COSMIC: Molecular Conformation Space Modeling in Internal Coordinates with an Adversarial Framework.

Area of Science:

• Computational chemistry
• cheminformatics
• drug discovery

Background:

• Accurate molecular conformation modeling is crucial for drug discovery.
• Current diffusion models offer good performance but are computationally expensive.
• There is a need for faster and efficient methods for conformation generation.

Purpose of the Study:

• To introduce COSMIC, a novel adversarial generative framework for molecular conformation.
• To demonstrate COSMIC's comparable generative performance with improved efficiency.
• To address the limitations of slow sampling in existing diffusion models.

Main Methods:

• COSMIC utilizes a two-stage generative process.
• The first stage generates conformations using internal coordinates (rotation and translation invariant).
• The second stage refines conformations by predicting atomic distances and applying optimization.

Main Results:

• COSMIC achieves comparable generative performance to state-of-the-art diffusion models.
• The framework offers significantly faster sampling and training times.
• The model considers conformational energy, ensuring physical plausibility.

Conclusions:

• COSMIC presents a time-efficient alternative for molecular conformation generation.
• The framework maintains high generative performance and diversity.
• COSMIC can accelerate computational drug discovery pipelines.