Psivant Therapeutics

Can quantum mechanics improve the accuracy and extensibility of machine learning models for drug discovery? We believe that better representations of molecules will produce more accurate ML models with less data. Traditionally, ML property prediction models (e.g. solubility, permeability, bioavailability, clearance, etc) use simple 1D or 2D molecular representations (e.g. SMILES, fingerprints, graphs, etc). However, we all know that molecules are not 1D or 2D objects. Quantum mechanics is the ground truth for molecules and we believe that describing molecules with QM will play a critical role in building more accurate ML models with a broader range of applicability. At Psivant Therapeutics, we are committed to designing better small molecule therapeutics for patients with unmet needs in inflammation and immunology. By combining physics-based simulations and machine learning, we tackle critical bottlenecks in drug discovery, targeting areas where antibodies have shown clinical success, but no small molecule drugs exist. Recently, we had the privilege of hosting Todd Martinez, the David Mulvane Ehrsam and Edward Curtis Franklin Professor of Chemistry at Stanford University, for a 1-month mini-sabbatical. As a pioneer in #quantum chemistry calculations, Todd collaborated with our team to explore how quantum mechanics can enhance the predictive power of #machinelearning (#ml) models in #drugdiscovery. Our approach? Leveraging quantum mechanical properties to improve molecule embeddings—the way molecules are represented for machine learning property predictions. By enhancing these representations, we aim to push the boundaries of what's possible in computational drug design and better predict critical properties like solubility, permeability, bioavailability, clearance, binding affinity, and selectivity. This collaboration exemplifies our philosophy: combining world-class expertise, cutting-edge technology, and a relentless focus on patient needs to advance the next generation of therapeutics. We're always looking to engage with leaders and innovators in computational chemistry, quantum chemistry, molecular dynamics, statistical thermodynamics, and machine learning. Let's shape the future of drug discovery together!

I will be part of the upcoming Google Cloud and EPAM event: "Harnessing AI to Accelerate Drug Discovery and Clinical Operations" We at Psivant Therapeutics work closely with Google Cloud and EPAM Systems, with great results. The event is free, but requires registration (link below). I am looking forward to participating in the panel discussion with Chris Waller and others. When: Tuesday, Oct 15, 2:30-6:30pm ET Where: Google Cambridge (325 Main Street, Cambridge, MA, USA) RSVP: https://lnkd.in/efKFgp3r Hope to see you there. I do not get to Cambridge all that often these days (yes, Boston Seaport is close, but still). If you are in Cambridge and want to connect, let me know via DM.

High-resolution protein structures, particularly those with bound ligands, are fundamental to advancing structure-based drug design. DeepMind’s #AlphaFold has demonstrated the transformative potential of computational methods in structural biology. However, with the latest iterations of AlphaFold not being available for commercial purposes and the code remaining inaccessible to the public, the scientific community faces a significant gap. It is imperative that we explore and develop alternative computational tools to ensure continued progress in this critical field. Psivant Therapeutics is happy to support #OpenFold, a non-profit consortium with a mission of democratizing AI for structural biology and drug discovery. The tools being developed by OpenFold can be used by academics, #biotech, and #pharmaceutical companies to accelerate basic research and ultimately bringing new cures to patients. Why did we choose to support OpenFold? *Open Source* An open-source project that can be used and improved by academics and companies alike. *Permissive License* A permissively licensed model that allows commercial and non-commercial use. *Training Pipeline* Provides the tools used to train the model under the same license. *Optimized for Performance* Optimized performance for use on state-of-the-art and widely available #GPUs. *PyTorch-Based* A supercomputer scale, distributed training, #pytorch-based training framework. *Weights Available* Utilize the existing pre-trained weights to get quickly get started fine-tuning your model. *Community* Stellar researchers from academia and industry, working together to solve high-impact problems at the interface of computers and biology. Read the press release: https://lnkd.in/eNyM9NSJ Learn more on the OpenFold website: https://openfold.io/ If you want to participate in the OpenFold consortium, please contact Mallory Tollefson, Ph. D.

We are excited to announce the publication of our STORMM manuscript in a special edition of Journal of Chemical Physics on "Modular and Interoperable Software for Chemical Physics": https://lnkd.in/eutM__3v STORMM (Structure and TOpology Replica Molecular Mechanics)?is engineered from the ground up to harness the full power of modern #GPUs, handling many independent simulations simultaneously, ensuring efficient use of the ever-growing GPU compute bandwidth. We thank NVIDIA for feedback and running some benchmarking. We also thank the journal editors for inviting us to participate in this highly relevant special issue. A huge acknowledgement to David Cerutti for writing roughly a quarter million lines of general-purpose, performant library code from scratch (and more to come)! Please contact Dave or myself if you have questions or want to get involved in the STORMM project.?STORMM is an open-source project under the hashtag #MIT license. STORMM publication: https://lnkd.in/eutM__3v #GitHub Repo: https://lnkd.in/ecacHxSE

Great to be representing Psivant Therapeutics at the Longwood MIT Conference, co-hosted by Massachusetts Institute of Technology professors Phil Sharp and Bob Langer. Check out the amazing lineup of speakers: https://lnkd.in/e4rUGEH8 I will be in the panel on "Ecosystem of External Innovation". This great event is hosted by the Longwood Fund and the Longwood Healthcare Leaders.