Predictive Machine Learning Models for Olfaction.

Area of Science:

• Neuroscience
• Computational Chemistry
• Machine Learning

Background:

• Linking odorant chemical structure to olfactory perception is a long-standing challenge.
• Subjectivity, incomplete physiological understanding, and lack of standardized descriptions hinder progress.
• Computational approaches, including machine learning, offer promising solutions.

Purpose of the Study:

• To present a comprehensive workflow for developing machine learning models for odor prediction.
• To cover the entire process from problem formulation to model deployment and interpretation.
• To provide a framework for synthetic chemists and data scientists in olfaction research and industry.

Main Methods:

• Development of a machine learning workflow for odor prediction.
• Integration of data-driven prediction enhancement and interpretation techniques.
• Evaluation of models for real-world applicability.

Main Results:

• A structured methodology for building and evaluating odor prediction models.
• Insights into recent advancements for improving and interpreting data-driven olfactory predictions.
• Demonstration of a framework applicable to synthetic chemistry and data science.

Conclusions:

• The presented workflow offers a robust framework for odor prediction.
• Advancements in machine learning enhance the accuracy and interpretability of olfactory models.
• This approach supports applications in the fragrance and perfume industries and broader olfaction science.