Prediction of synthetic accessibility based on commercially available compound databases

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Design

Background:

• Synthetic accessibility (SA) is crucial in drug design, as computer-generated compounds may be difficult to synthesize.
• Existing SA prediction methods often rely on retrosynthesis analysis and reaction databases.
• A practical and adaptable SA prediction method is needed.

Purpose of the Study:

• To develop a novel method for predicting compound synthetic accessibility (SA).
• To utilize commercially available compound databases and molecular descriptors for SA prediction.
• To assess the correlation between drug sales price and SA.

Main Methods:

• Estimated SA based on the probability of substructure existence within a compound library.
• Incorporated molecular descriptors such as symmetry atoms, graph complexity, and chiral centers.
• Validated predictions against expert manual assessments.

Main Results:

• The developed method achieved a Pearson correlation coefficient of 0.56 with expert SA assessments.
• The approach is adaptable for different compound vendors due to its reliance on compound databases, not reaction databases.
• A weak correlation was observed between the sales price of approved drugs and their SA values.

Conclusions:

• The new SA prediction method provides a feasible alternative to reaction-based approaches.
• This method can be readily customized for various compound vendors.
• Drug pricing appears to be influenced more by overall development costs than by synthetic accessibility alone.