Weekly Tech+Bio Highlights #26
Also: Market Myopia; Framework to Fight Antimicrobial Resistance; AI-driven Spatial Proteomics
Hi! This is
’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and the great companies driving the biopharma and medtech industries forward.If this newsletter is in your inbox, it’s because you subscribed or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:
Let’s get to this week’s topics!
Brief Insights
🔬 OpenAI develops GPT-4b micro, an AI model designed to enhance protein engineering for stem cell reprogramming, increasing the efficiency of Yamanaka factors by over 50%, marking its first foray into biological data and a step toward longevity science.
💰 Eli Lilly acquires Scorpion Therapeutics' PI3Kα inhibitor program, including lead candidate STX-478, for up to $2.5B to expand its oncology pipeline, while Scorpion spins out non-PI3Kα assets into a new precision oncology-focused company.
🔬 Baker Lab and collaborators develop AI-designed miniproteins that neutralize snake venom toxins, offering a cost-effective, stable alternative to traditional antivenoms with demonstrated 80-100% survival in preclinical studies.
🔬 Mayo Clinic partners with Microsoft and Cerebras Systems to develop foundation models for radiology and genomics, leveraging AI to enhance imaging workflows and analyze genomic data for personalized diagnostics, with initial applications in chest X-rays and rheumatoid arthritis.
🔬 Vevo Therapeutics plans to open source Tahoe-100M, the world's largest single-cell transcriptomic atlas with data on 100M cells, 1,200 drug treatments, and 50 tumor models, created in collaboration with NVIDIA and Parse Biosciences to accelerate AI-driven drug discovery. More about Tahoe dataset in our December 2024 issue.
💰 Bioptimus secures additional $41M in its latest round to develop a next-gen multi-scale, multi-modal AI foundation model for biology, building on the success of its H-Optimus-0 pathology model launched in 2024.
🔬 Synchron partners with NVIDIA to advance implantable brain-computer interface (BCI) technology, leveraging real-time edge AI and developing scalable brain-language foundation models to enhance neuroprosthetics and digital interaction for individuals with paralysis.
🔬 AbCellera expands its collaboration with AbbVie to develop novel T-cell engagers (TCEs) for oncology, leveraging its CD3-binding antibody platform, with AbCellera leading discovery and AbbVie handling development and commercialization.
🔬 Proscia's AI-powered Concentriq pathology is growing, as platform supported diagnoses for 2.4M patients in 2024, with adoption by major labs and pharma companies.
🔬 Insilico Medicine used its AI platform, PandaOmics, to identify two novel therapeutic targets, GBP2 and HCK, and repurpose the FDA-approved drug Lifitegrast as a potential treatment for endometriosis, with findings published in Advanced Science.
🔬 Rubedo Life Sciences announces Phase 1 trials for RLS-1496, a GPX4 modulator targeting aging cells and tissues, leveraging its AI-driven ALEMBIC platform to address age-related conditions, trials set to begin in the Netherlands in Spring 2025.
💰 Umoja Biopharma raises $100M in a Series C round to advance its in vivo CAR-T platforms, including UB-VV400 for oncology and autoimmune diseases.
🔬 Recursion partners with Faro Health to streamline clinical trial design using Faro’s AI-driven platform, reducing trial complexity and costs by optimizing study protocols and automating manual processes to accelerate drug development.
💰 GSK acquires IDRx for up to $1.15 billion, adding IDRX-42, a selective KIT tyrosine kinase inhibitor targeting all key mutations in gastrointestinal stromal tumors, to its oncology portfolio.
🔬 A28 Therapeutics and GATC Health completed AI-driven analysis of AT-101, a liver cancer peptide therapy, using GATC's MAT platform—to optimize clinical strategy by evaluating safety, efficacy, and off-target effects.
💰 London-based Lupa raises $4M in seed funding to develop its AI-powered Veterinary Operating System, featuring tools like an AI Co-Pilot and automated clinical documentation to streamline diagnostics and reduce administrative burdens for veterinarians.
🔬 Servier extends its partnership with Google Cloud for five years to integrate generative AI tools like Gemini into R&D, clinical trials, and production, aiming to accelerate drug development, improve efficiency, and focus on oncology and rare diseases.
💰 Be Biopharma raises $92M to advance its engineered B Cell Medicines (BCMs), supporting Phase 1/2 trials for BE-101 in hemophilia B and preparing BE-102 for hypophosphatasia, aiming for treatments with durable and redosable cell therapies.
💰 Normunity raises $75M to advance its lead T cell engager, targeting a novel tumor-specific mechanism in solid tumors, with Phase 1 trials planned for late 2025 and a broader anti-cancer pipeline in development.
💰 Tune Therapeutics raises over $175M to advance its epigenome editing platform TEMPO and lead clinical-stage program Tune-401 for chronic Hepatitis B, while expanding its pipeline of gene, cell, and regenerative therapies.
🔬 Researchers from EPFL and collaborators develop MaSIF-neosurf, an AI-driven tool for designing drug-controlled protein interactions by targeting "neosurfaces" formed by protein-ligand complexes, enabling applications in synthetic biology and innovative drug therapies, published in Nature.
🔬 Interius BioTherapeutics received European approval for the INVISE Phase 1 trial, marking the region's first in vivo CAR gene therapy study targeting B-cell malignancies with a single-dose therapy designed to create CAR-T and CAR-NK cells directly in patients.
🔬 Iktos and Cube Biotech partner to develop small molecule agonists for the Amylin Receptor, combining Iktos' AI-driven drug discovery platform with Cube's membrane protein technologies to address metabolic disorders like obesity and diabetes.
🔬 NVIDIA and IQVIA partner to develop custom foundation models and agentic AI workflows using NVIDIA AI Foundry, aiming to streamline clinical trials, accelerate drug development, and enhance healthcare decision-making with AI agents.
💰 Surbhi Sarna, former Y Combinator partner and founder of nVision Medical, launches Collate with $30M in seed funding at a valuation of over $100M. The startup aims to use AI and large language models to automate documentation processes and streamline workflows in clinical trials, product development, and manufacturing.
🔬 Lykos Therapeutics, after FDA rejection of its MDMA-assisted therapy for PTSD in 2024, plans an independent review of prior phase 3 data alongside a new phase 3 trial, following a "productive" meeting with the FDA, as part of efforts to resubmit for approval.
This newsletter reaches over 7.7K industry leaders from organizations like NVIDIA, Microsoft, Novo Nordisk, Pfizer, Novartis, FDA, the US Department of State, Illumina Ventures, and hundreds more.
Interested in sponsoring?
Contact us at info@biopharmatrend.com
Ryan Fukushima on Market Myopia
Ryan Fukushima, Chief Operating Officer at Tempus and CEO at Pathos, argues there’s a flaw in oncology drug development: decisions to advance or halt therapies are too often dictated by market forecasts rather than scientific merit. Fukushima posits that this reliance on projections—historically inaccurate by 40%—has led to the abandonment of treatments with great potential.
The development of Imatinib exemplifies the dangers of undervaluing scientific potential. Designed to target the BCR-ABL fusion protein driving chronic myeloid leukemia (CML), it initially faced skepticism. At the time, the pharmaceutical industry viewed kinase inhibitors as unfeasible due to concerns over specificity and side effects. Compounding the challenges, CML—a rare disease with an incidence rate of just 1–2 cases per 100,000 people annually in the U.S. (SEER)—was considered too small a market to justify significant investment.
Yet, Imatinib challenged these assumptions. Early trials revealed unprecedented results: in a Phase I study, all 31 patients achieved complete remission, with normal blood counts and cytogenetic remission in some cases. These outcomes, described by oncologist Brian Druker as “virtually unheard of,” highlighted the profound biological insight underlying the drug’s design.
By precisely targeting the BCR-ABL fusion protein—a tyrosine kinase driving unchecked cell proliferation—Imatinib blocked cancer at its molecular source. This success extended patient survival by more than a decade and achieved $4.7 billion in peak annual sales (Novartis Annual Report 2015; p. 189), overturning assumptions about the relationship between efficacy and market size.
''This really is the first clear example of a drug that targets an intracellular signaling molecule in cancer therapy and is almost through the approval process, and the results are just remarkable.'' — Dr. Ed Harlow, Harvard Medical School, NYT on May 8, 2001.
Fukushima connects this to contemporary drug development challenges, including Pathos’ work on P-300, a CBP/p300 inhibitor targeting cancer-driving proteins. Originally developed by Forma Therapeutics, P-300’s future became uncertain after Forma’s acquisition by Novo Nordisk in 2022. In 2023, Pathos secured worldwide rights to P-300, leveraging its AI-driven platform to optimize patient selection and clinical trial strategies. Fukushima’s approach seeks to expand the clinical relevance of P-300 beyond prostate cancer, emphasizing the importance of biological understanding in unlocking new applications.
For Fukushima, the Imatinib story underscores a broader critique of drug development: the industry’s high 90% attrition rate in clinical development is compounded by what he terms a "double penalty." The first penalty arises from the inherent challenges of drug development, with the majority of candidates failing during the lengthy and costly clinical trial process. The second penalty is imposed by flawed go/no-go decisions based on unreliable market models, which prematurely halt the development of promising therapies.
Together, these factors amplify inefficiencies to the point of limiting innovation and leaving potentially transformative treatments undiscovered. He argues that prioritizing biology-driven insights over theoretical market forecasts can uncover treatments capable of redefining patient outcomes and expanding markets—even for diseases initially perceived as too niche to warrant investment.
Using AI in Antimicrobial Resistance
A new report from the Fleming Initiative and Google DeepMind, Harnessing Artificial Intelligence in Antimicrobial Resistance, offers a detailed look at how AI could tackle one of the most pressing health challenges of our time. Antimicrobial resistance (AMR), a crisis projected to cause nearly 40 million deaths by 2050, is a slow-moving disaster already reshaping the way we think about global health. The report lays out a vision for how AI can help but underscores the stark reality—unlocking this potential will take more than just technology.
The report highlights several areas where AI could make a difference. AI could enable faster, more accurate diagnostics—think pinpointing resistant infections at the point of care, rather than waiting for lab cultures that take days. It could also predict how resistance spreads, helping health systems get ahead of outbreaks, and assist in designing new drugs by analyzing enormous datasets of microbial genomes and chemical compounds. But these opportunities come with challenges and require significant investments in infrastructure, such as high-performance computing, local data centers, and the training of an AI-literate workforce to ensure that the technology can be used effectively where it is needed most.
Some takeaways from the report:
AI’s potential: Rapid diagnostics, resistance prediction, and drug discovery are areas where AI could significantly improve AMR management, provided the right systems are in place.
Infrastructure gaps: Investment is needed in computing power, data-sharing frameworks, and local infrastructure to ensure AI’s equitable deployment globally.
Global collaboration is critical: Governments, industries, and healthcare systems must work together to develop shared data frameworks and ensure equitable access to AI tools.
Interdisciplinary expertise: Training programs and career pathways that bridge the gap between AI and AMR science will be key to progress.
Focused action on equity: AI systems must be designed and implemented in ways that reduce disparities, particularly in low- and middle-income countries.
The report draws from the United Nations High-Level Meeting on AMR. To operationalize these recommendations, the Fleming Initiative and Google DeepMind have established an Academic Fellowship to develop interdisciplinary expertise necessary for translating AI-driven solutions into practical outcomes.
Related: In October, we covered an effort by researchers in Hungary to map Acinetobacter baumannii resistance globally, making a map of antibiotic resistance. This genomic surveillance framework aims to support precision phage therapy, enabling hospitals to pre-emptively tackle antibiotic-resistant infections with tailored bacteriophage treatments.
Virtues of Spatial Proteomics
The idea of a virtual cell is gaining momentum across the biotechnology and AI communities. Initiatives like Noetik’s OCTO-VirtualCell and Recursion’s world model platform are exploring new ways to model biological systems, while the Chan Zuckerberg Initiative integrates datasets like the Human Cell Atlas to advance cellular modeling. Google DeepMind is also investigating virtual cells to simulate molecular dynamics and aid in drug discovery.
This time, we take a look at VirTues, a foundation model framework developed by researchers at EPFL and ETH Zurich, which aims to bring the concept of virtual tissues to life using spatial proteomics.
VirTues builds on the ideas outlined in the paper “How to Build the Virtual Cell with Artificial Intelligence” (Cell, 2024), which proposed AI models that integrate data from cells, tissues, and molecular systems to create detailed representations of biological systems. VirTues addresses multiplex tissue imaging by using a vision transformer to integrate spatial details with molecular markers with a protein language model. Importantly, the model can generalize to new datasets—like pancreatic tissue from type 1 diabetes—without retraining.
One of VirTues’ notable features is its explainability, highlighting molecular and spatial factors behind outputs like cancer grade, relapse risk, and cell type distributions. This clarity helps researchers and clinicians better understand tissue samples. The model performed consistently well across scales, excelling in tasks such as identifying underrepresented cell types and niche-level structures.
VirTues supports clinical decision-making through its Virtual Tissues Database, which compares tissue samples based on molecular and architectural profiles, helping researchers and clinicians find similar cases from a database of previous patients. By linking tissue phenotypes to clinical outcomes, the database could support more informed treatment decisions or research into new therapeutic strategies.
Another feature of the model is its ability to handle new markers without retraining, making it easier to integrate emerging biomarkers. While its performance may dip with markers that are very different from its training data, the framework still handles a wide range of data types to support cross-study comparisons and large-scale analyses.