Interactive Knowledge-Based Kernel PCA for Solvent Selection.

Area of Science:

• Green Chemistry
• Computational Chemistry
• Chemical Engineering

Background:

• Selecting sustainable solvents is vital for reducing the environmental footprint of chemical processes.
• Existing data-driven tools, like principal component analysis (PCA), aid solvent selection in the pharmaceutical industry.
• Interactive knowledge-based kernel PCA offers a way to incorporate expert knowledge into solvent mapping.

Purpose of the Study:

• To apply interactive PCA for improved solvent selection in chemical processes.
• To develop an intuitive interface for interactive PCA within an electronic laboratory notebook.
• To promote the adoption of sustainable chemistry practices.

Main Methods:

• Utilized interactive knowledge-based kernel PCA, a variant of PCA.
• Developed an intuitive user interface integrated into the AI4Green electronic laboratory notebook.
• Created evidence-based user guidelines to support the interactive PCA process.

Main Results:

• Successfully applied interactive PCA to the task of solvent selection.
• The AI4Green platform provides an intuitive interface for users.
• Identified four potential sustainable solvent substitutions for a thioesterification reaction.

Conclusions:

• Interactive PCA is a valuable tool for selecting sustainable solvents.
• The AI4Green platform effectively integrates this tool to encourage greener chemistry.
• The developed method and guidelines facilitate the identification of environmentally benign solvent alternatives.