compared with other deep learning classification models and existing recognition approaches and exhibits robustness against discontinuities in the mapping of molecule structures to ORs. In addition, we developed a space distribution map to elucidate the structural intricacies of diverse receptors, revealing the clustering patterns among receptors. Our method facilitates a deeper understanding of the interaction mechanisms between molecules and receptors, laying a foundation for the digitization of receptor reaction.