henhaobio
Sign in

bioRxivBiophysicsPreprint

Generating 3D Models of Carbohydrates with GLYCAM-Web

Oliver C. GrantiD, Daniel Wentworth, Samuel G. HolmesiD, Rajan KandeliD, David SehnaliD, Xiaocong WangiD, Yao XiaoiD, Preston Sheppard, Tobias Grelsson, Andrew Coulter, Grayson Miller, Bethany Lachele FoleyiD, Robert J. WoodsiD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

The carbohydrate 3D structure-prediction tools (builders) at GLYCAM-Web (glycam.org) are widely used for generating experimentally-consistent 3D structures of oligosaccharides suitable for data interpretation, hypothesis generation, simple visualization, and subsequent molecular dynamics (MD) simulation. The graphical user interface (GUI) enables users to create carbohydrate sequences (e.g. DGalpb1-4DGlcpb1-OH) that are converted to 3D models of the carbohydrate structures in multiple formats, including PDB and OFF (AMBER software format). The resulting structures are energy minimized prior to download and online visualization. There are advanced options for selecting which shapes (rotamers) of the oligosaccharide to generate, and for creating explicitly solvated structures for subsequent MD simulation. The GLYCAM-Web builders integrate known conformational preferences of oligosaccharides, summarized here, and employ the GLYCAM forcefield for energy minimization with algorithms tailored for speed and scalability. Even for large oligosaccharides (100 residues, ~2100 atoms) a 3D structure is typically returned to the user in less than a minute.

In-depth summary

Claude Opus 5.5 · xhigh effort · 2 Nov 2026 · from the full text of version 1 (9 May 2025) · CC BY-ND 4.0

GLYCAM-Web provides interactive and API-driven carbohydrate 3D structure builders that assemble, resolve steric overlaps, and energy-minimize oligosaccharides up to 100 residues in under a minute.

Although carbohydrate force fields and conformational preferences are well established, complex branching rules and specialized nomenclature create barriers for researchers outside glycobiology who need 3D glycan models for visualization or molecular dynamics simulations. GLYCAM-Web bridges this gap by automating sequence input, rotamer selection, steric overlap resolution, and GLYCAM06 energy minimization through point-and-click, text, library, and JSON-API interfaces.

Key results

  1. ShownEnergy-minimized carbohydrate structures generated by GLYCAM-Web closely reproduce experimental glycosidic linkage dihedral angle distributions from high-quality Protein Data Bank structures.

    Filtered against high-quality PDB carbohydrate structures with average B-factors < 30 Å and biologically valid linkages, GLYCAM-Web reproduced experimental phi and psi dihedral angle distributions across alpha and beta linkages, achieving phi Z-score = 0.27 and psi Z-score = 0.49 across n=3229 beta1-4 linkages. · Accuracy and experimental validation; Figure legends

  2. ShownThe GLYCAM Molecular Modeling Library (GMML) and AmberTools pipeline returns overlap-resolved, energy-minimized 3D structures in under a minute even for large polysaccharides.

    Using lookup templates from the GLYCAM06 force field, two-stage greedy dihedral overlap resolution in 1° increments up to ±20°, and AmberTools tleap/cpptraj minimization with TIP3P or TIP5P solvation options, 100-residue oligosaccharides (~2100 atoms) are generated in less than 1 minute. · Abstract; Sequence parsing and initial structure…; Overlap resolution; Energy minimization protocol and output

  3. ShownGLYCAM-Web supports pre-built glycan libraries, bidirectional sequence entry, chemical derivatives, and programmatic batch generation, achieving broad community adoption.

    Pre-built libraries cover all 611 glycans in version 5.0 of the Consortium for Functional Glycomics array, logged approximately 372,698 visits over a 5-year period (reaching 80,000 visits per year in 2020), processed ~21 direct build requests per day in 2023-2024, and was cited in ~350 publications from 2020 through 2024. · Types of builders on GLYCAM-Web; The GLYCAM-Web API; Community Adoption

Methods, data and reproducibility

GLYCAM-Web parses carbohydrate sequences via a JSON API into the C++ GLYCAM Molecular Modeling Library (GMML), assembling 3D models from GLYCAM06 templates using empirical glycosidic linkage preferences. A two-stage greedy algorithm resolves steric overlaps by adjusting dihedral angles (±20° in 1° steps), followed by AmberTools tleap and cpptraj energy minimization and optional TIP3P or TIP5P solvation.

  • High-quality carbohydrate-containing structures from the Protein Data Bank filtered via GlyFinder (average B-factor < 30 Å) · n = >14,000 PDB instances filtered; n=3229 beta1-4 and n=3 beta1-3 linkages
  • Consortium for Functional Glycomics (CFG) mammalian glycan array library (version 5.0) · n = 611 carbohydrates
  • GLYCAM-Web server usage logs (2016-2020 and 2023-2024) and Google Scholar citation survey (2020-2024) · n = 372,698 visits (5 years), ~21 builds/day (2023-2024), ~350 publications
Methods & tools
GLYCAM-Web, GMML, GEMS, GLYCAM06, AmberTools, tleap, cpptraj, Mol*, GlyFinder, GlyGen, GlyConnect
Software & environment
C++ standalone library (GMML2) and GEMS under GNU Lesser General Public License v3.0; integrates AmberTools (tleap, cpptraj) and GLYCAM06 force field.
Experimental conditions
Returns a 3D structure in less than 1 minute for a 100-residue oligosaccharide (~2100 atoms) on the GLYCAM-Web server; server hardware specifications not stated.
Reproducibility gaps
For security reasons, the front-end website source code and deployment pipeline are not publicly released, and visitor logs from 2021-2024 were not recorded due to system changes.
Limitations (4)
  • Glycans adopt multiple interconverting conformations in solution (especially furanoses and L-iduronic acid ring puckers), so static energy-minimized models require subsequent molecular dynamics simulations to capture conformational ensembles. · stated by the authors
  • Generating a deoxy monosaccharide via atom deletion and charge transfer yields partial atomic charges that are similar but not identical to de novo computed charges for common deoxy sugars like L-fucose. · stated by the authors
  • PDB validation of beta1-3 glycosidic linkages was limited by small sample size after quality filtering (n=3), preventing statistically significant conclusions for that linkage class. · stated by the authors
  • Direct export to other glycan nomenclature formats (such as WURCS or IUPAC Condensed), crystallographic CIF output, and broader bacterial monosaccharide coverage are not yet implemented. · stated by the authors

Read this if you need to build, visualize, or prepare explicitly solvated 3D atomic models of linear, branched, or derivatized carbohydrates for molecular dynamics simulations.

Citations
24

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISNieto‐Fabre '26Thomès '25Adolf‐Bryfo '21Bojar '20Naravane '23Sundar '26
76% semantic matchBioinformatics
Ferran Nieto‐Fabregat, Oliver C. Grant, Xiaocong Wang+3 more·bioRxiv·Jan 27, 2026· 1 citations