FutureGRIN NextGen

Reliable computational methods for discovering compounds with molecular glue properties is somewhat of the holy grail in the TPD field, and two recent papers have explored this approach. Geoffrey and colleagues developed a workflow they call (MOLDE) Molecular Glue-Design-Evaluator. The approach is based on protein-protein docking, and shortlisting of docked poses based the presence of a ligandable pocket in the PPI interface. This is followed by AI-driven design of a predicted glue compound fitting into the pocket. The authors exemplify the approach with DDB1 and CDK12, a well-studied protein pair, known to be glued by the small molecule degrader R-CR8 (incorrectly referred to as RC8 in the paper). In applying the workflow, the authors identify SAIT_MD_26121 as a novel molecular glue for this system, with a similar scaffold as R-CR8, and a fair few additional stereocenters. Overall, the described approach is attractive since it incorporates novel compound discovery into the methodology. Whilst the overall approach appears reasonable on the surface of it, it is noteworthy that the paper contains errors and inconsistencies. The second study is a pre-print on BioRxiv, where the authors develop YDS-Ternoplex, based on AlphaFold 3-type models. The authors validate the system on a number of different ternary complexes, for example the VHL-CDO1 complex glued by the Novartis compound. No component of novel glue predictions was included in this study. Whilst the above two studies certainly are interesting, an absolute key driver in computationally driven molecular glue discovery is an experimental readout of predicted models, which provide feedback for model refinement and validation. A system that can profile a therapeutically relevant target against a panel of ubiquitin-ligases, and use AI assisted methods for identification of candidate molecular glues for the top-ranking pairings will be a game-changer. Time will tell! Geoffrey et al:?https://lnkd.in/e8U6v_W8 BioRxiv study: ?https://lnkd.in/epXVmK5D #TPD #Drugdiscovery #Innovation?

As a #CADD scientist and a continuous learner in this field, I find the newly introduced #GROMACS_Copilot really interesting. This tool aims to bring automation to MD simulations using GROMACS, making the setup, simulation, and analysis process much easier. link: https://lnkd.in/dWCBKU_y From what I understand, GROMACS Copilot automates key steps like RMSD, RMSF, Rg, H-bonds analysis, and more. This could be a game-changer for those working in molecular dynamics, especially in protein simulations. I'm looking forward to exploring how this tool can improve efficiency in simulation workflows. If you've already tried it, I'd love to hear your thoughts! Known Issues ?? LLM sometimes struggles with selecting the correct group index. Double-checking the selection is recommended. ? The interaction between LLM and gmx prompt input isn't always seamless. Running commands based on suggestions can help you get the correct results more easily. Despite these minor issues, the potential time savings and improved workflow make it a tool worth exploring for anyone working with GROMACS! #MolecularDynamics #GROMACS #ProteinSimulations #ResearchTools #ScienceInnovation #CADD

From a tiny idea in a lab to the pharmacy shelf—ever wondered how medicines are made? ???? It’s a long journey, but AI is helping speed things up! #drugdiscovery #healthcare #artificialintelligence #pharma

At the heart of drug discovery lies the need to understand molecular behaviour at the quantum level. Density Functional Theory (DFT) analysis allows us to explore molecular properties at the quantum level, providing critical insights into drug interactions, stability, and reactivity. Now, imagine combining this with AI—faster calculations, deeper insights, and unprecedented efficiency in drug design. The future of precision medicine is unfolding before our eyes! With DFT, you can predict the electronic structure of molecules. At FutureGRIN NextGen, we integrate DFT to optimize drug design and uncover novel therapeutic solutions. If AI and quantum chemistry could solve one mystery in medicine today, what would you want it to be? Let’s discuss this in the comments! ?? #densityfunctionaltheory #DFT #drugdiscovery #artificialintelligence #healthcare #pharmaceutical #quantumchemistry

Imagine sifting through millions of compounds in record time to find the perfect match for a disease target. That’s the power of virtual screening. Virtual screening isn’t just a step in drug discovery—it’s a game-changer. Do you know the interesting thing? Gone are the days of screening thousands of compounds manually. Today, virtual screening, powered by machine learning, is helping scientists rapidly identify promising drug candidates with higher precision and efficiency. From structure-based to ligand-based screening, AI is optimizing the search for the next breakthrough drug! At FutureGRIN NextGen, we combine ML and advanced algorithms to accelerate the identification of promising drug candidates, saving time and resources in the process. What’s one disease you hope AI helps us conquer in the next decade? Let’s hear your thoughts! ?? #DrugDiscovery #AI #VirtualScreening #PharmaTech #InnovationInHealthcare #Healthcare #Pharmaceutical

Molecular docking is more than just a computational tool—it's the bridge between molecules and breakthroughs. It shows how a molecule/drug interacts with the protein. By simulating how potential drug compounds interact with target proteins, we’re unlocking new possibilities in drug discovery. Aside from helping us learn the interactions, it has been widely used in ranking which molecules are more promising among arrays of choice. It has made drug discovery more fun. Many software have been developed for molecular docking, such as: 1??AutoDock and AutoDock Vina by Scripps Research 2??DockThor by the Grupo de Modelagem Molecular em Sistemas Biológicos (Molecular Modeling Group of Biological Systems) at LNCC 3??GOLD by CCDC - The Cambridge Crystallographic Data Centre 4??FlexX by BioSolveIT 5??Molegro Virtual Docker by Dassault Systèmes BIOVIA 6??Discovery Studio by Molexus At FutureGRIN NextGen, we leverage docking in many of our research, enhancing the speed of predicting potential drug candidates. Have you used molecular docking before? Let’s discuss! ?? #DrugDiscovery #AI #MolecularDocking #InnovationInHealthcare #Healthcare #ComputationalBiology #Pharmaceutical

Some 20 terms in AI-driven drug discovery lately. Which one are you hearing for the first time? #AI #drugdiscovery #terms

Debunking the Myth: AI Will Replace Drug Scientists… There's a common misconception that artificial intelligence (AI) will replace drug scientists in the development of life-saving medications. But the truth is, AI isn't here to replace scientists – it's here to assist them. AI technologies are revolutionizing the drug development process, helping scientists analyze vast datasets, predict molecular interactions, and even speed up clinical trials. However, the human element – critical thinking, creativity, ethical decision-making, and expertise – remains essential in drug discovery. In short, AI cannot, rather, will enhance the work of drug scientists, empowering them to make more informed decisions and push the boundaries of medical advancements faster than ever before. #AI #DrugDiscovery #Pharma #Innovation #Science #Technology #AIinPharma

Rational drug design is hard, rational design of targeted protein degraders, especially molecular glues, is harder. A good degrader must stabilize a ternary complex between the ligase, degrader, and target. This complex often forms a new, flexible protein-protein interface. It must adopt shapes that allow efficient ubiquitin transfer and hold together long enough for adequate ubiquitination. Can we design molecular glue degraders for specific targets? In a new manuscript, we introduce GlueMap. This computational platform combines structural, thermodynamic, and pharmacodynamic modeling to find new molecular glues, provide mechanistic insights into their activity, and guide their design and optimization. We’ve used GlueMap successfully in our own discovery projects and are eager to apply it to even more. https://lnkd.in/egYv7vkE I want to highlight the power of combining physics-based simulations with machine learning. In GlueMap, we trained a variational autoencoder on extensive molecular dynamics simulations to capture a latent structural representation of ternary complexes. This helped us build a model that predicts degradation potency using training data from wet-lab measurements of just five molecular glues. Our results hint that, by pairing simulation data—cheap and quick to generate—with limited experimental data—costly and slow, physics-informed few-shot machine learning may create powerful predictive models. Team work by Jesus Izaguirre, Yujie Wu, Zach McDargh, PhD, Timothy Palpant, Asghar Razavi, PhD, Fabio Trovato, and Cheryl Koh from Atommap

Advancements in AI Technologies have enabled machines to learn from large datasets, improving speed of drug discovery, and disease diagnosis. One of the notable developments is the increasing use of Machine Learning (ML) and Deep Learning (DL) algorithms in healthcare. These technologies have enhanced drug discovery and disease diagnosis, reducing costs, and time. #FutureGRINNextGen #AI #Technology #ML #DL #Drugdisccovery #Diseases