Headed to ACS San Diego? Join us for Happy Hour!
← Back to glossary

De Novo Drug Design

Field
Field
Field

De novo drug design is a computational approach used to create novel drug-like molecules from scratch. Unlike traditional drug discovery methods that rely on libraries of known compounds or natural products, de novo design involves generating entirely new molecular structures that are optimized for specific biological targets. The downfall is that it is difficult to guarantee synthetic accessibility. The process typically involves:

  1. Generating Molecular Structures: Using algorithms to create novel chemical compounds.
  2. Scoring and Optimization: Evaluating the generated compounds based on their predicted binding affinity, drug-likeness, and other pharmacokinetic properties.
  3. Synthesis and Testing: Synthesizing the most promising candidates and testing them in biological assays.

Importance in Computational Drug Discovery

  1. Novelty: De novo design allows for the creation of entirely new chemical entities that may not be found in existing chemical libraries.
  2. Target Specificity: By designing molecules specifically for a given target, de novo methods can potentially yield compounds with high specificity and efficacy.
  3. Speed and Efficiency: Computational methods can rapidly generate and evaluate potential ligands, speeding up the early stages of drug discovery.
  4. Diversity: De novo design can explore a broader chemical space, increasing the likelihood of finding effective drug candidates.

Key Tools

  1. AutoGrow: An open-source tool for de novo drug design that uses genetic algorithms to evolve molecules towards desired properties.
  2. GANDI: A tool that uses generative adversarial networks (GANs) for de novo molecular design, focusing on generating drug-like molecules.
  3. REINVENT: A reinforcement learning framework for de novo drug design that optimizes molecular structures based on specified objectives.
  4. LigBuilder: A de novo design tool that generates novel ligands by assembling fragments in the active site of a target protein.
  5. DeepLigand: Utilizes deep learning techniques to generate novel drug-like molecules with desired properties.

Literature

"Accelerating factor Xa inhibitor discovery with a de novo drug design pipeline"

"Yin-yang in drug discovery: rethinking de novo design and development of predictive models"

  • Publication Date: 2023-06-21
  • DOI: 10.3389/fddsv.2023.1222655
  • Summary: This review discusses the importance of balancing the quantity and quality of data in de novo design and emphasizes the need to disclose inactive compounds and negative data in public repositories.

"De novo design of Na+-activated lipopeptides with selective antifungal activity: A promising strategy for antifungal drug discovery"

"De Novo Drug Design Using Transformer-Based Machine Translation and Reinforcement Learning of an Adaptive Monte Carlo Tree Search"

  • Publication Date: 2024-01-27
  • DOI: 10.3390/ph17020161
  • Summary: Introduces a novel drug design approach called drugAI that leverages the Encoder–Decoder Transformer architecture in tandem with Reinforcement Learning via Monte Carlo Tree Search (RL-MCTS) to expedite drug discovery.

"De novo drug design as GPT language modeling: large chemistry models with supervised and reinforcement learning"

  • Publication Date: 2024-04-22
  • DOI: 10.1007/s10822-024-00559-z
  • Summary: Explores the use of large chemistry models (LCMs) in drug discovery and highlights the potential for reinforcement learning with human feedback.

"Diffusion Models in De Novo Drug Design"

  • Publication Date: 2024-06-07
  • DOI: 10.1021/acs.jcim.4c01107
  • Summary: Reviews the state-of-the-art diffusion models for 3D molecular generation and their role in advancing drug discovery.

"Relevant Applications of Generative Adversarial Networks in Drug Design and Discovery: Molecular De Novo Design, Dimensionality Reduction, and De Novo Peptide and Protein Design"

  • Publication Date: 2020-07-01
  • DOI: 10.3390/molecules25143250
  • Summary: Examines the use of generative adversarial network (GAN) frameworks in drug design, including molecular de novo design and dimensionality reduction.

"Recognition of De Novo Drug Design for Targeted Chemical Libraries through Optimization Techniques"

  • Publication Date: 2024-09-18
  • DOI: 10.1109/ICOSEC61587.2024.10722441
  • Summary: Describes a system for rapid exploration of chemical space to form chemical compounds with personalized pharmacological profiles, improving therapeutic efficacy.

"Equivariant 3D-Conditional Diffusion Model for De Novo Drug Design"

  • Publication Date: 2024-11-04
  • DOI: 10.1109/JBHI.2024.3491318
  • Summary: Proposes an equivariant 3D-conditional diffusion model for generating pharmaceutical compounds based on 3D geometric information of target protein pockets.

"Enhancing De Novo Drug Design across Multiple Therapeutic Targets with CVAE Generative Models"

  • Publication Date: 2024-10-18
  • DOI: 10.1021/acsomega.4c08027
  • Summary: Introduces a Conditional Variational Autoencoder (CVAE) generative model for de novo molecular design, demonstrating its ability to generate molecules with specific property profiles.