[Computer-aided design of polyketides with the required properties]

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Rational drug design requires accurate prediction of biological activity for novel compounds.
• Polyketides represent a diverse class of natural products with significant therapeutic potential.
• Developing computational tools is crucial for efficient exploration of chemical space and identification of lead compounds.

Purpose of the Study:

• To present an approach for the rational design of new polyketides with tailored biological activity spectra.
• To introduce BioGenPharm software for generating polyketide libraries and predicting their activities.
• To optimize the selection criteria for identifying molecules with desired properties.

Main Methods:

• Development of BioGenPharm software for combinatorial library generation and activity spectrum prediction.
• Utilizing the PASS algorithm for polyketide activity spectrum prediction.
• Validation of the PASS algorithm's predictive accuracy using a dataset of 242 natural macrolides.
• Analysis of optimal probability of activity (Pa) thresholds for balancing sensitivity, specificity, and concordance.

Main Results:

• The PASS algorithm demonstrated a mean prediction accuracy of 75.5% for polyketide activity spectra.
• BioGenPharm software enables the generation of virtual polyketide libraries based on user-defined criteria.
• A virtual library of erythromycin analogues was generated, with a focus on selecting compounds with low predicted hepatotoxicity.

Conclusions:

• The proposed approach and BioGenPharm software facilitate the rational design of polyketides with predictable biological activities.
• The validated PASS algorithm is a reliable tool for predicting the activity spectra of polyketides.
• This method enables the selection of drug candidates with improved safety profiles, such as reduced hepatotoxicity.