Searcharxiv⌕ Search

arXiv · 2611.18259

CARVE: A glycan-specific framework for assessing protein-glycan binding-pose quality

Abstract

Protein-glycan interactions depend on the precise three-dimensional arrangement of carbohydrates at protein-binding interfaces, yet reliable assessment of predicted complex structures remains challenging. Existing evaluations commonly rely on fixed root-mean-square deviation (RMSD) thresholds, which do not fully account for the structural diversity and conformational flexibility of glycans. Here we developed CARVE, a glycan-specific framework for evaluating protein-glycan structure predictions. Through topology-aware glycan parsing, CARVE integrates orientation RMSD (oRMSD), interface-ring RMSD (irRMSD) and the fraction of native contacts (Fnat) to assess glycan scaffold geometry, interface-ring placement and native-contact recovery. These descriptors define four quality categories (CARVE-Level) with geometric thresholds that scale with the degree of polymerization (DP), and a continuous CARVE-Score, which reproduced those categories on 25,068 independently generated poses without recalibration (macro-F1 = 0.892). Pairing these measures with glycan-specific topological, chemical and geometric validity checks, we benchmarked nine prediction methods on CARVE-Bench, comprising 1,974 experimentally determined protein-glycan complexes. Among the top 25 predictions per target, HADDOCK3 achieved the highest Acceptable-level success rate (47.0%), whereas no method produced a high-quality, structurally valid pose for more than one in five. Collectively, CARVE provides a standardized basis for assessing protein-glycan structure prediction methods and guiding the development of more accurate docking and co-folding approaches.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Yifei Xiao, Huifeng Zhao, Silong Zhai, Zichang Jin, Tingjun Hou, Yu Kang. 2026-11-13. CARVE: A glycan-specific framework for assessing protein-glycan binding-pose quality. https://searcharxiv.com/abs/2611.18259

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

Discovery of new antiviral agents through artificial intelligence: in vitro and in vivo results

In this research, we employed a deep reinforcement learning (RL)-based molecule design platform to generate a diverse set of compounds targeting the neuraminidase (NA) of influenza A and B viruses. A total of 60291 compounds were generated, of which 86.5% displayed superior physicochemical properties compared to oseltamivir. After narrowing down the selection through computational filters, nine compounds with non-sialic acid-like structures were selected for in vitro experiments. We identified two compounds, DS-22-inf-009 and DS-22-inf-021 that effectively inhibited the NAs of both influenza A and B viruses (IAV and IBV), including H275Y mutant strains at low micromolar concentrations. Molecular dynamics simulations revealed a similar pattern of interaction with amino acid residues as oseltamivir. In cell-based assays, DS-22-inf-009 and DS-22-inf-021 inhibited IAV and IBV in a dose-dependent manner with EC50 values ranging from 0.29 uM to 2.31 uM. Furthermore, animal experiments showed that both DS-22-inf-009 and DS-22-inf-021 exerted antiviral activity in mice, conferring 65% and 85% protection from IAV (H1N1 pdm09), and 65% and 100% protection from IBV (Yamagata lineage), respectively. Thus, these findings demonstrate the potential of RL to generate compounds with promising antiviral properties.

q-bio.BM↗

Agent-guided epitope excision enables economical de novo binder design

De novo protein design campaigns remain bottlenecked by the inference compute time required to generate and rank candidates, limiting wider functional exploration. Here, we introduce and evaluate an epitope excision approach that truncates targets to local interface windows to accelerate both backbone diffusion and structure prediction. Across 15 targets and 3,666 designs, excised windows closely reproduced full-target binding interfaces (median CαRMSD of 0.37 Å) and made it practical to design binders against the 719-residue extracellular domain of fibroblast activation protein alpha (FAP) by cutting RFdiffusion and AlphaFold3 GPU time by ~ 88% in one of the epitopes. Pairing this workflow with an agentic optimization loop allowed the system to autonomously identify catalytic-site targeting strategies. Under a fixed compute budget, a human-in-the-loop (HITL) workflow raised the computational hit rate from 5.06% to 9.33% while further cutting GPU hours by ~ 42%. We also found that omitting target multiple sequence alignments during refolding roughly substantially increased the apparent yield. Wet-lab validation with yeast display identified nanomolar binders from the generated designs with the HITL and the fully autonomous agent (human-out-of-the-loop, HOOTL) arms recovering more enriched designs than the PI-alone baseline. Together, target excision and agent-guided search offer a practical way to scale de novo binder design to large, therapeutically relevant proteins.

q-bio.BM↗

Cross-species digital xenocortication model delineates molecular correlates of human brain connectome

Cross-species digital xenocortication (CSDX) model based on pairing of a human brain structure with one or more mouse brain structures was constructed and implemented to dissect the molecular correlates of the human brain connectome. Design of the CSDX model is based on the assumption that a human brain region and its named mouse counterpart occupy corresponding positions in an evolutionary conserved mammalian brain connectome plan, so that a projection strength measured between two mouse structures is informative about the relation between the two human structures paired with them. In the CSDX model, gene expression for the node was measured in the human brain, while connection strength between two nodes was measured in the mouse brain. CSDX models of different parcellations were employed to explore the potential regulatory associations of brain connectome patterns and constituents of protein-protein interaction (PPI) cascades of hybrid assembly pathways of endogenous multiprotein complexes. Metrics of continuous engagement of the PPI cascade nucleated protein dimers of CORUM multiprotein complexes have been identified as statistical components of the multivariate transcriptome-connectome associations. It accounts for additional connectome associated variations beyond the five measured axes in the 82 nodes CSDX model. Comparisons of the seven arm models of transcriptome-connectome associations demonstrate that the mouse derived model generally fits the mouse brain connectome better, while the human derived model assigns a significantly larger fraction of its transcriptome-connectome association to unique rather than net shared contributions. This study nominates elite panels of top scoring candidate protein dimers and CORUM multiprotein complexes for follow up mechanistic studies of their potential roles in structural functional features governing mammalian brain connectome.

q-bio.BM↗