Structure

Protein structure predictors achieve high single-state accuracy, but it remains unclear whether they can recover functionally relevant conformational ensembles or account for the presence of ligands and/or binding partners. Here, we benchmark AlphaFold3, Boltz-2, Chai-1, and BioEmu on four canonical multi-state proteins (Pf-MATE, LAO, SecA, and {beta}2AR), quantifying state bias and sampling breadth against experimental reference structures. Models frequently default to a dominant state represented in the PDB; small-molecule ligands have weak or inconsistent effects, while large protein partners drive clear conformational switching between states. Multiple sequence alignment (MSA)-based approaches (AF-Cluster and random subsampling) recapitulate similar biases, indicating that this behavior is not unique to newer architectures. These results underscore current limitations for multi-state protein structure prediction and structure-guided ligand discovery. TOC Graphic O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=111 SRC="FIGDIR/small/737860v1_ufig1.gif" ALT="Figure 1"> View larger version (12K): org.highwire.dtl.DTLVardef@3bf389org.highwire.dtl.DTLVardef@1f1c436org.highwire.dtl.DTLVardef@188ea8aorg.highwire.dtl.DTLVardef@1de236e_HPS_FORMAT_FIGEXP M_FIG C_FIG

Peptidoglycan biogenesis requires membrane flippases to translocate lipid-linked precursors across the cytoplasmic membrane for processing (1). This essential step is mediated by MurJ, the lipid II flippase conserved across all peptidoglycan-producing bacteria (2). While MurJ from diderm bacteria has been structurally resolved in multiple conformational states (3-6), its monoderm homolog remains uncharacterized. Monoderm MurJ homologs exhibit substantial sequence divergence yet retain the same lipid II flipping function (7) and are promising antibiotic targets. Here we report structures of Staphylococcus aureus MurJ (SaMurJ) captured in both outward- and inward-facing conformations. These structures show that SaMurJ adopts the conserved MOP family fold and undergoes conformational transitions consistent with an alternating-access mechanism. Our findings reveal conserved and divergent features of MurJ between diderm and monoderm bacteria that are critical for lipid II flipping and provide a structural framework for probing substrate recognition and specific inhibition. Significance StatementThe growing global threat of antibiotic resistance and the limited development of new antibacterial therapies underscore the urgent need to identify and mechanistically characterize new antibiotic targets and mechanisms. MurJ is an essential membrane transporter required for cell wall biosynthesis and represents an attractive but unexplored antibiotic target. Here we determine the structures of MurJ from a clinically critical monoderm pathogen Staphylococcus aureus in key conformational states during its transport cycle. This work advances our understanding of an essential step in bacterial cell wall synthesis, reveals key distinctions between monoderm and diderm MurJ, and defines structural features that can be exploited for antibiotic discovery.

The human sodium-citrate cotransporter NaCT imports various tri- and dicarboxylates into the cell as TCA cycle intermediates. This substrate uptake process is driven by an inward sodium gradient. The protein is a member of the Divalent Anion-Sodium Symporter (DASS) family. Whereas extensive biochemical and structural studies have been carried out for NaCT, how the substrate binding and translocation is coupled to the sodium gradient remains unclear. Here using single particle cryo-electron microscopy, we determined the structures of the human NaCT protein in three states: sodium-free, in the presence of sodium, and sodium- and substrate-bound. These structures suggest a simultaneous binding mechanism for sodium-substrate coupling, distinct from the sequential binding, conformational selection mechanism previously observed for the bacterial DASS protein VcINDY.

Cryogenic-electron microscopy (cryo-EM) has revolutionized the field of protein structural biology. The structures of large membrane proteins are now routinely determined by cryo-EM to near atomic resolution. However, in the medium resolution range of cryo-EM maps (>[~]2 [A]), negatively charged sidechains of acidic residues are not well-resolved due to the negative electrostatic potential of the region. This may lead to incorrect sidechain models for residues like glutamic acid or aspartic acid that are central for proton transfer activity in various respiratory and photosynthetic enzymes. We previously proposed that the acidic residues with weak or non-existent cryo-EM density can be modeled to represent their low proton affinity conformations. Here, we tested this hypothesis on a larger data set of acidic amino acid residues in two high-resolution respiratory complex I structures. By using faster sidechain modeling and proton affinity prediction tools, we created a workflow that generates sidechain conformations of selected amino acid residues. We validated the sidechain conformation predictions by Q-score analysis and atomistic molecular dynamics simulations in different charged states. The proposed workflow provides a way to rapidly obtain sidechain conformations of acidic residues with weak cryo-EM densities and can be integrated into the existing cryo-EM modeling pipelines to speed up sidechain rotamer prediction.

Neuroendocrine cells communicate with other cells by releasing neurotransmitters or hormones by exocytosis, which involves SNARE-mediated fusion between secretory vesicles and the plasma membranes of the secreting cells. In neurons two plasma membrane SNARE proteins, Syntaxin-1a and SNAP25, join with the vesicle membrane SNARE protein Synaptobrevin-2 to form a four-helix bundle, which drives membrane fusion. The assembly of these SNAREs, which is highly orchestrated in cells, has been intensely studied in solution using fragments of the SNARE proteins without their transmembrane domains or lipid anchors. However, in cell and model membranes, Syntaxin and SNAP25 are known to oligomerize and cluster, and little is known about how clustering affects their incorporation into SNARE complexes. In cells, the SM protein Munc18 has been implicated in aiding secretory vesicle docking and facilitating SNARE complex assembly through its interactions with Syntaxin. To understand how Munc18 orchestrates SNARE complex assembly on membranes, we employed protein reconstitution in model membranes as well as biochemical and biophysical assays to show that lipid-dependent oligomerization of Syntaxin affects Munc18-Syntaxin binding and SNAP25 insertion into the plasma membrane acceptor SNARE complex. We showcase the consequences of the different modes of Munc18-Syntaxin and SNAP25 interaction on Syntaxins oligomerization and orientation relative to the membrane surface, as well as on docking and fusion of purified insulin granules. We also determined low-resolution structures by cryoEM in nanodiscs and on the surface of proteoliposomes of membrane-bound assembly states of Munc18/Syntaxin and Munc18/Syntaxin/SNAP25 complexes.

The inhibitor of {kappa}B kinase (IKK) complex integrates diverse cellular inflammatory responses, and induces transcription factor NF-{kappa}B. The molecular mechanism by which IKK becomes catalytically active in response to signaling remains unclear despite structural knowledge of the individual IKK1/, IKK2/{beta}, and NEMO/IKK{gamma} protein components within its hetero-oligomeric assembly. Cryo-EM of the IKK2/{beta} homodimer bound to an associating NEMO/IKK{gamma} protein fragment, reveals multiple conformers. Mutual exclusivity of dimeric conformers, canonical versus alternate, is reflected in and dependent upon order-to-disorder transition of the canonical 6-helical bundle dimerization interface. Correlation of this unusual structural plasticity of IKK2/{beta} with its biochemical and cellular activities suggests mechanistic possibilities for how association with its partner scaffold protein NEMO/IKK{gamma} and polyubiquitin chains might dictate catalytic activation of IKK through distinct IKK2/{beta} conformers.

DnaK, a prokaryotic Hsp70 chaperone, plays a central role in proteostasis by restoring native structures to heat-denatured proteins in an ATP-hydrolysis-dependent manner. While structures of DnaK in complex with nucleotides, co-chaperones, and short peptides have been resolved, structures with larger, stably folded substrates--such as firefly luciferase (Fluc, 61 kDa)--are lacking, limiting mechanistic understanding of how DnaK refolds such proteins. Here, we generated models of the DnaK-Fluc complex using AlphaFold3 and evaluated their mechanistic relevance. In one of three major model clusters, Fluc is unexpectedly immobilized beneath the DnaK -helical lid against the nucleotide-binding domain (NBD), rather than interacting primarily with the substrate-binding domain {beta} (SBD{beta}), as commonly assumed. All-atom molecular dynamics simulations indicate that, in this configuration, the lid can engage a thermally destabilized Fluc helix (residues 405-411), which we recently identified as the first--and likely the only--helix to irreversibly melt at 42 {degrees}C. Upon binding, the lid forms extensive hydrogen-bonding interactions with the melted helix. These interactions persist during lid movement toward SBD{beta} (following ATP hydrolysis), enabling the lid to actively extract the helix from the Fluc surface. In contrast, simulations with the helix in its native folded state show that the lid cannot extract it, leaving the native structure unaffected. Equilibrium simulations further indicate that, once extracted and mechanically stretched, the melted helix can refold to its native conformation. Together, these findings suggest a revised mechanism for DnaK-mediated protein refolding, in which the -helical lid selectively recognizes structurally compromised segments, forms stabilizing hydrogen bonds, and--powered by ATP hydrolysis--mechanically pulls them away from the protein surface to facilitate their refolding. SIGNIFICANCEDnaK is a model chaperone, which can reactivate thermally denatured proteins. Over the span of 40 years, significant findings have been made about DnaKs structure, dynamics and interactions with its co-chaperones, the exact molecular mechanism by which DnaK refolds misfolded proteins remains a mystery. This work exploited Alphafold3 to generate atomistic models of complexes between DnaK and Firefly luciferase. Molecular dynamics simulations directly captured how DnaK may assist thermally denatured proteins by mechanically pulling out their misfolded helices. This study provides a new insight into the DnaK mechanism.

Monobodies are engineered binding proteins that recognize extended protein surfaces and offer advantages over small-molecule inhibitors for targeting challenging KRAS oncoproteins. Monobody 12D4 exhibits high affinity and selectivity for the oncogenic KRAS(G12D) mutant, but the molecular determinants governing its recognition and the basis for its mutant selectivity remain poorly understood. Here, we combined molecular dynamics simulations and energy calculations to characterize the interactions between monobody 12D4 and WT KRAS as well as four clinically relevant oncogenic variants (G12C, G12D, G12V, and G12R) in both GTP- and GDP-bound states. Our simulations revealed that 12D4 recognition depends on a conserved hydrophobic interaction network centered on the monobody FG loop (residues L77, F78, and W79). This network forms stable contacts with KARS Switch II and 3-helix. The energy calculations also showed that residue K75 of 12D4 formed a mutation-specific electrostatic interaction with KRAS G12D. This interaction contributed significantly to the affinity of 12D4 toward this mutant, whereas this interaction was absent in other variants. No monobody currently exists for targeting KRAS G12R in either nucleotide state, and no monobody selectively targets KRAS G12C and G12V in the GDP-bound inactive state. To address these, we performed computational redesign at residues 75. We identified mutations (K75Q, K75Y, and K75M) that enhanced predicted binding to G12C, G12R, and G12V variants through reorganization of interfacial contacts. Our work establishes a structural framework for understanding KRAS-monobody recognition and provides a rational foundation for engineering variant-selective monobodies with improved affinity toward previously untargetable KRAS mutants.

T cells are important components of the adaptive immune system and develop through a selection process regulated by signaling through the T-cell receptor (TCR). Thymocyte-Expressed Molecule Expressed in Selection (THEMIS) is a TCR-proximal protein that modulates the activity of Shp1 phosphatase to influence TCR signaling during development. THEMIS has been shown to both activate and inhibit Shp1, but the molecular mechanisms of these functions are poorly understood. THEMIS contains two rare Cysteine All-Beta In THEMIS (CABIT) domains, the N-terminal of which interacts with Shp1 and is likely responsible for modulation of its phosphatase activity. Herein, we report the first crystal structure of the THEMIS CABIT1 domain. While a portion of the CABIT1 domain is poorly resolved, it appears to share the same overall fold observed in our recent CABIT2 crystal structure and AlphaFold predictions. We show that phosphorylation of the CABIT1 domain by LCK is required for association with SHP1 and that phosphorylated CABIT1 can protect Shp1 from oxidation and inhibition by reactive oxygen species (ROS), which may serve as a mechanism by which THEMIS enhances Shp1 activity.

Vacuolar-type ATPases (V-ATPases) are evolutionarily conserved rotary proton pumps that play essential roles in the eukaryotic cell. By coupling ATP hydrolysis in their cytosolic V1 region to proton translocation through their membrane-embedded VO region, V-ATPases establish and maintain an acidic pH in the lumen of several different organelles. Functional diversity in the pump is enabled by multiple paralogous genes for the subunits of the complex, which are expressed in a tissue- and organelle-specific manner. Interactions between V-ATPase and TLDc domain-containing proteins have been shown to regulate the enzyme in yeast and mammals but their relevance in plants has remained unclear. We isolated the endogenous V-ATPase from Arabidopsis thaliana leaves and determined its structure by electron cryomicroscopy. Mass spectrometry showed that most of the enzyme originated from the tonoplast. The structural analysis revealed the full rotary catalytic cycle of the plant V-ATPase, and a combination of structural and biochemical experiments showed S-acylation of subunits AP1 and the tonoplast-specific subunit a3 isoform. A subpopulation of complexes derived from the trans-Golgi network/early endosome was identified and found to bind the TLDc protein OXR5. Together, these findings reveal plant-specific features in V-ATPase and suggest organelle-specific interactions with TLDc proteins, pointing to conserved but context-dependent V-ATPase regulation in eukaryotes.

Structure prediction models have moved from single proteins to assemblies that include diverse biomolecules and their modifications. AlphaFold3 (AF3) and related models extended structural modelling via an all-atom framework, opening many new potential applications in structural biology. We evaluate how well the new capabilities of AF3 translate into application tasks in diverse areas: prediction of ubiquitinated protein structures, T-cell receptor (TCR)-epitope recognition, antibody-antigen complexes, protein-RNA and protein-lipid interactions. We find that, while AF3 can perform well in favourable settings, this performance is uneven across applications. In RNA-target predictions, the model confidence fails to separate genuine from decoy interaction partners and in several tasks accuracy depends on the presence of related complexes in the training set. Taken together, our assessment is more cautious than for AF2, whose gains in modelling monomers and complexes were clear and broadly generalisable. AF3s extension to new biomolecule types shows less consistent performance and generalisation. AF3 can be a powerful tool for hypothesis generation and prioritisation, but its predictions and use of confidence metrics will depend strongly on the specific application area and must be interpreted with respect to training-set overlap. We expect that the benchmarks provided here will serve for testing of future developments in the structure prediction field.

Fatty Acid Transport Protein 2 (FATP2) couples fatty acid uptake to intracellular activation and is associated with pathological lipid accumulation in cancer and nonalcoholic fatty liver disease. Here, we present cryo-electron microscopy structures of human FATP2 across its reaction cycle. Our structures suggest that FATP2 recruits fatty acids directly from the membrane interface through a hydrophobic tunnel. Catalysis involves a [~]130{degrees} rotation of the C-terminal domain, a transition trapped by the antihypertensive drugs isradipine and benidipine. Both drugs lock the enzyme in a thioester-forming state, but benidipine exhibits superior efficacy by extending a bulky moiety into the primary catalytic tunnel to sterically block substrate entry. Furthermore, we identify a product inhibition mechanism where excess acyl-CoA traps the enzyme, potentially limiting metabolic overload. These findings provide a structural framework for understanding vectorial fatty acid channeling and a scaffold for developing modulators of metabolic flux. HighlightsO_LICryo-EM structures of human FATP2 reveal a membrane-anchored lollipop topology C_LIO_LIEndogenous fatty acids within a hydrophobic tunnel delineate the fatty acid uptake pathway C_LIO_LIIsradipine and benidipine displace fatty acids to trap a non-productive conformation C_LIO_LIAcyl-CoA product inhibition may provide negative feedback via steric occlusion C_LI

In single particle cryo-EM, analysis of continuous conformational heterogeneity has always been challenging. Both linear and deep learning-based methods treat conformational heterogeneity as perturbations to the consensus average conformation, limiting their capability in analyzing large protein motions. While classic conformational classifications are capable of handling large domain motion, they bin continuous protein dynamics into discrete static substates. Here, we present cryoROLE, a computational tool that extracts the continuous conformational dynamics embedded in the static composite map constructed from multi-body refinement into a landscape of relative orientation between the moving domains. Depicted in real space, the landscape allows intuitive interpretations of domain motion and the population of poses in the conformational space. Applying it to various biological systems reveals hidden conformational dynamics that are relevant to protein functions.

Biomolecular structure prediction tools such as AlphaFold have achieved remarkable success in predicting structures of single proteins and multiprotein complexes. AlphaFold3 now incorporates the capability to model complexes containing nucleic acids and chemically modified side chains. Investigators can now predict structures of proteins bound to chromatin, where interactions with nucleosomal DNA and histone post-translational modifications converge to control genome function. To evaluate its robustness in modeling chromatin complexes, we benchmarked AlphaFold3 on 115 structures containing nucleosomes whose coordinates were released by the Protein Data Bank after the training set cutoff date. We find that AlphaFold3 excels at predicting histone-driven interactions and accurately models complexes that deposit and recognize post-translational modifications. By contrast, AlphaFold3 struggles to predict structures of chromatin factors that primarily engage nucleosomal DNA, notably transcription factors and chromatin remodelers. Finally, we show that AlphaFold3 can faithfully recapitulate known post-translational modification recognition patterns, matching experimentally determined specificity profiles. This assessment of the capabilities and limitations of AF3 in chromatin structural biology provides a roadmap for its effective application to studies of chromatin regulation and PTM readout, while identifying key areas for future algorithmic refinement. SignificanceStructure prediction with AlphaFold has become an invaluable tool in experimental biology, and the accuracy of many of its predictions has been verified in structural and biochemical studies. With the recent incorporation into AlphaFold3 of nucleic acids and post-translational modifications, this prediction tool can now be applied to chromatin structural biology. Our benchmarking of AlphaFold3 reveals its strengths and weaknesses in predicting structures of proteins bound to nucleosomes, thereby providing a framework for using these models in mechanistic studies of chromatin regulation. We introduce metrics for evaluating structures of nucleosome complexes that highlight AlphaFold3s strengths in predicting protein-nucleosome interactions and post-translational modification specificity.

More than two thirds of proteins in the human proteome are predicted to contain intrinsically disordered regions (IDRs), which lack stable folded structure. IDRs are critical for biological regulation and organization, as targets for post-translational modifications, and as mediators of biomolecular condensates. To address the pressing need for better structural models enabling functional insight, we developed AlphaFlex to model fully atomistic conformer ensembles for proteins predicted to have IDRs, modeled in the context of AlphaFold folded domains and an implicit bilayer for transmembrane proteins. The AlphaFlex resource provides conformational ensembles of human proteins from the AlphaFold database with identified IDRs in the Protein Ensemble Database that is mirrored in UniProt. This transformative resource of AlphaFlex ensembles provides physically and biologically relevant full-length models for IDR proteins, including scaffold proteins, those with IDR:folded-domain interactions, regulatory and condensate proteins requiring exposed binding elements, conditionally folding IDRs, and transmembrane proteins containing IDRs.

Glutamate dehydrogenase (GDH) is a highly regulated key enzyme that catalyzes the reversible oxidative deamination of glutamate to alpha-ketoglutarate, positioning it at a critical hub linking amino acid catabolism to energy production while supplying ammonia for urea synthesis and other nitrogen pathways. Early investigations have shown that bovine GDH (bGDH), which shares 98% sequence identity with its human homolog, assembles into polymeric filaments with altered allosteric responses. Filamentation has only relatively recently been appreciated as a widespread mechanism of enzyme regulation, prompting a reevaluation of these early observations in GDH. Here, we use high resolution cryogenic electron microscopy (cryo EM) to show that bGDH hexamers assemble via reciprocal antenna interactions that oppose the conformational changes associated with GTP inhibition, revealing how filamentation reshapes GDH allostery and with implications for the treatment of human disease.

Cryogenic electron microscopy (cryoEM) is now routinely used for high-resolution structure determination of biological macromolecules. However, many biological specimens exhibit varying degrees of preferred orientation on cryoEM grids, resulting in uneven sampling of three-dimensional Fourier space. This orientation bias produces anisotropic reconstruction artifacts and, in severe cases, can exacerbate particle misalignment during iterative refinement, thereby limiting the success rate of near-atomic resolution cryoEM structure determination. This protocol provides a practical guide for applying spIsoNet, a self-supervised deep-learning method, to mitigate preferred-orientation issues in cryoEM reconstructions. We describe two complementary workflows: (1) map Anisotropy Correction to correct anisotropic artifacts of cryoEM maps and (2) particle Misalignment Correction, which integrates spIsoNet with RELION external reconstruction to improve particle-pose estimation. We demonstrate these workflows using two influenza hemagglutinin (HA) trimer datasets representing moderate and severe degrees of preferred-orientation bias. The protocol includes installation instructions, parameter-selection guidance, quality-control checkpoints and troubleshooting advice, and can typically be completed in ~7 hours on a workstation equipped with four NVIDIA A100 GPUs. Together, these workflows provide step-by-step guidance for using the open-source spIsoNet software to mitigate the preferred-orientation problem directly from experimental data.

Membrane contact sites are organized by protein assemblies that physically couple organelles and coordinate lipid metabolism, yet the structural principles that enable lipid exchange across these junctions remain poorly defined. At the nuclear-vacuolar junction (NVJ) in budding yeast, the tethering protein Mdm1 and its binding partner Nvj3 form a complex that regulates lipid metabolic pathways, but the structural features underlying their interaction have not been resolved. Here, we use AlphaFold-based complex prediction and comparative structural analysis to define the organization of Nvj3-Mdm1 complex assembly. We identify a high-confidence heterodimer in which conserved PXA and PXC domains generate an extended tunnel spanning both proteins. Tunnel analysis predicts a core hydrophobic conduit traversing the Nvj3-Mdm1 interface, consistent with a lipid-compatible architecture. Evolutionary conservation is enriched at the Nvj3-Mdm1 interface. The predicted conduit shares geometric and physicochemical properties with bridge-like lipid transfer proteins, including Atg2, Fmp27, and Hob2, suggesting that heteromeric tether assemblies may contribute directly to inter-organelle lipid transfer. Cophylogenetic analysis reveals coordinated coevolution of Nvj3 and Mdm1 across Saccharomycetes. Together, these findings define Nvj3 as a structural partner of Mdm1 and support a conduit-based model of lipid transfer at the NVJ.

Cryo-electron microscopy (cryo-EM) of biological specimens is limited by radiation damage and a low signal-to-noise ratio (SNR). Here, we show that reducing the illuminated area substantially slows the observed diffraction decay in protein microcrystals. We further show that narrow parallel-beam electron diffraction from thin non-crystalline biological specimens provides substantially higher reciprocal-space SNR than conventional cryo-EM imaging. We developed a multimodal scanning workflow, 4D-para-STEM, that records narrow-beam diffraction patterns together with corresponding images. Using viruses, peptide assemblies, and microtubules, we demonstrate interpretable diffraction signals from both crystalline and non-crystalline biological specimens. Together, these results show that narrow parallel-beam scanning reduces observed radiation damage and improves the SNR in cryo-EM.

T cell receptor recognition of peptide-MHC depends on sequence, interface chemistry, and three-dimensional geometry, but docking geometry is often summarized at the whole-receptor level, leaving CDR3-local pose difficult to compare across structures. We introduce TCR-FramePose, a local-frame descriptor set that represents each TCR-pMHC complex as three bodies - whole TCR, CDR3a, and CDR3b - measured relative to a pMHC groove frame. For each body, FramePose decomposes the native pose into reach, offset direction on S^2, and orientation on SO(3); for tangent-space analyses, these components are mapped to six coordinates per body and 18 coordinates per complex. Applied to 378 curated abTCR-pMHC crystal structures, FramePose recovers known class-associated receptor-placement differences and additionally resolved whole-TCR and CDR3b orientation shifts that were not captured by crossing angle. The same orientation coordinates identified reverse-polarity and off-axis outliers as distinct modes. In cross-validated association analyses, FramePose added nonredundant BSA- and affinity-associated information beyond conventional descriptors, and the modest affinity gain was concentrated in CDR3 orientation blocks which were least recoverable from conventional descriptors. Biological grouping analyses showed that shared receptor pose over peptide-MHC was organized primarily by germline V-region framework. TCRs recognizing the same peptide-MHC target favors shared FramePose geometries rather than strong receptor-specific divergence, whereas CDR3 sequence did not detectably reposition the rigid-body pose after antigen context and germline framework were fixed. MHC allele and peptide length contributed smaller adjustments, localized mainly to CDR3b and groove-normal orientation axes. Finally, interface analyses showed that affinity tracked interface burial, with CDR3b reach linking FramePose geometry to binding through buried surface area. Within engineered panels, mutation-level effects were panel-specific, with CDR3b remodeling localizing to a recurrent interface region but varying in direction across receptors. These properties enable FramePose to serve as a geometric filter for in silico TCR-pMHC models and as a feature layer for structure-guided TCR engineering. Together, TCR-FramePose provides a nonredundant geometric layer for structure-guided TCR-pMHC analysis, linking germline-scaffolded recognition, CDR3-local pose, and interface organization without replacing sequence, contact, or energetic descriptors.