Advanced fingerprint methods for similarity searching: balancing molecular complexity effects

Area of Science:

• Computational Chemistry
• Cheminformatics
• Drug Discovery

Background:

• Molecular fingerprints are early computational tools for similarity searching, valued for efficiency and ease of use.
• Despite simplicity, 2D molecular fingerprints have aided in identifying novel active compounds.
• However, their application in searching for active molecules presents significant complexities.

Purpose of the Study:

• To describe known caveats associated with fingerprint similarity searching.
• To present strategies for overcoming these difficulties in computational drug discovery.
• To introduce complexity-independent fingerprint representations and similarity metrics.

Main Methods:

• Review of established computational methods for molecular similarity searching.
• Analysis of biases in fingerprint search calculations, including molecular size and complexity effects.
• Development and description of novel complexity-independent fingerprint representations and similarity metrics.

Main Results:

• Identification of limitations in current similarity functions and threshold values for predicting biological activity.
• Demonstration of how molecular size and complexity can increase false-positive rates in fingerprint searches.
• Introduction of strategies to mitigate these issues, enhancing the reliability of similarity searching.

Conclusions:

• Standard 2D fingerprint similarity searching has inherent limitations that complicate reliable application.
• Addressing biases related to molecular size and complexity is crucial for improving accuracy.
• The proposed complexity-independent approaches offer enhanced strategies for effective molecular similarity searching in drug discovery.