Weekly Tech+Bio Highlights #42: Proteins, the Universe, and Everything
Neuralink's $650M Raise, Interactive Map of Digital-Pathology Deals, ASCO AI Highlights, and a 10-Year Quest to Map Cell Behavior
Over the last week, a protein-design model broke the 20-amino-acid barrier by designing proteins with 29 noncanonical residues, MIT’s foundation model posted benchmark-leading accuracy in small-molecule property prediction, and a reinforcement-learning framework generated a novel antibiotic active against MRSA in mice.
At ASCO 2025, Tempus, Artera, Massive Bio, BostonGene and Median Technologies previewed multimodal AI tools for trial matching, prognosis and imaging; prior to that, Intelligencia AI used its predictive platform to highlight five emerging oncology programs from non–big pharma developers.
Capital—Neuralink raised $650 million to drive brain-computer interface trials; ProteinQure secured $11 million for an AI-designed peptide therapeutic; and bidding reopened for 23andMe, with a mystery backer reportedly holding $17 billion in cash. Roche’s investor day featured news on genome run times, flow cell reuse, and assay development across diagnostics and pharma divisions.
Beyond AI, Harvard chemists built self-replicating synthetic cells from polymers; and the Allen Institute launched a decade-long effort to model multicellular behaviour (Allen Institute’s imagery featured on the cover). Meanwhile, emerging startup activity spanned brain diagnostics, iPSC therapies, AI-powered tissue platforms, robotic drug production, and clinical automation—driven by a mix of regional rebounds and targeted raises across Europe and North America.
The highlights above are a quick summary—hyperlinked stories with more detail are organized in the sections below.
Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.
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:
🤖 AI x Bio
(AI applications in drug discovery, biotech, and healthcare)
🔹 MIT researchers Jackson Burns and Akshat Zalte release CheMeleon, a descriptor-based foundation model for predicting molecular properties, integrated into ChemProp v2.2, achieving top performance in 89% of 57 benchmarks with strong results even on datasets as small as 100 entries.
🔹 Structure prediction breaks the 20-amino-acid barrier—RareFold becomes the first deep learning model to accurately predict and design protein structures with 29 noncanonical amino acids, enabling scalable therapeutic peptide engineering with side-chain precision rivaling AlphaFold3.
🔹 Researchers developed SyntheMol-RL, a reinforcement learning framework that generated a novel antibiotic, synthecin, effective against MRSA in mice.
🔹 Foundation models stir debate in AI pathology—a Nature feature highlights both promise and skepticism around large AI models like UNI2, CONCH, and PathChat, which aim to automate and enhance cancer diagnosis from pathology slides. While developers cite breakthroughs in classification and workflow automation, critics warn of overhype, poor generalizability, and a lack of clinical validation.
🔹 Todd J. Herron and Deok-Ho Kim (Curi Bio) report an AI-guided laser purification platform for hiPSC-derived cardiomyocytes, enabling faster, label-free cell selection for cardiac therapy applications (Nature Communications Biology).
🔹 Medical imaging AI gains spatial intelligence: PRS-Med, a new vision-language model, integrates tumor segmentation with positional reasoning from free-text prompts—outperforming prior models like LLaVA-Med and BiomedParse across six modalities and introducing a benchmark dataset (MMRS) for spatially grounded QA in medicine.
🔹 Lantern Pharma develops AI-driven method to decode hidden patterns in ctDNA, enabling 86% predictive accuracy in treatment response and identifying novel genomic biomarkers, now being tested in three active clinical trials. The company also reported preclinical results for its experimental brain cancer therapy in rare pediatric tumors.
🔹 Designing new cyclic peptides with AI—Gaurav Bhardwaj shares peer-reviewed work (Nature Communications) on a tool for structure prediction, sequence redesign, and de novo generation of cyclic peptides, expanding applications in binder and therapeutic design.
🔹 Harvard and MIT launch MedBrowseComp, a benchmark for testing AI agents on real-world medical web browsing, revealing that top models struggle with multi-hop reasoning and GUI tasks.
🔹 Cracking the dark proteome with AI—a new perspective in ACS Publications by Percudani & De Rito outlines how deep learning, structure prediction, and language models are converging to decode unknown protein functions, with 214 million structures now mapped but most still lacking known roles.
🔹 Tempus unveiled a new program leveraging its massive multimodal dataset to build a foundational AI model aimed at improving diagnosis, prognosis, and personalized treatment selection in precision medicine. The company also expanded its AI clinical tool with EHR integration and ASCO guidelines.
🔹 SoundHound AI and Allina Health launched a voice AI agent that now handles most patient calls to reduce wait times.
🔹 Transparent AI for diagnostics—Diagnostics.ai released the first CE-IVDR certified AI platform for molecular diagnostics, offering per-test auditability and algorithm transparency for PCR testing as new EU regulations take effect.
🔹 AI-discovered anti-inflammatory clears early safety trial—Enveda’s novel oral anti-inflammatory passed Phase 1 safety testing and is advancing to trials in atopic dermatitis.
🔹 Yonsei University launches AI-powered cancer data library across 171,000 cases—The YCDL integrates 800+ real-time data points per patient across 11 cancer types, delivering high-accuracy NLP, 3D tumor maps, and interactive decision support.
🔹 MIT Jameel Clinic unveils a protein localization model for single cells—Combining ESM-2 with stained cell images, PUPS predicts pixel-level protein placement from sequence data alone—even for unseen proteins.
🔹 Precise design of cell-type-specific regulatory DNA—Bo Wang, CIFAR AI Chair at Vector Institute, introduces new framework that fine-tunes DNA language models using Lagrangian-guided reinforcement learning to generate cis-regulatory sequences with high on-target expression and minimized off-target effects,
🔹 Insilico Medicine’s AI-designed oral ENPP1 inhibitor reactivates the STING pathway to overcome resistance to chemo and immunotherapy.
🔹 AI diabetes care—IBM and Roche co-developed an app that uses real-time glucose data and AI to predict hypoglycemia.
This newsletter reaches over 8.7K industry professionals from leading organizations across the globe. Interested in sponsoring?
Contact us at info@biopharmatrend.com
🚜 Market Movers
(News from established pharma and tech giants)
🔹 Alex Dickinson shares updates from Roche investor day—noting chip reuse for lower costs, 4.5h genome runs, strong FFPE/MRD results, early long-read capability (>1,000bp), and broad assay development across Roche’s diagnostics and pharma units.
🔹 BioAge preps its oral, brain-penetrant NLRP3 inhibitor for IND filing in obesity; in mice, it matched semaglutide in weight loss and showed additive effects when combined.
💰 Money Flows
(Funding rounds, IPOs, and M&A for startups and smaller companies)
🔹 Neuralink raises $650M in Series E at $9B valuation amid brain-computer interface trials—securing fresh funding as it continues testing its brain implant, which enabled a quadriplegic patient to browse the internet using only neural signals; the company also recently received FDA breakthrough designation.
🔹 AI-designed peptide drug heads to clinic—ProteinQure raised $11M to launch trials for what it calls the first AI-designed peptide therapeutic, targeting treatment-resistant breast cancer, and expanding its pipeline in neurology and kidney disease.
🔹 Speculation swirls over 23andMe’s mystery backer—As 23andMe’s founder reenters the bidding for her company, observers including Yaniv Erlich and Alex Dickinson are parsing court clues—like $400B+ market cap and $17B in cash—to speculate which major company is financing her counteroffer.
🔹 Central and Eastern European (CEE) healthcare startups rebound in 2025, raising €75M YTD across 16 deals—already surpassing 2023 totals—driven by major rounds from Atrandi Biosciences, Ataraxis AI, Jutro Medical, and Ingenix.ai, according to Kris Przybylak from Inovo.vc.
🔹 Danish startup Brainreader raises $7.3M to expand global adoption of its FDA-cleared AI brain diagnostics tool, focusing on U.S. market growth and enhanced clinical capabilities for conditions like Alzheimer’s and brain injuries.
🔹 France-based TreeFrog Therapeutics secures €30M from the European Investment Bank (EIB) to advance its Parkinson’s disease cell therapy toward clinical trials and expand its iPSC-based pipeline, with backing from the EU’s InvestEU program.
🔹 $40M to replace animal testing with human tissue AI—Vivodyne raised $40M (led by Khosla Ventures) to scale its robotic platform that grows and tests human tissues, aiming to improve drug success rates and reduce reliance on animal models in preclinical research.
🔹 AssistIQ raised $11.5M (Battery Ventures) to scale its computer vision platform that automates surgical supply tracking.
🔹 French Startup Diamidex raises €5M to expand AI-powered microbial detection—it’s MICA platform accelerates detection for key pathogens and is scaling globally with new funding from PureTerra Ventures and Invesdor.
🔹 Finnish CurifyLabs raises €6.7M to automate drug production— it replaces manual compounding with a robotic platform, aiming to scale in the U.S. and lead a €5.6M EU project for decentralized pharma manufacturing.
⚙️ Other Tech
(Innovations across quantum computing, BCIs, gene editing, and more)
🔹 “Bridge recombinases let us rewrite the genome, not just edit it”, explains Niko McCarty (Asimov Press). Unlike CRISPR, these RNA-guided enzymes can cut, invert, or delete massive genomic regions—up to 930,000 bases—without relying on cellular repair, enabling predictable, large-scale genome engineering in human cells.
🔹 Harvard researchers create self-replicating synthetic cells made from polymers—demonstrating lifelike behavior without biochemistry, suggesting how non-carbon-based life might arise and inspiring new nanotech and astrobiology approaches.
🔹 Neuralink rival enters human testing—Paradromics successfully implanted its brain-computer interface in a human for the first time, confirming neural signal capture in a brief trial—joining Neuralink in advancing high-resolution BCIs for speech restoration.
🔹 Allen Institute expands its enhancer-AAV toolkit to target nearly any brain cell type in rodents and primates, enabling precise gene delivery without transgenics and unlocking cross-species studies in neurology and gene therapy.
🔹 Researchers led by Bernhard Kuster (TUM) present a high-resolution map of human proteins with direct DNA contact, using a novel zero-distance photo-crosslinking method to identify over 1,000 DNA-bound proteins and 688 crosslinked peptides at single-amino-acid resolution, offering insights into genomic regulation and protein-DNA interactions.
🔹 Benjamin McLeod shares insights from an AAV review, highlighting breakthroughs in capsid engineering, post-translational modifications, epigenetic effects, and emerging synthetic production platforms like E. coli-based VLPs.
🏛️ Bioeconomy & Society
(News on centers, regulatory updates, and broader biotech ecosystem developments)
🔹 Allen Institute launches CellScapes, a 10-year initiative to map how human cells cooperate to form tissues, using dynamic imaging and mathematical models—aiming to predict, and eventually design, cell behavior for applications in cancer, regenerative medicine, and synthetic biology.
🔹 Animal models remain essential to biomedical research, argues Carole LaBonne in STAT, highlighting their irreplaceable role in understanding complex diseases like cancer and diabetes—despite advances in AI and in vitro tools.
🔹 David Furman and co-authors argue immune aging drives chronic inflammation and multimorbidity, proposing biomarkers like iAge and immune clocks to quantify and potentially reverse immunosenescence—positioning immune health as a core metric of biological aging.
🔹 Montana opens U.S. gateway for experimental anti-aging therapies—A new state law permits clinics to offer unapproved treatments after Phase 1 safety trials, creating a legal path for in-state use and positioning Montana as a potential hub for medical tourism and longevity biotech.
🔹 DeciBio’s Katie Maloney shares a new interactive network chart highlighting over 40 digital pathology partnerships in 2025—led by PathAI and AstraZeneca—spotlighting a growing focus on integration, the rise of digital cytology, and first-time players like Pictor Labs and DigitCells entering the scene.
🚀 A New Kid on the Block
(Emerging startups with a focus on technology)
🚀 Biotech shows early-stage spark in April/May—Max Robinson lists 20 new startups that emerged in Q2 2025 with strong plays in gene editing, ADCs/PDCs, and platform tech; oncology and autoimmune lead, with launches spanning the US, Europe, and Israel.
🏥 ASCO 2025
(Some highlights from the 2025 ASCO Annual Meeting)
🔹 AI flags rising oncology innovators—Intelligencia AI identified five standout non–big pharma cancer programs ahead of ASCO 2025, featuring novel cell therapies, radiotherapies, and ADCs with strong early results and regulatory traction.
🔹 Tempus presented ten studies at ASCO 2025, highlighting AI-enabled advances in trial matching, ctDNA monitoring, genomic profiling, and EHR data abstraction to support more personalized oncology treatment.
🔹 AI predicts who benefits from prostate cancer drugs—Artera presented validation data at ASCO 2025 showing its multimodal AI tool can identify which high-risk prostate cancer patients benefit from added therapy.
🔹 Median Technologies showcased its AI-powered imaging services for oncology clinical trials at ASCO 2025, including new radiopharmaceutical image processing tools and abstracts on tumor growth modeling and muscle index analysis.
🔹 BostonGene presented six studies at ASCO 2025 showing how its multimodal AI platform enhances biomarker discovery, patient stratification, and therapeutic response prediction to accelerate oncology drug development.
🔹 Massive Bio showcased its AI-powered platform for clinical trial matching and oncology data at ASCO 2025, highlighting tools that use real-world evidence and digital workflows to improve research access and decision-making.
Cover: From Allen Institute’s CellScapes project page imagery
Read also:
13 Foundation Models: Startups, Industry Updates and the Nobel Prize