Vecura

🧬 How can researchers screen an entire generation of de novo protein designs before moving to the next iteration? This clip shows how Boltz-2.1 is accessed via API directly within Vecura to fold designed protein sequences and return predicted 3D structures together with confidence metrics. In the complete workflow, #RFdiffusion generated candidate backbones, #LigandMPNN designed the corresponding sequences, and Boltz-2.1 provided structure-confidence signals to support the ranking and filtering of nine designs targeting the ATP pocket of the LRRK2 kinase domain. These scores are used for relative triage, helping researchers identify more credible designs and candidates that may require further iteration. They are not treated as definitive proof that a design will fold successfully. By connecting external APIs with pre-integrated models and tools, Vecura enables research teams to build complete computational workflows without moving files between separate services or maintaining a different environment for every model. ▶️ Watch Boltz-2.1 in action below. ▶️ Read the full workflow and results: https://lnkd.in/ghNrA9yt ▶️ Start your research: https://app.vecura.com #Boltz #ProteinDesign #DeNovoProteinDesign #StructuralBiology #AIforScience #ComputationalBiology #Vecura

Accurate prediction of molecular electronic structure remains a computational bottleneck in early-stage drug discovery. Vecura has onboarded #Psi4, an open-source ab initio quantum chemistry suite supporting over 200 methods, including DFT, MP2, and CCSD(T), for high-accuracy energy, gradient, and frequency calculations. This enables rigorous first-principles simulation of molecular properties, supporting compound screening, reaction energetics, and precise thermochemical profiling without heuristic approximations. Vecura’s secure, no-code Agentic AI platform allows research teams to deploy and orchestrate these computationally intensive quantum workflows instantly, eliminating infrastructure setup. Explore the model in the Vecura blog: https://lnkd.in/g2fMGyE4 Many thanks to Daniel Smith, Lori Burns, Andrew Simmonett, Robert Parrish, and many other co-authors for developing Psi4. #Vecura #AIForScience #ComputationalChemistry #DrugDiscovery #MolecularDiscovery

📢 Excited to share Vecura Biotech Insiders #03, featuring a workflow shared with the Vecura team by Tran Duy Thanh. This workflow explored a practical challenge in structure-based discovery: What should researchers do when the human target is biologically important, but no experimental structure is available? The case focused on human tyrosinase, a copper-dependent enzyme involved in melanin biosynthesis and a key target in pigmentation-related research. Instead of relying only on mushroom tyrosinase as a surrogate, the workflow built a more human-relevant model by: 👉 starting from the AlphaFold human TYR model ➡️ aligning it to a mushroom tyrosinase reference structure ➡️ transplanting the catalytic copper ions ➡️ preparing and minimizing the system in OpenMM ➡️ benchmarking the active site ➡️ validating the model through docking with a six-compound panel The broader lesson: Do not start docking too early. For metalloproteins like tyrosinase, a raw predicted structure may not be chemically complete. Cofactors, active-site geometry, substrate-channel access, and scoring limitations all need to be considered before docking results can be interpreted responsibly. This is an exploratory in-silico workflow, not an experimentally resolved human TYR structure or a validated inhibitor discovery result. But it shows how Vecura can help researchers organize a transparent structure-based modeling workflow from target preparation to docking validation. Thank you, Tran Duy Thanh, for sharing this workflow with the Vecura team. Have you built a workflow on Vecura? Share it with us and use #VecuraBiotechInsiders for a chance to be featured in a future community story. 👉 Read the full feature here: https://lnkd.in/gYRvb89x #Vecura #VecuraBiotechInsiders #Tyrosinase #MolecularDocking #StructuralBiology #ComputationalBiology #DrugDiscovery #AIforScience #LifeScienceResearch

Screening massive compound libraries for molecular similarity or substructure matches is a persistent computational bottleneck in virtual screening. Vecura has onboarded #FPSim2, a high-throughput molecular fingerprint search library built on RDKit. It leverages sublinear bounds to efficiently skip database segments, supporting threshold, top-K, Tversky, and substructure searches. This enables rapid hit expansion and structural analogue identification across multi-million compound databases, accelerating molecular discovery workflows. Vecura provides a secure, no-code Agentic AI environment, letting research teams deploy and test #FPSim2 without complex infrastructure setup. Explore the model in the Vecura blog: https://lnkd.in/gWP-HFUt Many thanks to: ChEMBL Team (European Bioinformatics Institute | EMBL-EBI) - Eloy Félix, Andrew Dalke, Greg Landrum, Roman Bushuiev, authors of FPSim2. #Vecura #AIForScience #DrugDiscovery #MolecularDiscovery #ComputationalBiology

Identifying druggable cavities in protein structures often requires complex geometry-based tools or heavy computational resources. Vecura has onboarded #P2Rank, a machine-learning tool that predicts and ranks ligand-binding sites directly from PDB, mmCIF, or BinaryCIF files. Using a Random Forest model on physicochemical features, it scores solvent-accessible surfaces in seconds on a single CPU, requiring no GPU or sequence alignment. This enables rapid, calibrated binding-probability estimation and can rescore geometry-based pockets to recover sites pure geometry misses, accelerating protein modeling workflows. Vecura delivers this as a secure, no-code Agentic AI platform, letting research teams deploy #P2Rank without infrastructure setup. Read more in the Vecura blog: https://lnkd.in/gb-S_Uc9 Many thanks to: Radoslav Krivak, David Hoksza, authors of P2Rank. #Vecura #AIForScience #DrugDiscovery #ProteinModeling #ComputationalBiology

Predicting solid solubility is a bottleneck in API formulation, often requiring extensive wet-lab experiments. Vecura has onboarded #FastSolv, a solid solubility predictor. It estimates logarithmic solubility (logS) of a solute in a solvent at a given temperature using an ensemble of fastprop-based neural networks with Sobolev training, outputting predictions and uncertainty estimates directly from 2D SMILES descriptors. This accelerates formulation design and compound screening by reducing physical testing needs while providing confidence intervals for data-driven decisions. As a secure, no-code Agentic AI platform, Vecura lets research teams test and deploy models like #FastSolv instantly, without infrastructure setup. Explore the model in the Vecura blog: https://lnkd.in/gnmYFuH9 Many thanks to: Lucas Attia, Jackson Burns, Patrick Doyle, William Green, authors of FastSolv. #Vecura #AIForScience #DrugDiscovery #FormulationDesign #AgenticAI

🧬 How do you turn a cloud of cryo-EM density into a simulation-ready protein structure? In our latest Vecura Insight, we reconstruct the active human sweet taste receptor, TAS1R2/TAS1R3, from a 4.28 Å density map for which no fitted atomic model was available. The workflow combines AI-assisted density enhancement, atom tracing, structural modeling, and molecular dynamics to build a full-length, sucrose-bound receptor model, while clearly distinguishing experimentally observed features from computationally constructed ones. The article also explores a glucose-bound state through molecular dynamics, showing why a single visually convincing simulation frame is not enough. Stability, receptor conformation, and binding-pocket integrity must be evaluated together. The broader lesson is simple: modern structural biology advances by combining experimental evidence, computational inference, and rigorous validation. A model’s reliability depends on being explicit about what was observed, what was constructed, and what still requires further validation. 👉 Explore how experimental evidence and computational modeling can work together to advance structural biology: https://lnkd.in/gbbTAWh7 👉 Explore the workflow: https://lnkd.in/grZwhaEp #CryoEM #StructuralBiology #ProteinStructure #MolecularDynamics #AIforScience #Vecura

Thank you to everyone who joined our technical webinar on NVIDIA #BioNeMo and generative AI in life sciences, and a special thank you to Dr. Ying-Ja Chen and our host Giang Nguyen for walking us through the platform. The discussion highlighted the growing impact of generative AI in computational biology, how NVIDIA's framework supports molecular design, and what researchers should consider when integrating these advanced tools into practical workflows. Key themes included: ▪️ The framing and impact of generative AI in life sciences ▪️ How BioNeMo’s pillars support discovery workflows ▪️ Practical case studies showing BioNeMo in action ▪️ Opportunities and limitations of AI‑driven biology Recording: https://lnkd.in/gMaFspQv Explore Vecura: https://vecura.com/en

Designing high-affinity VHH (nanobody) binders traditionally requires weeks of experimental library screening. 𝗺𝗕𝗘𝗥 𝗢𝗽𝗲𝗻 is now available on Vecura. It generates de novo VHH sequences by optimizing AlphaFold-Multimer backpropagation with ESM2 sequence conditioning, yielding structurally confident binder-target complexes. This accelerates protein modeling and molecular discovery by compressing the design-to-candidate timeline, providing directly testable sequences and relaxed 3D structures. Vecura offers a secure, no-code Agentic AI platform to test and deploy this framework without complex infrastructure setup. Explore the model in the Vecura blog: https://lnkd.in/gWeJd4t5 Many thanks to: Erik Swanson, Mike Nichols, Supriya Ravichandran, Pierce Ogden, authors of mBER Open. #Vecura #AIForScience #ProteinModeling #DrugDiscovery #AgenticAI