BoltzMol-1: Towards Reliable Virtual Screening for Fast and Cost-Effective Hit Discovery

Area of Science:

• Drug Discovery and Development
• Computational Chemistry
• Medicinal Chemistry

Background:

• Current small-molecule discovery methods often focus on well-characterized targets.
• This limits the range of biology accessible to small-molecule intervention.
• Resource-intensive high-throughput screening (HTS) is a common but costly approach.

Purpose of the Study:

• To present BoltzMol-1, a small-molecule hit discovery pipeline optimized for prospective discovery.
• To enable reliable hit discovery across diverse target classes.
• To reduce reliance on HTS by prioritizing compounds for experimental validation.

Main Methods:

• BoltzMol-1 couples model-driven compound ranking with streamlined procurement from commercial catalogs.
• Integrated ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) models predict kinetic solubility (logS), lipophilicity (logD), and Caco-2 permeability.
• These models act as an early triage layer, filtering compounds with unfavorable physicochemical and absorption properties.

Main Results:

• BoltzMol-1 demonstrated strong prospective performance across ten diverse targets, including those with no prior representation in training data.
• Functional actives or binders were identified for 6 out of 10 targets.
• Successes were achieved on receptors and enzymes often considered difficult for traditional structure- or ligand-based approaches, with modest experimental budgets.

Conclusions:

• BoltzMol-1 provides a practical framework for low-throughput, cost-constrained discovery campaigns.
• The pipeline delivers chemically tractable binders with favorable property profiles.
• This approach broadens the scope of targets amenable to small-molecule intervention.