Smoothed Spherical Truncation based on Fuzzy Membership Functions: Application to the Molecular Encoding

Area of Science:

• Computational chemistry
• Cheminformatics
• Molecular modeling

Background:

• Traditional methods for truncating molecular interactions often focus on atom-groups.
• Existing molecular descriptors can be limited in their ability to discriminate between structurally diverse molecules.
• Fuzzy logic has not been extensively applied to molecular truncation methods.

Purpose of the Study:

• To introduce a novel spherical truncation method using fuzzy membership functions for molecular encodings.
• To evaluate the effectiveness of this method in discriminating between molecules and improving predictive modeling.
• To compare the proposed method against existing truncation techniques.

Main Methods:

• A spherical truncation method was developed, circumscribing molecules into spheres centered on their geometric centers.
• Fuzzy membership degrees were computed for each atom/amino acid based on its distance from the molecular center.
• Smoothing values for truncating interatomic relations were derived from the average fuzzy membership degrees.

Main Results:

• The fuzzy truncation method demonstrated superior ability in discriminating between structurally different molecules compared to non-truncated or non-fuzzy methods.
• Principal component analysis indicated that the method encodes orthogonal chemical information.
• Modeling studies showed improved robustness and statistically better results compared to 12 existing procedures.

Conclusions:

• The proposed fuzzy spherical truncation method is a valuable strategy for generating enhanced molecular encodings.
• This method improves the modeling ability of existing geometric molecular descriptors for various molecule sizes.
• The approach has potential applications in cheminformatics, drug discovery, and the development of predictive models.