Recursion

Meet the newest cohort of healthcare startups at Altitude Lab! Following a record number of applications, our Utah-based incubator has announced 7 new startups addressing critical needs, from improving patient outcomes and developing new therapeutic modalities to making diagnostics more accessible. They include:  ◾ Freyya - developing wearable technology for pelvic floor rehabilitation; ◾ Conduit Bio - using multimodal data and AI to increase the success rate of experiments; ◾ Sethera - advancing therapeutic discovery with a peptide research and development platform; Light Hope Diagnostics - using liquid biopsy to detect cancer-specific RNAs;  ◾ Anziva - redesigning existing medications to enhance drug performance, reduce side effects, and improve stability at lower costs;  ◾ Intactus Bio - revolutionizing biocomputation and regenerative medicine through novel tissue engineering; and  ◾ Vira Regen - developing therapies and devices for bone fracture healing, musculoskeletal injuries, osteoarthritis/osteoporosis, and cancer. Learn more: https://lnkd.in/eSn-gMVx #ai #healthcare #startups #utah #incubator #founders Chandana Haque

A career at the forefront of AI and biology. A new video from AI Utah highlights Imran Haque, VP of AI and Data Science, who leads a team of data scientists, machine learning experts, computational biologists and chemists to build AI systems that map biology and identify novel ways to treat disease. #career #techbio #biotech #ai #ml #biology

The Recursion mindset is about pushing boundaries: “An unwavering drive to challenge and disrupt the status quo.” ◾ Every year, we honor that mindset with the Tech Symposium – encouraging all of our creative data scientists, software engineers, chemists, biostatisticians, machine learning scientists, and product managers to share the big ideas they’ve been working on.  ◾ Over the past 2 days, Recursionauts gave more than 80 presentations for hundreds attending in person at our Salt Lake City headquarters and online under the theme “Connecting People to Technology.” ◾ They shared how they are creating tech-enabled “superteams” and industrialized workflows; how they’re building multimodal models and driving advances in mass spectrometry and differential gene expression. ◾ They gave talks on agile clinical trials and scaling quantum chemistry. They provided a how-to on using public data, a debate on data science notebooks Hex vs. Looker, and a talk on transforming compound management with automation. ◾ Interns shared their work in a poster session – and there were also lab tours, happy hours, craft sessions, tailgates and digital scavenger hunts. Keynotes included company tech leaders like Laura Schaevitz, Berton Earnshaw, Marissa Saunders and Najat Khan, PhD. With the fast pace of discovery at Recursion, the Tech Symposium offers everyone a chance to learn and ask questions about what others are working on – and it provides a prime opportunity for collaboration, and the next creative tech breakthroughs. #ai #ml #tech #techbio #techsymposium #data #datascience #creative #collaboration #engineer

Data is the differentiator. ◾ Today, we’re kicking off a series called Data Deep Dive – sharing how our bilingual team of over 500 Recursionauts – equal parts life scientists and computational experts – collect, model and analyze data in order to solve challenging biological problems and transform drug discovery. 🗺 We approach these problems through mapping – not unlike how Amazon predicts your shopping preferences or Google Maps calculates when you will arrive at your destination: (1) observe and collect massive datasets on real-world interactions;  (2) develop models to make predictions based on that dataset; (3) combine these two steps into a virtuous cycle where the model is continuously improving. ◾ But because biology is so unpredictable and multi-layered, our data needs to be extremely custom-built. High-quality, high scale biological data simply doesn’t exist publicly. ◾ First, we built our automated wet lab which now performs over 2 million experiments per week, generating images of human cells and extracting signatures of what is happening in those cells using computer vision and machine learning. Using this lab, we’ve amassed one of the largest phenomics (human cellular image based) datasets in the world. ◾ Over the last several years, we’ve been adding data layers – using the same philosophy that we brought to phenomics. We now generate data from 1,000 digitized animal cages and integrate multimodal precision oncology patient data. We’ve sequenced over 1 million transcriptomes and run tens of thousands of ADME experiments using our automated DMPK module. ◾ All together, we now have access to over 50 petabytes of fit-for-purpose proprietary data (and growing). ◾ Using this data, we are building foundation models that predict and navigate over 6 trillion biological and chemical relationships. These models stitch together multiple point solutions into an industrialized workflow. ◾ Leveraging our supercomputer, BioHive-2, our team of Recursionauts is building the purpose-built data and designing the models to create biological maps that can drive new insights. To turn drug discovery into a search problem and massively accelerate new treatments. #data #datadeepdive #ai #ml #drugdiscovery #supercomputer #biology #chemistry #tech Chris Gibson Ben Mabey Imran Haque Najat Khan, PhD Laura Schaevitz Lina Nilsson

In biology, unpaired data is the norm as experiments are destructive. This lack of paired data limits our ability to build powerful multimodal models for drug discovery. To fuel a multimodal future in TechBio, we propose a simple algorithm for matching. In this work from Jason Hartford's team, we fit a classifier that outputs the conditional probabilities of specific perturbations across single-cell assays like phenomics and transcriptomics. This gives us a common space where matching algorithms like OT-based matching can be used to align modalities. These aligned samples can then be used as proxies for paired samples where existing multimodal representation learning techniques can be applied. See the experiments: https://lnkd.in/gmkTwHju Read the paper: https://lnkd.in/gv-fErT5

Inside the rise of Recursion. In a new feature in STAT, Allison DeAngelis dives deep into the Recursion story – and how co-founder and CEO Chris Gibson followed his instinct about the possibility of using image analysis technology in biology and went on to build a massive proprietary dataset to “become a frontrunner in the buzzy field of AI drug development, leading the pack of first-generation companies behind a whole new kind of research.” Drawing on interviews with more than 30 executives, investors, former employees, and industry experts, the story follows the company from its early beginnings – from building out screening tools and software developed by Anne Carpenter at Cambridge’s Broad Institute of MIT and Harvard, to creating an industry-leading database, to advancing a pipeline of AI-enabled drugs, and forming partnerships with leading pharma companies including Bayer and Roche Genentech. She writes: “The real value, investors say, is the database that Recursion built from those imaging tools, plus its staff’s know-how and trade secrets… Today, the company has used robots and automated systems to collect over 50 petabytes of data, close to three-quarters of which is proprietary to Recursion.” The story includes insights into our pipeline, including the coming Phase II clinical trial readout for Cerebral Cavernous Malformation in September. It also speaks to our new chemical entities, created from scratch, for the bacterial infection C. difficile as well as a novel cancer drug targeting RBM39. As Chief R&D and Chief Commercial Officer Najat Khan, PhD, notes, the company’s focus on improving existing medicines and developing new treatments for conditions that have none is strategic. “Patients need both,” she says, adding that she “wants to continue ‘upping the game’ on Recursion’s chemistry work and adding new drug candidates to the pipeline.” The story also includes Recursion’s close partnership with NVIDIA, who invested $50 million – and provides an inside look at our supercomputer that Nvidia helped build.    And it touches on the exciting developments in our collaboration with Genentech and Roche – building a neuroscience phenotype or “Neuromap” that required us to grow our own stem cell-derived neurons, making us one of the largest producers in the field.. “I wasn’t sure that the scales that we were talking about would be feasible,” said Andrew Chan, senior vice president of research biology at Genentech. “Because you’re growing cells at an astronomical volume. Not everything’s going to work, but we have been pleasantly surprised.” Read more: https://lnkd.in/eJ9vGtgk  #ai #tech #techbio #investment #drugdiscovery #pipeline #pharma #biology #chemistry

Want to understand how we make our digital maps of biology? Our updated preprint explains. “Building, Benchmarking, and Exploring Perturbative Maps of Transcriptional and Morphological Data” reveals how scientists at Recursion and Genentech are scaling genetic perturbation techniques with high-dimensional assays such as cellular microscopy and RNA-sequencing in order to create digital “maps of biology” that both capture known biological relationships and uncover new associations that can lead to new therapeutic discoveries. The paper shows how datasets from various experiments (including CRISPR-Cas9 knockout, CRISPRi knockdown, and compound treatment) are combined to create these maps – unified, relatable embedding spaces allowing for the generation of genome-scale sets of pairwise comparisons. In the paper, the researchers:  (1) establish a standardized terminology for the steps involved in perturbative map building (2) introduce key classes of benchmarks to assess the quality of such maps (3) construct maps from four genome-scale datasets employing different cell types, perturbation technologies, and data readout modalities (4) generate benchmark metrics for the constructed maps and investigate the reasons for performance variations, and  (5) demonstrate utility of these maps to discover new biology by suggesting roles for two largely uncharacterized genes. Learn more: https://lnkd.in/e9UmvZQ8 #ai #ml #maps #tech #drugdiscovery #techbio #digitalmaps #biology Safiye C. Jan-Christian Hütter, Nathan Lazar, PhD, Rahul Mohan, Conor Tillinghast, Tommaso Biancalani, Marta Fay, Berton Earnshaw, Imran Haque

Staff Machine Learning Scientist Ali Madani presents on "AI and Data Science" at the 4th Transatlantic ECI GPCR symposium on Sept. 5 - a 2-day symposium for early career investigators (ECIs) designed to promote collaboration among early-career scientists in the GPCR field and provide guidance from more established researchers. 👇

The synergy between technological innovation and regulatory advancement. At the 15th Annual Drug Discovery Innovation Forum in Berlin Sept. 4-5, Karen Sayal - MD PhD consultant clinical oncologist and industry-first scientist in AI-driven Clinical Translation and Clinical Development at Recursion - will give a talk on the synergistic advancement of AI-enabled drug discovery and regulation. She'll discuss: ◾ How industrialising drug development is unveiling hidden biological complexities to propel a paradigm shift in treating human disease. ◾ How technology-driven discovery is synergistic to the needs of our evolving next-generation regulatory ecosystem. ◾ And how we can achieve seamless interoperability between the needs of developing and deploying AI technologies and the needs of regulation through pragmatic bipartisan dialogue and coordinated evolution. Learn more & register: https://lnkd.in/d8EFpTuh #ai #ddif #ddif2024 #drugdiscovery #innovation #tech #techbio #regulation

Building the datasets to drive better treatments.  In a new episode of NVIDIA's AI Podcast, Chris Gibson talks to host Noah Kravitz about Recursion’s journey from an idea he began exploring while working on his MD/PhD at the University of Utah, to a leading TechBio company with one of the largest proprietary datasets in the industry and pharma’s fastest supercomputer. Some highlights: 1. The problem: Patients need better treatments. “Everybody knows somebody who has a disease where there’s not a good treatment,” Chris says. “You’ve got a relative who’s died of cancer or someone suffering from Alzheimer’s or Parkinson’s, and despite all of the incredible technology in the world, 90% of drugs that go into clinical trials fail before.” 2. How existing solutions fall short. “Traditionally in our industry, people have gone after one disease at a time. If we wanted to go after a disease together, we’d read the literature, we’d build a team. And then we would build specific experiments for that one disease. And we would generate data. And 5 or maybe 10 years later we’d have a drug going into the clinic – and there’s a 90% failure rate.” 3. How Recursion is betting on data. “We’re taking this approach to invest more initially to build really large complex datasets…over time you start to build these network effects where today if we get some results in our automated laboratory we can compare it to over 250 million experiments that we’ve run over the past few years and all of that data’s relatable. So instead of one disease at a time, instead of slices of biology, we’re building a volume and we’re sampling this volume and using AI to fill in what we can predict about the rest of it.” 4. The uniqueness of Recursion's approach. “I’m unaware of any company that is building a dataset that is this large. We’ve knocked out with CRISPR-Cas9 every gene in the human genome in multiple different cell types. We’ve profiled millions of molecules, and all of these data exist in multiple layers. We call it “omics.” People may have heard of genomics, well we’re building phenomics and transcriptomics and invivomics and proteomics – all of these omics data layers.” 5. How it compares to the evolution of Google maps. “Go back to the Google map days. We started with maps from airplanes and knew where all the streets were. Then street view, and cars were driving around. We’re building all these same layers [at Recursion], but instead of doing it in the physical world, we’re using a massive automated laboratory full of robots to do millions of experiments to figure out what the roads and intersections are of biology and chemistry.” Learn more and listen to the full episode here: https://lnkd.in/eFApAKiP #ai #podcast #tech #techbio #drugdiscovery #pharma #maps #biology #chemistry