Dimensionality reduction of COSMO-RS molecular descriptor using functional principal component analysis (FPCA) for...

Area of Science:

• Green Chemistry
• Computational Chemistry
• Chemical Engineering

Background:

• Traditional solvent screening relies on experimental data or computational methods like COSMO-RS.
• Principal Component Analysis (PCA) has limitations with complex molecular descriptors such as σ-potentials.
• Similarity maps are useful for exploring alternative molecules similar to existing solvents.

Purpose of the Study:

• To introduce Functional Principal Component Analysis (FPCA) as an advanced dimensionality reduction technique for solvent mapping.
• To develop a framework for efficient solvent screening and substitution towards greener and safer alternatives.
• To leverage the functional nature of σ-potentials for improved solvent space representation.

Main Methods:

• Application of Functional Principal Component Analysis (FPCA) to a database of 1588 solvents.
• Utilizing σ-potentials derived from COSMO-RS theory as molecular descriptors.
• Developing a two-dimensional solvent map for clustering and similarity analysis.

Main Results:

• FPCA achieved a 2D solvent space representation with minimal information loss (0.5%).
• Principal components were directly associated with electron donor and acceptor characteristics.
• Natural solvent clustering emerged, facilitating identification of structurally and functionally similar solvents.

Conclusions:

• FPCA provides a suitable framework for solvent substitution and computer-aided solvent design.
• The methodology supports the transition towards more sustainable chemical practices.
• This approach enhances preliminary solvent screening by revealing solvent relationships effectively.