Computational researchers have introduced G-screen, a scalable virtual screening framework designed to evaluate candidate molecules within ultra-large chemical spaces by combining flexible global ligand alignment with reference-based protein interaction modeling. The new open-access method addresses a longstanding trade-off in rational drug discovery between high-throughput ligand-based evaluations that ignore receptor geometry and computationally demanding molecular docking workflows that explicitly model protein-ligand complexes.
Virtual screening has long served as an essential computational method for prioritizing candidate compounds from extensive chemical collections, yet balancing physical interaction detail with numerical throughput remains difficult. Conventional ligand-based strategies operate rapidly and reliably across massive molecule collections, but they lack direct consideration of receptor binding pockets. Conversely, standard docking algorithms explicitly sample conformational poses and optimize interaction energies against receptor structures, incurring substantial computational costs that impede routine application across ultra-large libraries.
G-screen operates specifically in settings where an experimentally solved or computationally predicted reference protein-ligand complex structure is accessible. Rather than performing full molecular docking simulations that require extensive three-dimensional pose sampling and atomic energy minimization, the framework generates alignment-guided pose hypotheses using an integrated algorithm designated as G-align. This approach rapidly positions screening candidates relative to the known binding mode of the reference ligand.
Following candidate positioning, G-screen evaluates the resulting aligned conformations using protein-aware pharmacophore interaction terms established from the reference binding site. This analytical strategy enables atomic-level contact evaluation while retaining the execution efficiency and numerical robustness characteristic of ligand alignment routines, avoiding the extensive conformational search spaces that slow down traditional structure-based docking pipelines.
The development team assessed the screening capabilities of G-screen across three established benchmarking datasets: DUD-E, LIT-PCBA, and MUV. Validation across these diverse benchmark systems demonstrated that the framework achieves effective protein-aware prioritization and beneficial early enrichment of active compounds, validating the predictive utility of alignment-guided pharmacophore scoring relative to reference complexes.
Computational efficiency measurements confirmed that G-screen maintains per-molecule processing speeds on the millisecond scale when deployed under multi-threaded parallel execution. This rapid execution profile permits extensive library evaluation without the extreme resource demands typically required for dense structural docking simulations across vast chemical databases.
The authors position G-screen as an intermediary prefiltering tool intended to bridge the operational gap between conventional ligand-based screening and computationally intensive structural docking.
The utility of G-screen remains bounded by its requirement for a reliable reference protein-ligand structure, making it applicable to target systems with available experimental or predicted co-complex models rather than uncharacterized binding sites.
The authors released G-screen and its underlying flexible alignment algorithm, G-align, through dedicated public code repositories at https://github.com/seoklab/gscreen and https://github.com/seoklab/galign.
Published in the Journal of Cheminformatics, the research demonstrates that alignment-guided pose generation bridges the operational gap between fast ligand comparisons and intensive docking calculations. By capturing explicit atomic-level interaction data while maintaining millisecond-scale execution runtimes per molecule under parallel processing, the tool provides a scalable strategy for prioritizing candidate compounds across ultra-large chemical collections whenever an experimentally determined or predicted reference complex structure is available.
Subject of Research: Scalable protein-aware computational virtual screening
Article Title: G-screen: scalable protein-aware virtual screening through flexible ligand alignment
Article References: Jung, N., Park, H., Yang, J., & Seok, C. (2026). G-screen: scalable protein-aware virtual screening through flexible ligand alignment. Journal of Cheminformatics. https://doi.org/10.1186/s13321-026-01297-0
Image Credits: AI Generated
DOI: 10.1186/s13321-026-01297-0
Keywords: virtual screening, computational chemistry, ligand alignment, G-screen, G-align, cheminformatics, protein-ligand interactions, pharmacophore modeling, drug discovery, molecular docking, chemical libraries, DUD-E benchmark, LIT-PCBA, MUV dataset, pose sampling, structural biology
Cite Scienmag News
Jason Bradley. (November 19, 2026). Flexible Alignment Tool Accelerates Protein-Aware Virtual Screening for Chemical Libraries. Scienmag. https://scienmag.com/flexible-alignment-tool-accelerates-protein-aware-virtual-screening-for-chemical-libraries/
Jason Bradley. "Flexible Alignment Tool Accelerates Protein-Aware Virtual Screening for Chemical Libraries." Scienmag, 19 November 2026, https://scienmag.com/flexible-alignment-tool-accelerates-protein-aware-virtual-screening-for-chemical-libraries/. Accessed 19 November 2026.
Jason Bradley. "Flexible Alignment Tool Accelerates Protein-Aware Virtual Screening for Chemical Libraries." Scienmag. November 19, 2026. https://scienmag.com/flexible-alignment-tool-accelerates-protein-aware-virtual-screening-for-chemical-libraries/








