Structure

Upon RAS-driven membrane recruitment, RAF kinases ARAF, BRAF and CRAF are activated via formation of homo- or hetero-dimers to initiate signaling through the MAP kinase cascade. Although RAF heterodimers are important for both physiologic and oncogenic signaling, they have been little studied at a structural and biochemical level. Here we report the preparation, biochemical characterization, and the cryo-EM structure of a 14-3-3-bound BRAF/CRAF heterodimer complex. The heterodimer exhibited kinetic parameters and sensitivity to a panel of twelve structurally diverse RAF inhibitors that were closely similar to, or intermediate between, those of BRAF and CRAF homodimers. Cryo-EM structures of the heterodimer with and without MEK1 revealed an overall organization essentially identical to that of RAF homodimers, but with an asymmetric interaction in the MEK1-bound structure in which the BRAF N-terminal acidic (NtA) motif extends across the dimer interface to engage the CRAF RKTR motif. Mutagenesis of this interface unexpectedly revealed that replacing the acidic NtA sequence with a basic RARA sequence yields highly active RAF homodimers and heterodimers, demonstrating that negative charge in the NtA motif is not required for activity. Collectively, our findings suggest that the charge state of the NtA motif influences RAF activity through effects on local backbone dynamics and the stability of the inactive kinase conformation, rather than via stereospecific recognition across the dimer interface.

Double-stranded DNA minicircles have been observed in a variety of biological settings and are also widely employed in biotechnology, therapeutic applications, and basic research. Here, we report a cryo-EM structure of a 95-basepair minicircle (dsMC95) at a 5.3 [A] resolution. dsMC95 forms a closed ring as designed and no local deformation is observed. The two DNA strands are fully resolved, with the major and minor grooves clearly distinguishable. Analysis reveals a nine-fold periodicity in the helical twist, which corresponds to approximately 10.56 base pairs per turn. Together with groove width analysis, the data indicate that dsMC95 maintains a B-DNA configuration. The dsMC95 ring exhibits an in-plane ellipticity of 1.13 and an out-of-plane displacement of 15{degrees}, with differences in out-of-plane displacements observed between the two half-segments. The dsMC95 structure, which is the only free DNA cryo-EM structure with a resolution better than 6 [A] to date, allows comparison to other structures to better understand DNA physical features such as bending. The findings advance our understanding of DNA structure under topological constraints and may inform studies of naturally occurring small circular DNA as well as the manipulation of DNA in nanotechnology applications.

Mutations in MAPT, the tau gene, give rise to forms of frontotemporal dementia and parkinsonism linked to chromosome 17 (FTDP-17T), with abundant filamentous tau inclusions in brain cells. Some mutations that encode missense and deletion variants can give rise to a clinical picture of Picks disease and filaments made of three-repeat tau. Here we report the electron cryo-microscopy (cryo-EM) structures of tau filaments from individuals with MAPT mutations D252V, G272V, S320F and {Delta}G389-I392. The two-layered Pick fold was present in the individuals with mutations D252V and {Delta}G389-I392. By contrast, mutations G272V and S320F gave rise to a more open variant of the Pick fold, with residues 272-341 rotated by 20-25{degrees} with respect to the rest of the structure. These findings show that missense mutations within the filament core can modify the Pick fold, generating closely related structural variants. In addition, we were able to reconstitute the Pick fold and some of its variants using seeded assembly with recombinant 0N3R tau carrying 12 serine or threonine to aspartate substitutions (PAD12) and missense mutations D252V, G272V or S320F. This work provides a foundation for the development of structure-based diagnostic and therapeutic approaches.

In vertebrate phototransduction, the G protein-coupled receptor rhodopsin activates the -subunit of transducin (GT), which, upon binding the {gamma} subunits of phosphodiesterase-6 (PDE6), stimulates the hydrolysis of cGMP. We reported a cryoEM structure for a complex containing two constitutively active GT (GT*) subunits coupled by a bivalent antibody bound to PDE6 that demonstrated a striking displacement of both PDE{gamma} subunits from the PDE/PDE{beta} catalytic sites and suggested an alternating-site mechanism for PDE6 activation. Here, we use site-directed spin labeling (SDSL) and double electron-electron resonance spectroscopy (DEER) to probe PDE6 conformational changes upon GT* binding. Both spin-labelled Cys68 on wild-type PDE{gamma} and a spin-labelled cysteine residue substituted for Ile64 on PDE{gamma} demonstrate that PDE{gamma} has highly flexible C-termini that transiently bind to the PDE/PDE{beta} heterodimer. Binding of GT* to PDE6 with the competitive inhibitor udenafil occupying its catalytic sites alters the positions of the PDE{gamma} subunits in agreement with the striking changes shown in the cryoEM structure for this complex, whereas coupling the GT* subunits to the bivalent antibody does not affect the DEER distributions observed for PDE6 bound to GT*. However, binding of the slow hydrolyzing 8-Br-cGMP substrate in the presence of GT* causes a dramatic increase in the separation and spread of the spin-labelled PDE{gamma} subunits, thereby revealing a previously unobserved conformation of PDE6 associated with catalysis. These studies indicate that whereas inhibitors trap GT*-PDE6 complexes in an inactive state as represented by the cryoEM structure, the binding of both substrate and GT* produces a dynamic active state consistent with an alternating site mechanism.

G protein-coupled receptors (GPCRs) are critical regulators of human physiology and major drug targets. Although structural studies have provided valuable insights, determining GPCR structures remains challenging, especially for inactive state receptors. Recent advances in cryo-electron microscopy (cryo-EM) have enabled structural determination of small GPCRs by using fusion partner proteins and binders to increase molecular weight. However, current methods require extensive experimental screening of fusion constructs. Widely adopted strategies, such as BRIL-Fab complexes, also face limitations due to inherent flexibility. Here, we introduce a streamlined and universal pipeline that integrates an in silico fusion construct screening program, NOAH (NOAH: NOn-experimental, AI-assisted High-throughput construct screening), with a de novo designed fusion protein called ARK1 (ARtificially-designed fiducial marKer). We validate the efficacy of NOAH by determining the structures of the vasopressin V2 receptor (V2R) bound to the clinical antagonist tolvaptan and the partial agonist OPC51803, as well as the bradykinin B2 receptor (B2R) bound to the clinical antagonist icatibant, thereby elucidating their activation and deactivation mechanisms. Furthermore, we demonstrate the capability of NOAH-ARK1 by solving the tolvaptan-bound V2R structure at higher resolution and showcase the methods versatility by determining the structure of lysophosphatidic acid receptor 2 (LPA2) bound to the antagonist Ki16425. This approach eliminates the need for time-consuming and labor-intensive construct optimization, providing a rapid and widely applicable solution for high-resolution GPCR structure determination and drug discovery.

The brain-derived neurotrophic factor (BDNF)-tropomyosin receptor kinase B (TrkB) signalling axis is a key effector of synaptic plasticity and neuroprotection. While TrkB activation is a major objective towards preventing dysfunction of the nervous system, it cannot be reached with exogenous BDNF administration given the unfavourable physiochemical properties of BDNF. In addition, BDNF also activates a tumour necrosis factor pathway by binding to the neurotrophin receptor p75. The TrkB agonist ZEB85 provides an alternative route to the selective activation of TrkB. We report here the structural basis for the interaction between human TrkB, and both ZEB85 and BDNF, and reveal that a sulfated tyrosine modification is indispensable for ZEB85 activation of TrkB signalling. Using structure-guided BDNF- and ZEB85-binding deficient TrkB mutants, we assessed their ability to sequester ligands from full-length TrkB in cultured human neurons. We found that the BDNF binding site extends into the extracellular juxtamembrane domain of TrkB but does not require the sulfotyrosine at residue 400 to activate TrkB. Together with biophysical analysis and AlphaFold modelling these results also explain how BDNF can displace ZEB85 from TrkB through an overlapping epitope. Our findings reveal unique features of TrkB, not present in the related neurotrophin receptors TrkA and TrkC, and suggest new directions to explore the role of sulfotyrosine in TrkB signalling and identify new TrkB-specific protein ligands. One Sentence SummaryInvestigation of the mechanism of action of TrkB agonist ZEB85 extends molecular understanding of TrkB activation.

Phosphatidylinositol 4-kinase alpha (PI4KA) is an essential lipid kinase that generates phosphatidylinositol 4-phosphate (PI4P) from phosphatidylinositol (PI) at the plasma membrane (PM). PI4P is a precursor for PIP2 and PIP3 lipid signalling, with PI4P serving a critical role in maintaining PM identity and asymmetry. Given the important roles of PI4KA in myriad processes, understanding how it is regulated is of immense importance. Here, we have identified that PI4KA can be phosphorylated in its dimerization domain (pY1154) and kinase domain (pY2090) by a cohort of tyrosine kinases. Phosphorylation of pY2090 significantly impairs lipid kinase activity of PI4KA but does not alter its recruitment to membranes by its regulatory protein EFR3. Cryo-EM and HDX-MS analysis reveals that phosphorylation does not result in dramatic conformational changes but instead causes localised changes in the k12 C-terminal helix of PI4KA. Phosphorylation of the C-terminal helix is found in multiple PI3Ks and PI4Ks, suggesting this may be a evolutionarily conserved regulatory mechanism. Overall, our work reveals a novel regulatory mechanism for PI4KA that directly alters its lipid kinase activity.

The emergence of single-particle cryoEM as a powerful method for structure determination has in large part been fueled by its ability to resolve both single static structures and complex conformational landscapes. Indeed, modern approaches to the heterogeneous reconstruction task can resolve 100s-1,000s of different maps from a single cryoEM dataset. How accurate these algorithms are, however, has proven difficult to rigorously assess, due to a lack of suitable benchmark datasets containing both realistic noise features and ground-truth labels. To address this obstacle, we recently developed a series of benchmark datasets that leverage the targeting power of Cas9 and the programmable heterogeneity of DNA to newly offer access to ground-truth per-particle structural labels in real data. Here, we challenged two popular heterogeneous reconstruction algorithms with mixed particle stacks resampled in silico from these datasets, finding that existing approaches resolve the encoded heterogeneity with limited accuracy. In particular, in realistic particle stacks with complex, multi-scale, and multi-axis heterogeneity, we observed that reconstruction of encoded heterogeneity depended strongly on the application of prior information about where heterogeneity was expected, and that individual particle assignments were made with significant error even when the correct structural states were reconstructed. Both molecular breathing motions and data collection features, such as defocus and projection angle, contributed to the observed particle assignment error. These results highlight important shortcomings of existing heterogeneous reconstruction methods and suggest new avenues for method development in both data collection strategies and in heterogeneous classification and reconstruction algorithms.

{beta}-turns are among the most common structural motifs in proteins, yet their conformational dynamics and sequence determinants remain incompletely understood. Here we present a data-driven classification and dynamic analysis of {beta}-turn conformations using large-scale molecular dynamics trajectories from the mdCATH database. Clustering of backbone dihedral angles using a cross-bond Ramachandran representation identifies six {beta}-turn types, including a previously uncharacterized hybrid I/I' cluster that combines geometric features of canonical type I and I' conformations. Time-resolved analysis indicates that this hybrid state acts as a transient intermediate state of {beta}-turns. Transitions observed in molecular dynamics simulations closely match NMR ensembles and altlocs detected in X-ray crystal structures, with the most dominant exchanges occurring between type I and II, and between type I' and II' turns. Sequence analysis shows that each turn type exhibits characteristic amino acid preferences at the central residues (i + 1 and i + 2). Within these overall preferences, specific residue pairs display distinct biases toward static or dynamic behavior. Targeted in silico substitutions that interchange dynamic- and static-enriched residue pairs shift the conformational behavior of turns accordingly, providing direct support for these sequence-dynamics relationships. Analysis of flanking secondary-structure environments reveals that structural context further modulates turn flexibility, with strand- and coil-associated turns exhibiting higher dynamic propensity than helix-associated turns. Together, these results reveal how sequence composition and structural context jointly shape the conformational landscape of {beta}-turns.

The air-water interface (AWI) remains the primary barrier to routine high-resolution cryo-EM structure determination, driving protein adsorption, structural denaturation, and restricted particle orientations during vitrification. Here, we describe a simple and broadly applicable strategy to mitigate these effects using the mild non-ionic detergent n-decyl-{beta}-D-maltopyranoside (DM). Addition of DM at low millimolar concentrations immediately prior to vitrification consistently suppresses AWI-driven artifacts, resulting in improved angular sampling, reduced structural damage, and enhanced reconstruction quality across diverse macromolecular systems. Using this approach, we obtained a high-resolution reconstruction of the 65 kDa Nucleophosmin 1 pentamer, a target previously limited by severe preferred orientation issues. We further show that DM promotes isotropic particle distributions for high-resolution reconstruction of hemagglutinin, transthyretin, as well as suppressing denaturation of aldolase while stabilizing its C-terminus. Our results indicate that DM effectively passivates deleterious air-water interface interactions without compromising particle integrity. These results establish DM as an effective additive for improving the robustness of single-particle cryo-EM sample preparation. O_FIG O_LINKSMALLFIG WIDTH=174 HEIGHT=200 SRC="FIGDIR/small/716008v1_ufig1.gif" ALT="Figure 1"> View larger version (56K): org.highwire.dtl.DTLVardef@108a6edorg.highwire.dtl.DTLVardef@10728b4org.highwire.dtl.DTLVardef@1014b2eorg.highwire.dtl.DTLVardef@1eed745_HPS_FORMAT_FIGEXP M_FIG C_FIG

Microtubules are dynamic filaments of tubulin heterodimers that comprise an essential part of the eukaryotic cytoskeleton1. The nucleotide state of tubulin controls microtubule dynamics: stable GTP-microtubules favor polymerization, whereas unstable GDP-microtubules drive depolymerization2. Anticancer compounds such as Taxol (paclitaxel) target microtubule dynamicity by preventing microtubule depolymerization3,4. Despite decades of work, the molecular basis of microtubule dynamics remains poorly defined. Using cryo-EM, we determined [~]2.2 [A] structures of human microtubules in GTP-like (GMPCPP) and GDP states. Comparison of these two states revealed switch-like structural changes as tubulins transition from the pre-hydrolysis (GMPCPP) to the post-hydrolysis (GDP) state. Additional structure determination of Taxol-bound microtubules at [~]2.2 [A] showed that Taxol binding converts the microtubule lattice into a pre-hydrolysis state by reversing the structural switches flipped during GTP hydrolysis. Focusing our analysis on the microtubule seam shows that the pre-hydrolysis conformation of GMPCPP or Taxol-GDP exhibits favorable lateral interactions at the seam, with lattice deformations clearly visible at the GDP seam. Together, our data show the existence of structural switches in tubulin that are coupled to the nucleotide state and are exploited by Taxol to stabilize microtubules into a pre-hydrolysis-like state. (191 words)

The voltage-gated potassium channel hERG (Kv11.1) plays a central role in cardiac repolarisation by mediating the rapid delayed rectifier K+ current (IKr). Blockage of hERG by small molecules can lead to delayed repolarisation, QT interval prolongation, and potentially fatal arrhythmias, making the channel a critical focus in drug safety screening. Despite extensive pharmacological and electrophysiological characterisation, a complete structural understanding of hERG gating remains limited by the absence of an experimentally determined closed-state structure. Here, we use AI-based structural modelling to predict and compare candidate closed conformations of hERG. Building on recent work in which AlphaFold2 (AF2) predictions guided by engineered structural templates captured closed and inactivated states, we applied the emerging protein structure predictor, Chai-1, which employs a single-sequence, language model-based approach independent of multiple-sequence alignments. The resulting Chai-1 hERG model was compared with the AF2-derived closed structure, a homology model based on the Rattus norvegicus EAG channel, and an experimentally resolved open-state cryo-EM structure. We assessed these models using a combination of all-atom and coarse-grained molecular dynamics simulations, analysing protein dynamics, pore geometry, gating residue orientation, hydration, and lipid interactions. The Chai-1 and AF2 models displayed strong structural and dynamic agreement, both adopting compact, non-conductive conformations consistent with a physiologically closed state. Our data reveal insights into VSD dynamics, as well as suggesting a state dependence for ceramide binding at the previously identified M651 residue. Our findings support the validity of AI-derived closed-state hERG models and underscore the growing potential of deep learning-based protein structure prediction to identify previously uncharacterised, pharmacologically relevant conformations of membrane proteins. Further, our Chai-1 derived closed state model expands our structural insights into hERG gating and may have utility for investigation of drug-hERG interactions.

Single-particle cryogenic electron microscopy (cryo-EM) has enabled near-atomic resolution structure determination of diverse biomolecules. Because the high vacuum required for electron microscopy prevents the imaging of liquid-phase samples, cryo-EM samples are prepared by plunging the sample into a cryogen, rapidly cooling the sample and suspending the ensemble of biomolecules in a matrix of water glass. However, the effects of this vitrification on the biomolecular ensemble are unknown, complicating efforts to use cryo-EM to derive conformational ensembles of biomolecules. To study these effects, we carried out extensive molecular dynamics simulations (over 50 milliseconds) of the Trp-cage miniprotein at equilibrium and undergoing rapid cooling. We simulated seven cooling rates spanning three orders of magnitude, with the slowest coolings matching experimental rates. By inspecting molecular mobility and density-temperature equations of state for water with and without protein, we found that water vitrification is unaltered by the protein. To track protein conformation changes, and to relate them to conformational kinetics, we made a Markov State Model (MSM) of Trp-cage from 5.4 milliseconds of equilibrium sampling at 277 K. We observed that MSM states with a characteristic time longer than the duration of the non-equilibrium cooling, tend to be more robust to artefacts induced by such cooling. Critically, although we observe that some states vanish in the equilibrium ensemble at 230 K, none do in our nonequilibrium cooled ensembles. However, to account for perturbations induced by nonequilibrium cooling for more labile states, we developed a thermodynamic inference framework for recovering equilibrium populations from the measured vitrified ensembles. These results indicate that cryo-EM has the capacity to be a reliable and accurate biophysical technique for the study of biomolecular ensembles. SignificanceCryogenic electron microscopy images biomolecules trapped in vitreous ice. To vitrify the sample, it must be cooled over the course of 22 microseconds. However, the degree to which this cooling causes the ensemble of the molecules to be perturbed from equilibrium is unknown. Here we present extensive molecular dynamics simulations to quantify the equilibrium dynamics of the Trp-cage miniprotein and the effects of cooling on its conformational ensemble. By simulating cooling at seven different rates, including the slowest experimental rates that still result in vitrification, we connect the kinetic properties of a proteins conformational state to the change in state population from cooling. We show that cooling-induced population shifts are small but observable. We further introduce a thermodynamic-inference method to recover equilibrium populations from the cooled ensembles.

Macromolecular crystallography is limited by the phase problem: diffraction experiments measure amplitudes but not the phases required to reconstruct electron density. Existing phasing routes usually seek enough continuous phase information for density modification and model building to converge. Here, we ask how much phase information can be discarded while preserving convergence. We analyzed 14,148 diffraction datasets from chiral crystals to characterize centric reflections in reciprocal-space asymmetric units. After conditioning by centric trace and, where required, index parity, the two theoretical symmetry-allowed phase values were populated near equally, close to 50:50, independent of space group, defining a compact symmetry scaffold. We then retained this exact scaffold while compressing reference acentric phases to a one-bit alphabet {0, {pi}}; as expected from their diffuse parent distribution, the assignments were also near-balanced. Although this binary representation, with fixed attenuation 2/{pi}, introduces large angular errors (mean of 52{degrees}), it frequently supported automated structure solution: in paired Phenix AutoBuild tests, 705 of 894 binary initializers met a conservative joint criterion of final Free R [≤] 30% and relative chain recovery [≥] 70%, within a 20.0-2.5 [A] resolution window. To rank candidate seeds without rebuilding, we developed a branch-balanced Basin Score from inexpensive density-modification and map-connectivity observables computed at 20.0-3.5 [A]. The empirical score quickly separates productive from unproductive initializers before AutoBuild. Controlled phase inversion shows that basin compatibility decays gradually and can reappear in an anti-phase-related branch, indicating that buildability is not confined to a single neighborhood around the reference phase set but extends to a much broader field. These results recast phase initialization as basin entry and support future symmetry-aware, binary phase-search strategies.

The coronavirus envelope (E) protein is a viroporin that plays a key role in viral assembly, release, budding and pathogenesis. E protein forms oligomeric ion channels that can activate immune responses. However, high-resolution structural data for its extramembrane domains is limited. The C-terminal domain of SARS-CoV has been shown previously to form amyloid fibers, and here we show that these fibers can modulate the shape of liposomes. The propensity to form fibrils, and their effect on liposomes, was examined for sequences belonging to the four clades of coronaviruses. Electron microscopy data shows that the C-terminal domain in E protein adopts a filamentous structure. These findings demonstrate the potential of these peptides to modulate membranes, providing a possible mechanism by which E protein interacts with membranes in the host cell.

Presynaptic inhibitory GPCR (Gi/o GPCR) signaling is an essential regulatory mechanism in vertebrate physiology. Near the presynaptic active zone, Gi/o GPCR activation releases G-protein {beta}{gamma} heterodimers (G{beta}{gamma}) which act to inhibit synaptic vesicle fusion through either modulation of Ca2+ entry via voltage-gated Ca2+ channels, or by direct interactions with the core exocytotic machinery comprised of the ternary SNARE complex downstream of Ca2+ influx. The precise molecular mechanism underlying G{beta}{gamma}-SNARE mediated inhibition has remained unclear due to lack of structural data for the G{beta}{gamma}-SNARE complex. We address this long-standing question here by stabilizing the interaction between G{beta}1{gamma}2 and a pre-fusion ternary SNARE mimetic and determining the structure using single-particle cryo-EM. We used our cryo-EM envelope to build an atomic level prediction of the interaction interface. We validated key interaction sites predicted by our model at the C-terminus of SNAP-25 using site directed mutagenesis and biochemical affinity measurements. Additionally, we found that G{beta}1{gamma}2 and a fragment of the regulatory protein complexin can engage the pre-fusion SNARE complex simultaneously. On the basis of these results, we propose a model in which G{beta}1{gamma}2 acts on the partially zipped SNARE complex at a late stage in the vesicle docking and priming cycle. In the model, the amino-terminal coiled-coil of G{beta}1{gamma}2 forms an interface with the C-terminus of the target membrane SNARE (t-SNARE) complex to prevent complete incorporation of the vesicle SNARE (v-SNARE) into the core SNARE helical bundle, thus blocking vesicle approach to the plasma membrane. The {beta}-propeller domain of G{beta}1 may also sterically hinder vesicle approach. Together these results provide crucial structural insights into the mechanism of binding of G{beta}{gamma} to the SNARE complex, and lends essential insights into the critical role of GPCR signaling to the SNARE complex in modulating synaptic vesicle fusion.

TRPC3 is a calcium-permeable, non-selective cation channel that is activated by DAG. It is expressed in several tissues, especially in the cerebellum, and has been implicated in various human diseases. Despite recent progress in understanding the structural mechanism of TRPC3, how the channel opens remains elusive. Here, we present structures of hTRPC3 in an agonist-free resting state, determined using a DAG-binding site mutant. We also present the structure of hTRPC3 in a DAG-bound open state, determined using a constitutively active "moonwalker" (T561A) mutant. These structures, together with electrophysiological results, reveal that the T561A mutation activates hTRPC3 by disrupting a polar interaction with N652. A newly formed {pi}-bulge in S6 leads to rotation and outward tilting of the lower half of S6, resulting in dilation of the pore and thus channel opening. Agonist DAG stabilizes hTRPC3 in the open conformation. BTDM exerts its inhibitory effect by pushing S5 and S6 back to the center to close the pore, while preserving the {pi}-bulge. These results shed light on the opening mechanism of hTRPC3.

Adenylyl cyclases (ACs) convert ATP into the second messenger cAMP, thus directly influencing cellular signaling in response to a wide variety of stimuli. Despite their physiological importance, structural studies of isoform-specific AC regulation are compounded by difficulties in AC expression and purification. Here, we designed a chimeric construct AC95, combining human AC9 as a molecular scaffold and incorporating the catalytic-allosteric core of human AC5. Cryo-EM analysis of AC95 at 3.5 [A] resolution revealed a state of AC95 bound to both ATPS and forskolin, demonstrating that the chimera reproduces AC5-like allosteric regulation while retaining the structural features of the AC9 scaffold. Although AC95 chimera retained the ability to bind to and be activated by forskolin, it lost the ability to be autoinhibited by the C2b domain of AC9. Moreover, AC95 is insensitive to inhibition by specific AC5 inhibitors SQ22,536 and NKY80, suggesting that these molecules may target a site distinct from the catalytic-allosteric core of AC5 grafted into the AC95 chimeric construct. Our results establish a generalizable approach for investigating isoform-specific regulation of membrane ACs by small molecules, offering a potential path for structure-based drug discovery targeting distinct AC isoforms.

Although machine learning has transformed protein structure prediction of folded protein ground states with remarkable accuracy, intrinsically disordered proteins and regions (IDPs/IDRs) are defined by diverse and dynamical structural ensembles that are predicted with low confidence by algorithms such as AlphaFold and RoseTTAFold. We present a new machine learning method, IDPForge (Intrinsically Disordered Protein, FOlded and disordered Region GEnerator), that exploits a transformer protein language diffusion model to create all-atom IDP ensembles and IDR disordered ensembles that maintains the folded domains. IDPForge does not require sequence-specific training, back transformations from coarse-grained representations, nor ensemble reweighting, as in general the created IDP/IDR conformational ensembles show good agreement with solution experimental data, and options for biasing with experimental restraints are provided if desired. We envision that IDPForge with these diverse capabilities will facilitate integrative and structural studies for proteins that contain intrinsic disorder, and is available as an open source resource for general use.

Macromolecular crystallographic refinement underpins structural biology, yet existing software packages often lack accessible, modular codebases amenable to rapid method development. Here, we introduce TorchRef, a PyTorch-based crystallographic refinement framework that exposes all refinable parameters, atomic coordinates, displacement parameters, occupancies, and scale factors to automatic differentiation. The framework implements FFT-based structure-factor calculations, the French-Wilson treatment of intensities, bulk-solvent modeling with established mask parameters, and stereochemical restraints from the CCP4 Monomer Library. A modular target architecture allows loss functions to be combined, weighted, and extended independently of the core refinement machinery. Validation against 1,000 PDB structures demonstrates that TorchRef-based refinement reproduces a median R-free within 1% of Phenix while maintaining comparable model quality. Structure factor calculation in TorchRef scales readily across multiple CPU cores and is over 100 times faster on modern GPUs than CCTBX. To showcase how modern methods like time-resolved crystallography can benefit from the flexibility that TorchRef provides, we implemented direct refinement of a typical time-resolved model against amplitude differences, a use case currently not explored by classic refinement programs. TorchRef is released under the MIT license with full API documentation and tutorials, providing an accessible platform for developing and testing new crystallographic refinement protocols. SynopsisTorchRef is an open-source PyTorch-based crystallographic refinement framework that exposes all refinable parameters to automatic differentiation, delivers GPU-accelerated structure-factor evaluation more than 100x faster than CCTBX, and enables new workflows, such as direct refinement against amplitude differences in time-resolved crystallography.