DeepOlf: Deep Neural Network Based Architecture for Predicting Odorants and Their Interacting Olfactory Receptors

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning in olfaction research

Background:

• Olfactory receptors (ORs) in the nasal cavity bind odorants, initiating the olfaction transduction mechanism.
• Odorant-OR interactions are specific, driven by molecular shape, physical, and chemical properties.

Purpose of the Study:

• To propose a deep neural network architecture, DeepOlf, for predicting potential odorants and their interacting ORs.
• To model odorant identification and odorant-OR interaction as binary classification problems.

Main Methods:

• Utilized a deep neural network architecture (DeepOlf) incorporating molecular features and fingerprints of odorants and ORs.
• Employed multiple classifiers for binary classification tasks.
• Compared DeepOlf's performance against classical methods like Support Vector Machines (SVM), Random Forests (RF), and k-Nearest Neighbors (k-NN).

Main Results:

• DeepOlf achieved high accuracy: 94.83% for odorant prediction and 99.92% for odorant-OR interaction prediction.
• The deep neural network framework demonstrated superior accuracy compared to SVM, RF, and k-NN.
• DeepOlf outperformed the existing SVM-based prediction server, ODORactor.

Conclusions:

• DeepOlf represents the first application of deep learning for predicting odorants and their interacting ORs.
• The proposed deep learning approach offers enhanced accuracy for odorant identification and OR interaction prediction.
• The DeepOlf tool is publicly accessible for further research and application.