OpenSynBio? Computing Foundation

AI is changing our lives. We used the LLM notebook to generate a podcast from our blog. Now I am biased but this is incredible. Have a listen & let us know what you think. “A new era of medicine being born”. Using AI to unlock the secrets of RNA. We have a series of blog posts lined up. More podcasts coming! https://lnkd.in/eeRSfb3Z Sandy Kory thought you may enjoy this!

Modeling PROTAC Degradation Activity with Machine Learning Stefano Ribes,?Eva Nittinger,?Christian Tyrchan,?Rocio Mercado Paper:?https://lnkd.in/eY9yGkcy Code:??https://lnkd.in/eXsfvx2B #deeplearning #drugdiscovery #machinelearning

?? ? Achieve stable out-of-core GPU computing and maximum performance with RAPIDS Dask. Tech blog ?? https://lnkd.in/gEPSBX6u Learn how optimal memory settings, accelerated networking with UCX, and multi-hardware backend support make scaling multi-GPU data analysis faster and seamless.

??AI offers immense potential but also biosecurity risks if misused. ??In Nature, our fellow Dr. Jassi Pannu, deputy director Anita Cicero, director Dr. Tom Inglesby, and colleagues stress the need for collaboration among governments, AI developers, and biosecurity experts to make AI safer. ?? Read the full commentary: https://lnkd.in/e-DJWyhi

Wohoo!! ?? You can now generate 3D structures of proteins, RNA, DNA, and small molecules. ?? Check out BOLTZ on Hugging Face! — the first open-source and commercially available model to achieve AlphaFold3-level accuracy in biomolecular structure prediction. ?? Try it yourself on Gradio ! Explore and generate different proteins and viruses : https://lnkd.in/d9wxermm For faster inference (and to try BOLTZ), check it out here (on L4): https://lnkd.in/dZz7rBez (credit to Simon Duer!) ?? Let’s make biology more open on HF as Clem Delangue ?? says! If there’s anything to fear, it’s tech being closed :) ! Cr: MIT's Jeremy Wohlwend - Gabriele Corso ?? - cc Ahsen Khaliq ??

The more AI/ML advances, the clearer it becomes: physics-based methods are essential for generating the high-quality data that drives meaningful progress. My newest blog... #artificialintelligence #computationalchemistry #drugdiscovery

How can AI + Physics Accelerate #Molecular #Glue Discovery? Designing molecular glues requires prior knowledge of specific protein conformations that reveal binding pockets at protein-protein interfaces. These pockets are critical targets for small molecules that induce proximity between proteins. However, identifying such pockets using traditional AI-based protein structure prediction models is challenging because the necessary conformational dynamics occur on microsecond to millisecond timescales. Combining AI and Physics to Identify Targetable Pockets Can we combine AI with physics-based simulations to capture these elusive pockets? Indeed, we can. The interaction between BRAF and MEK1 serves as an example. Under normal conditions, BRAF interacts with MEK1, and both proteins adopt inactive conformations, leading to normal cell growth and division. The BRAF V600E gene mutation is associated with more rapid cancer growth and a higher death rate in papillary thyroid cancer. Impact of the BRAF V600E Mutation: So how does the BRAF V600E mutation affect the BRAF-MEK1 interaction? The V600E mutation shifts BRAF from an inactive to an active state, which in turn activates MEK1 and the MAPK pathway. This activation leads to increased cell proliferation and tumorigenesis. Counteracting Activation with Molecular Glues: To counter this activation, we need to stabilize the BRAF(V600E)-MEK1 complex where both proteins adopt inactive conformations. One way to achieve this is by targeting these proteins with molecular glues that can stabilize their interaction. Our Approach and Findings Using a combination of AI and physics-based simulations, we managed to sample the conformational ensemble of the BRAF(V600E)-MEK1 complex in its native state. This allowed us to find druggable binding pockets at the protein-protein interface that can be targeted by molecular glues. Advancing #Molecular #Glue Discovery Traditionally, molecular glue discovery has been led by serendipity. Our results demonstrate how we can use AI and physics to discover druggable binding pockets in protein-protein interfaces. This method opens new avenues for developing molecular glues that can alter conformational ensembles and restore normal cellular function. Data To build a molecular glue targeting BRAF(V600E)-MEK1, you can utilize our conformational ensemble data and corresponding pocket analysis. Link: https://osf.io/ehx2b/ Connect with Us If you are working on molecular glues and want to understand how the dynamics of protein-protein interactions play a role in cellular signaling—or if you wish to discover druggable pockets that can be targeted by small-molecule glues—please feel free to reach out. #AI #ProteinDynamics #compchem #molecularglue #cancer #CancerResearch

Can Generative AI replace traditional Molecular Dynamics simulations? Despite significant advancements, why do machine learning models face challenges similar to traditional methods, and how can researchers overcome them? Our latest article discusses this hot topic in drug discovery, exploring successful applications, technological nuances, and our industry insights. Discover how AI is transforming this field, and connect with us to share your thoughts or collaborate! #ai #md #moleculardynamics #artificialintelligence #generativeai #drugdiscovery #biotech #pharmainnovation #ml #machinelearning

I recently had the privilege of presenting to NASA GeneLab's Alzheimer’s Disease and Brain Resilience Analysis Working Group, where I shared insights on multi-omics data analysis. In my latest article on?Decoding Biology?(linked below), I dive into the key concepts and takeaways from my talk. Spaceflight profoundly impacts biological systems, influencing multiple "omics" layers: genomics, epigenomics, transcriptomics, proteomics, metabolomics, and phenomics. Each layer provides unique insights into how the body responds to the extreme environment of space—microgravity, radiation, isolation, and altered circadian rhythms. These "omics" data types also capture biological information across different spatial and temporal scales, from long-term genomic stability to short-term metabolic responses. To understand how stressors like microgravity and spaceflight impact biological systems—and to effectively leverage this data for identifying biomarkers—we need to integrate these diverse layers of information. In this article, I explore: -The different multi-omics data types and their unique insights - The time scales on which these changes occur - Potential cause-and-effect relationships between the layers Check out the full article to learn more: https://lnkd.in/e2MkcX5Q #compbio #bioinformatics #omics #biotech #datascience #spaceflight

I have posted a bunch of NVIDIA's AI drug discovery applications but we are also working on large language models that help with multi-modal literature review. Check out this technical blog with work done in colab with Indian Institute of Technology, Madras that highlights a multimodal LLM that can do visual question answering using neuroscience images ?? https://lnkd.in/gW8hNwtk