Decoding Smell from Receptor Structure.

Area of Science:

• Neuroscience
• Biochemistry
• Computational Biology

Background:

• Understanding how individual odorant receptors (ORs) achieve ligand selectivity is crucial for deciphering the sense of smell.
• Despite advances, the precise mechanisms by which OR structure dictates odor recognition remain largely unknown.

Purpose of the Study:

• To develop a predictive framework for odorant receptor-ligand interactions using structural and sequence data.
• To investigate the role of three-dimensional receptor features in determining odorant selectivity.

Main Methods:

• Combined AlphaFold3-predicted OR structures with ESM2 protein sequence embeddings.
• Utilized large-scale sequencing data of in vivo olfactory sensory neuron activation.
• Developed a structure-informed deep learning framework to predict OR-ligand interactions.

Main Results:

• The deep learning model successfully predicted odorant receptor-ligand interactions.
• The learned receptor representation clustered ORs by functional similarity, outperforming sequence-based clustering.
• Identified specific binding-cavity subregions crucial for chemical recognition through feature attribution.

Conclusions:

• Odorant receptor ligand selectivity is fundamentally encoded in the 3D structural features of the receptors.
• The developed framework provides a predictive tool for interpreting the complex coding of olfactory information.