Multistep Reaction Based De Novo Drug Design: Generating Synthetically Feasible Design Ideas.

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Drug Discovery

Background:

• De novo molecular design is crucial for identifying novel drug candidates.
• Assessing synthetic feasibility computationally remains a significant challenge in drug discovery.

Purpose of the Study:

• To present a "multistep reaction driven" evolutionary algorithm for de novo molecular design.
• To integrate synthetic pathway generation with molecular design to improve feasibility assessment.
• To enable multicriteria drug design considering various pharmacological and physicochemical properties.

Main Methods:

• Development of a "multistep reaction driven" evolutionary algorithm.
• Utilizing a library of 88 generic reactions and approximately 20,000 reactants.
• Employing common computer-aided drug discovery scoring methodologies (e.g., similarity, docking, QSAR, machine learning).

Main Results:

• The algorithm generates molecular structures with proposed synthesis pathways.
• Approximately 50% of known small molecule drugs (MW < 300) had "close" analogs identified.
• In silico generated synthetic pathways were rated as highly plausible by synthesis chemists, comparable to literature schemes.

Conclusions:

• The developed evolutionary algorithm effectively generates novel molecular designs with synthetically viable pathways.
• This approach aids in assessing the feasibility of computationally designed molecules.
• The methodology supports multicriteria optimization for drug design, enhancing the efficiency of drug discovery.