Bits In Bio Weekly - June 17th 2025
Updates From and Around the Bits in Bio Community
📢 Highlights
Xaira drops a X-Atlas/Orion - an 8-million-cell perturb-seq atlas to turbocharge AI-driven biology
23andMe's cofounder and ex-CEO outbids Regeneron in $305M deal to buy back her company
AstraZeneca provide $5.3B for Chinese AI engine co-development with CSPC to accelerate small molecule discovery
Not yet a member of our super awesome slack community of >9500? Join HERE 🤗
👀 In Case You Missed it …
Xaira releases an 8-million-cell Perturb-seq dataset (X-Atlas/Orion) to power next generation of AI-driven "virtual cell" models in biology.
Bay Area AI-enabled biotech co-founded by Nobel laureate David Baker, Xaira, has unveiled "X-Atlas/Orion," the largest publicly available genome-wide Perturb-seq dataset to date. Comprising > 8 million single cells with CRISPR perturbations targeting all human protein-coding genes, the atlas provides an unprecedented resource for training AI foundation models of cell biology. The dataset was generated via Xaira's new FiCS Perturb-seq platform, which fixes and cryopreserves cells for high-throughput single-cell RNA sequencing – enabling both massive scale and high data quality. Notably, Xaira's team also developed a method to measure dose-dependent gene effects by quantifying guide RNA per cell, offering a finer-grained view of gene knockdown impacts. This release follows the broader trend of "data philanthropy" in biotech, joining efforts like the Chan Zuckerberg Initiative's CellxGene corpus and Meta's ESM protein database. The timing aligns with growing interest in virtual cell models – including efforts by companies like Genentech (GAVEL) and academics pushing for a "Human Cell Atlas 2.0" – all racing to build AI that can predict cellular behavior. By sharing this trove openly, Xaira hopes to accelerate "virtual cell" simulations that predict how complex genetic interactions drive disease, ultimately speeding the discovery of new drug targets and therapies. The move also positions Xaira strategically in the foundation model race, as companies with the best data almost always build the best models.
Behind the Deal: AstraZeneca Taps China’s CSPC in $5.3B AI Drug Discovery Pact
AstraZeneca has signed a deal worth up to $5.3B with CSPC Pharmaceutical to co-develop AI-designed oral therapies for chronic diseases. The agreement—$110M upfront, $1.6B in R&D milestones, and $3.6B tied to sales—gives AZ global rights to drugs discovered using CSPC’s AI “dual-engine” platform. It’s AstraZeneca’s biggest AI partnership yet and follows a string of tech-enabled tie-ups: Verge Genomics for neuro drugs, Tempus for trial design, and Absci for generative antibody discovery. Unlike those, CSPC brings a full-stack small molecule engine, aimed at scaling hit-to-lead timelines for inflammation and immune diseases.
The move mirrors a broader trend: Sanofi dropped $1.9B on AI-engineered immunology via Earendil; Incyte bet $885M on Genesis AI; and Roche continues to bulk up with its own AI-native investments. AZ is clearly signaling: the next blockbuster may come from code, not just the clinic. For CSPC, the deal is a breakout moment—offering validation on the global stage and a potential royalty windfall. First candidates could hit the pipeline within a year. The clock’s ticking to show this engine runs as well in practice as it does in pitch decks.
Anne Wojcicki to reclaim 23andMe, keeping genetic data under founder control despite public and regulator concerns
In what continues to be the most dramatic destruction of shareholder value in the history of consumer diagnostics formerly high-profile genomics pioneer, Anne Wojcicki is poised to reacquire 23andMe through her new purpose-made nonprofit entity, offering $305 million for essentially all the company's assets through the backing of an undisclosed set of financiers. This comes after 23andMe entered bankruptcy proceedings earlier this year following revenue struggles and a major 2023 data breach. Wojcicki's bid – which notably outbid Regeneron's $256M offer for the company, coming only after the auction was complete – appears driven by a desire to keep 23andMe's gigantic trove of consumer genetic data out of a pharmaceutical giant's hands and under her exclusive control. The drama echoes other consumer genomics pivots: Ancestry partnered with Blackstone for drug development, Nebula Genomics was acquired by ProPhase Labs, and Helix shifted to population health – all trying to monetize genetic databases while navigating privacy concerns. Regulators and state officials had raised privacy concerns about the sale of DNA data, prompting Wojcicki to testify in Congress that customers' genomic info wouldn't be sold without consent, however not mentioning what would happen to the samples themselves (from which more than just DNA could be obtained. The bankruptcy auction attracted interest from multiple parties including GSK (23andMe's former drug discovery partner) and private equity firms, highlighting the value of genetic databases in the AI era. By taking 23andMe private (the company had gone public via SPAC at a $3.5B+ valuation in 2021), Wojcicki shows her tenacity. The nonprofit structure may allow more focus on research over short-term profit, especially if equipped with a strong board. If approved by the bankruptcy court, the deal will represent a full-circle moment – the company's founder retaking the helm to "reset" a personal genetics revolution that she helped start with co-founders Paul Cusenza and the progenitor of the entire concept, Linda Avey.
Eli Lilly to acquire CRISPR base-editing biotech Verve Therapeutics for up to $1.3 billion
Pharma giant Eli Lilly has struck a deal to buy Verve Therapeutics – a genome-editing startup – for as much as $1.3 billion, providing a ˜5x return on capital invested. Verve specializes in base editing therapies for cardiovascular disease, including a one-time gene treatment to permanently lower "bad" LDL cholesterol by editing the PCSK9 gene. The acquisition, reportedly valued at $13.50/share if milestones are met, represents roughly double Verve's recent average stock price, signaling Lilly's strong conviction in the technology. It also outbids an earlier offer by Regeneron (Verve's original partner on PCSK9 programs), underscoring Lilly's urgency to secure Verve's pipeline and proprietary data. This marks Lilly's second major gene therapy acquisition after buying Prevail Therapeutics for $1B+ in 2020, and comes as the pharma giant races Novo Nordisk in the cardiometabolic space – where both companies are printing money from GLP-1 drugs. The deal validates the clinical promise of base editing following Verve's positive Phase 1b data showing 55% LDL reduction lasting 6+ months. The move highlights big pharma's appetite for genomics: Lilly is effectively betting that Verve's CRISPR-based therapies could become blockbuster treatments for heart disease, complementing Lilly's existing cardiovascular portfolio. If successful, the deal could validate base-editing in the clinic – and demonstrates Lilly's strategic shift toward in-house genomic medicine expertise, joining peers like Vertex (who partnered with CRISPR Therapeutics) and Intellia's go-it-alone approach.
Basecamp's Global expedition expands life's catalog 10×, yielding 9.8 billion new proteins for AI
London-based Basecamp Research announced a landmark preprint reporting the discovery of over 1.2 million previously unknown microbial species found in some of Earth's most remote, extreme environments. Over years of global 'bioprospecting’ expeditions through partnership with 125+ local teams, they collected and sequenced environmental samples, compiling BaseData™, now the world's largest protein sequence database at 9.8 billion sequences. This clean, non-redundant dataset is 10× bigger than all public protein databases combined, effectively shattering the "data wall" that has limited AI models trained on narrow genetic diversity. The achievement puts Basecamp in elite company with ESMFold (which used 250M sequences) and AlphaFold (trained on ~200k structures), but with orders of magnitude more raw material. Basecamp's trove captures a vast range of novel enzymes and biochemical pathways – from a metal-scavenging bacteria on a sunken WWII ship to an Antarctic microbe that survives by extracting water from air. This discovery comes as companies like Arzeda, Generate Biomedicines, and EvolutionaryScale are hungry for novel protein sequences to train their generative models. The timing also coincides with growing interest in bioprospecting for climate solutions – think carbon-fixing enzymes or plastic-degrading catalysts hiding in nature's untapped corners. By openly sharing BaseData™, the team aims to fuel next-gen foundation models and "virtual chemists" in biotech. The real-world payoff could be significant: with AI now able to learn from nature's hidden 90%, researchers envision new antibiotics, sustainable bioplastics, carbon-fixing enzymes and more – all inspired by the planet's newly revealed genomic dark matter.
Neuropsychiatry startup Draig debuts with $140M Series A to advance novel therapeutics
Draig Therapeutics, a new biotech out of Wales, just hatched with one of the largest Series A rounds in recent memory at $140 million. The company – whose name means dragon in Welsh – is tackling stubborn neuropsychiatric disorders by aiming at core neurological mechanisms. Its lead compound, DT-101, is a positive allosteric modulator of AMPA receptors (crucial glutamate channels in the brain) and has already shown clear target engagement in a 60-patient Phase 1 trial for major depression. This mechanism differs from traditional SSRIs and follows renewed interest in glutamatergic approaches after ketamine's success – joining companies like Alto Neuroscience (NMDA modulators) and Sage Therapeutics (GABA modulators) in pursuing rapid-acting antidepressants. With Phase 2 trials for DT-101 set to start this year, Draig's war chest will also push two other preclinical programs (both novel GABA receptor modulators) into human studies by 2026. The massive funding reflects both the dire need for new psychiatric drugs and investor confidence in mechanism-based approaches: the last truly novel antidepressant mechanism (SSRIs) was discovered in the 1980s. The startup was co-founded by prominent neuroscientists at Cardiff University's Medicines Discovery Institute in partnership with SV Health Investors, and its syndicate features heavyweights like Access Biotechnology (lead), Canaan, SR One, Sanofi Ventures, and more. Draig also benefits from the UK's strength in neuroscience – home to companies like Compass Pathways (psychedelics) and GW Pharmaceuticals (cannabinoids, acquired by Jazz for $7.2B). Investors cite the huge unmet need in mental health – many patients fail to get relief from existing antidepressants – and Draig's differentiated strategy of targeting the glutamatergic and GABAergic systems as reasons for the strong backing. If Draig's "dragon-sized" bet pays off, it could herald a new wave of treatments for depression and related disorders that work faster and better for those not helped by current meds.
Evogene + Google Cloud unveil 38-billion-molecule AI for faster drug and agchem design
The Israeli computational biology company, Evogene, has announced completion of its ChemPass AI foundation model for molecule design, built in collaboration with Google Cloud's AI team. Trained on 38 billion molecular structures, the model can propose de novo compounds that allegedly satisfy multiple constraints (potency, safety, patentability, etc.), a task that normally takes chemists many iterative cycles. Impressively, Evogene reports the model achieved ~90% precision in generating viable, novel molecules meeting all targets in silico – versus ~29% using prior GPT-style approaches. This puts ChemPass in the same league as other chemistry foundation models like IBM's MoLFormer (1.1B molecules) and Recursion's LOWE (1T+ relationships), but with a focus on multi-parameter optimization. The AI was developed on Google Cloud's high-performance infrastructure (joining other GCP biotech partnerships like Colossal Biosciences and InSilico Medicine) and is now integrated into Evogene's ChemPass platform, which is used for both drug and crop-protection compound discovery. The dual pharma-agri focus is notable – while most AI drug discovery focuses on human therapeutics, Evogene's model can tackle pesticides, herbicides, and other agricultural chemicals where the $65B crop protection market desperately needs innovation. By searching a vastly larger chemical space (and avoiding "well-trodden" structures), the system aims to deliver more innovative hits and reduce late-stage failure rates. This advance hints at a future where designing a drug or pesticide is less trial-and-error and more like engineering – with AI rapidly searching chemical space for optimal candidates that humans and robots can then make and test.
Parallel Bio's organoid + AI approach gets $21M series A to speed up new therapies
Cambridge-based Parallel Bio closed a $21 million Series A, led by AIX Ventures with participation from investors like Marc Benioff (Salesforce CEO), to ramp up its human-first organoid drug discovery platform. The startup grows miniature human immune systems that mimic the diversity of real patients, then uses AI and automation to evaluate how drug candidates perform on these life-like models. By doing so, Parallel Bio aims to replace a chunk of traditional animal testing with more predictive human data early on – potentially shaving years and hundreds of millions off development. The approach addresses a critical pain point: 90%+ of drugs that work in mice fail in humans, largely due to species differences in immune responses. Parallel joins a wave of "human-on-chip" companies like Emulate and Hesperos, but focuses specifically on immune system modeling – crucial for vaccines, immunotherapies, and inflammatory disease drugs. Notably, three Big Pharma companies are already piloting the approach, and Parallel's tech helped partner Centivax validate a novel flu vaccine (Centi-Flu) entirely in vitro; that vaccine is now headed into human trials with confidence from organoid results. The platform builds on advances in organoid technology (similar to those used by Hubrecht Organoid Technology and Organoid Therapeutics) but adds sophisticated immune cell integration and AI-driven analysis. The fresh funds will go toward scaling up the AI/automation behind the platform, hiring more scientists and engineers, and expanding partnerships with pharma/biotech (beyond the initial immunotherapy use cases). If successful, Parallel Bio's "immune system in a dish" could accelerate drug R&D and improve success rates by revealing efficacy and toxicity signals on human tissue before entering costly clinical trials – a boon for patients waiting for new treatments.
New neurogenomics tie-up as Tempus looks to sift Northwestern University for Alzheimer's clues
Tempus has announced a partnership with Northwestern's Abrams Neurogenomics Center to apply artificial intelligence in the fight against Alzheimer's. Using Tempus's Lens analytics platform, the collaborators will crunch Northwestern's large repository of Alzheimer's patient genomic data to identify patterns and molecular signatures linked to the disease. The idea is to let AI sift through genetic variants, gene expression changes, etc., to uncover hidden contributors to Alzheimer's pathology – potentially revealing new drug targets or diagnostic biomarkers. This partnership adds to a growing list of AI-for-Alzheimer's efforts, including Verge Genomics's ALS/Alzheimer's programs, Insitro's neurodegeneration collaboration with Gilead, and the recent NIH BRAIN Initiative's cell atlas projects. The collaboration is particularly timely given the mixed success of recent Alzheimer's drugs (Leqembi, Aduhelm) and the field's pivot toward precision medicine approaches. This effort is notable because Alzheimer's is complex and heterogeneous; traditional approaches have struggled to yield effective therapies. Northwestern brings deep expertise through the Mesulam Center for Cognitive Neurology, while Tempus contributes its AI platform that's already proven in oncology (powering 50%+ of academic cancer centers). By marrying big data with machine learning, Tempus and Northwestern hope to speed up discoveries about which genes and cell types drive neurodegeneration, and why some people respond to treatments (or not). In the long run, such insights could inform more precise interventions – for example, stratifying patients by genetic subtype or pointing to entirely new therapeutic strategies – demonstrating how AI might accelerate progress in diseases where conventional R&D has largely hit a wall.
Liverpool ChiroChem bought by XtalPi to merge AI drug design with automated chemistry and compound synthesis
XtalPi, known for its AI and physics-driven drug design, is beefing up its wet lab capabilities by purchasing Liverpool ChiroChem (LCC), a specialist in high-throughput chiral molecule synthesis. LCC (a University of Liverpool spinout) brings a proprietary tech platform called PACE that uses robotics and AI to rapidly make diverse stereochemically-defined building blocks – basically, an engine for churning out novel chemical scaffolds with precise 3D geometry. The deal follows a trend of AI drug discovery companies acquiring synthesis capabilities – like Schrödinger buying Morphic Therapeutic and Recursion's acquisition of Cyclica and Valence – recognizing that great algorithms need great chemistry. By joining forces, XtalPi will span AI prediction and real-world synthesis across its bases in the US, Europe, and Asia. The strategic aim is to significantly widen the searchable chemical space for new drugs (and materials): XtalPi's algorithms can propose wild new molecular candidates, and LCC's automated labs can physically make and test them at speed. This addresses a key bottleneck in AI-driven drug discovery where computational predictions often suggest molecules that are difficult or impossible to synthesize using traditional chemistry. XtalPi's COO highlighted that integrating LCC's chiral chemistry expertise enables an end-to-end loop where AI predictions quickly become tangible compounds, reducing trial-and-error and uncovering structures medicinal chemists might never have considered. The combined entity will compete with integrated platforms like Atomwise's AIMS and Exscientia's full-stack approach. It's a marriage of bits and atoms that could accelerate innovation in pharmaceuticals, agrochemicals, and beyond by bridging the long-standing gap between design and synthesis.
New startup "Sesen" targets pharma's language barrier, unveils life-science vocabulary optimized LLM
A mysterious Boston-based AI company, Sesen, has opened its doors with a singular focus: providing translation and localization services exclusively for life-science content. That means everything from clinical trial protocols, regulatory submissions and drug labels to patient-facing materials can be expertly translated into target languages while maintaining scientific accuracy and compliance. The companies finances and backing are not being circulated, but given AI funding trend's and the broad aperture of the companies business model it is fairly safe to say it is substantial. Sesen's approach blends a worldwide team of native medical/pharmaceutical linguists with AI-enhanced workflows – including a proprietary large language model dubbed SesenGPT that's fine-tuned on life sciences terminology. This launch joins a growing ecosystem of specialized biotech language models including BioGPT, BioBERT, and BioMedLM, but focuses specifically on multilingual capabilities rather than just English comprehension. The timing aligns with pharma's globalization push: clinical trials increasingly span dozens of countries, FDA requires multilingual labeling, and the EU's Clinical Trials Regulation demands translations for patient documents. The result is faster turnaround on critical documents (think: a protocol for a global Phase III trial or an FDA submission) with rigorous quality control meeting ISO certifications for translation services. Sesen enters a market dominated by generalist translation giants like TransPerfect and Lionbridge, betting that life sciences expertise will win over pharma clients burned by mistranslations of critical safety data.
🗓️ Upcoming BiB Events
See our handy dandy Lu.ma event calendar HERE, please RSVP so folks can plan accordingly!
BiB Paris: RNA Therapeutics with A51 and Tycho Bio, June 19th at 7 pm
Copenhagen Bits in Bio – Summer Gathering, June 20th at 3:30 pm
Bits in Bio <> Toronto Tech Week Happy Hour, June 24th at 6:30 pm
Seattle Summer Biotech Happy Hour | BioBeers & Bits in Bio, June 27th at 4 pm
📰 Top Community Conversations
@Kumar is looking for a new project ideas - what techbio tools should exist that don't?
link to post
🏢 New Job Openings
Senior Scientist at Full Circle Labs
Bioinformatician/Data Scientist at SNIPR Biome
Founding Engineers at ScienceMachine
Various Roles at Certara
Senior Software Engineer at LabGenius
Senior Scientific Implementation Manager at Ganymede
🙋 🙏 Community Asks
Feedback: How is the Newsletter doing? We’re trying different formats/content. In case the hyperlink above didn’t get your attention, maybe a bright orange button will!
Volunteer: Want to get involved with Bits in Bio, meet new members across the community, and learn about the ecosystem? We are looking for volunteers to help us create great content and manage the community.
🙏 Thank you for being a BiB Weekly reader!
We want to deliver what matters most in Bio AI and would love your feedback on how we can do better. Please weigh in as anon here or DM me directly!
Vince Alessi
BiB Editor in Chief
 | A guest post by
|