How Startups Are Innovating AI-Designed Medicines

Biotech company Insilico Medicine is using AI to rethink drug discovery. Now, with a fresh $110 million in funding pushing its valuation past $1 billion, the startup is considering a Hong Kong IPO. Developing a new drug is notoriously slow and expensive—it can take over a decade and billions of dollars before a single treatment reaches patients. Insilico wants to change that with AI, making the process faster, cheaper, and more precise. ► At the core of Insilico’s approach is Pharma.AI, which analyzes vast biological datasets and predicts which molecules are most likely to work, reducing the need for excessive trial and error.  ► Insilico is already delivering results—its leading drug candidate, Rentosertib, for a serious lung disease, reached early clinical trials in just 2.5 years, a process that normally takes up to six.  ► The company is making big deals by licensing its AI-generated drugs to pharmaceutical giants like Sanofi and Fosun Pharma, securing $3.5 billion in contract value. The startup’s pipeline now includes 30 drug candidates, with 10 receiving clearance from the US Food and Drug Administration (FDA) to proceed with human trials. Beyond discovery, Insilico is using AI to speed up lab work. The company is testing humanoid robots in its China lab to automate repetitive tasks, collect data, and reduce human error, all in an attempt to speed up research. If Insilico succeeds, it could reshape the entire drug discovery process. By combining AI, real drug candidates, and even lab robots, the company is tackling the whole process, not just one piece of it. Faster, cheaper, and maybe even better—this could be a glimpse of how future drugs come to life. #artificialintelligence #innovation

NEW: The next generation of AI-focused biotechs is sprinting into the clinic. By reviewing the pipelines of 11 of these startups, mainly founded between 2018 and 2021, I found the number of clinical-stage drugs is expected to nearly triple from today's total of eight to 21 in 2026. The clinical focus reflects a shift in both founders and investors valuing molecules more than models. There's plenty of examples, from Generate:Biomedicines starting Phase 3 trials to Iambic Therapeutics raising $100 million on the heels of its first clinical readout, planning on using the cash to bring two more drugs into the clinic. Next year will start providing some human data that may start to answer the question if the latest AI advances are actually meaningful enough to boost R&D productivity. Read more in my latest feature for Endpoints News, previewing 2026 for the AI bio world. Thanks to AbCellera CEO Carl Hansen and Amplify's Elliot Hershberg for sharing some thoughts: https://lnkd.in/eaWEVHh2

🧬 𝐀𝐈 𝐢𝐬 𝐪𝐮𝐢𝐞𝐭𝐥𝐲 𝐫𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐢𝐳𝐢𝐧𝐠 𝐚𝐧𝐭𝐢𝐛𝐨𝐝𝐲 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭. If you’ve been watching closely, you’ll know that the way we discover and design antibodies is undergoing a seismic shift. What used to take years in the lab is now being compressed into weeks, thanks to the rise of AI/ML platforms that are not just accelerating discovery, they’re fundamentally changing how we think about biologics. Let’s start with de novo antibody design. This is where AI builds antibodies from scratch, without relying on natural templates. It’s like designing a key without ever seeing the lock, except the key fits perfectly. One of the most exciting players here is Nabla Bio, which recently signed a $1B+ deal with Takeda to deploy its Joint Atomic Model (JAM) platform. JAM can generate functional antibodies in just a few weeks, including binders to notoriously difficult targets like GPCRs. Then there’s Absci pioneering zero-shot generative AI for antibody design. Their platform doesn’t need prior examples to create high-affinity binders, and they’ve already partnered with Merck, AstraZeneca, and Twist Bioscience. And now, meet DenovAI Biotech, a rising star out of Israel’s AION Labs. With $4M in seed funding and support from AstraZeneca, Merck, Pfizer, Teva, and AWS. The are developing an AI-powered biophysics platform to design antibodies completely de novo. On the other side of the spectrum, we have epitope discovery, the art of finding new regions on antigens that antibodies can bind to. This is especially critical in cancers where traditional targets have failed. Sanavia Oncology Inc. is making waves here, having raised $23M to develop its AI-powered discovery engine. Their lead candidate, SANA-01, targets a novel epitope found in a majority of lung, breast, and pancreatic cancers. And don’t overlook BigHat Biosciences, which is using its Milliner™ platform to optimize antibodies for affinity, manufacturability, and immunogenicity with Eli Lilly as partner. ⚠️ But It’s Not All Sunshine and Rainbows: Challenges ahead ➡️Stability: AI-generated sequences may fold correctly in silico but fail under physiological conditions. ➡️Immunogenicity: Even well-designed antibodies can trigger unwanted immune responses if not carefully screened. ➡️Solubility: Poor solubility can derail promising candidates during formulation or delivery. ➡️Internalization: For therapeutic efficacy, especially in oncology, antibodies often need to be internalized by target cells, something AI models don’t always predict well. ➡️Affinity: High binding affinity is essential, but AI must balance it with specificity and off-target risks. 💬 Let’s spark a conversation: 👉 Which AI-native biotech do you think will lead the next wave of antibody innovation? 👉 Is your portfolio or pipeline aligned with the future of programmable biologics? Would love to hear your thoughts, let’s connect. #Biotech #AI #VC #Investment #BD #investor #FoF

Gold Diggers of Pharma: Turning Abandoned Drugs into Blockbusters In the shadows of pharmaceutical R&D lies a quiet revolution. A new breed of biotech startups—Pharma’s Gold Diggers—are mining failed, forgotten, or shelved drugs and turning them into clinical and commercial successes. Why Drugs Fail (Yet Still Hold Value) Not all failures are final. Many drugs stumble in trials due to: * Narrow misses on clinical endpoints * Side effects in limited populations * Lack of commercial fit for Big Pharma * Being ahead of their time Startups now scout these “scrap heap” molecules, reviving them at a fraction of original R&D costs—and often in a fifth of the time. Meet the Drug Resurrectors 1. Ignota Labs – AI-Powered Resurrection
Uses AI to identify why Phase 2/3 drugs failed (e.g., liver toxicity), tweaks the chemistry, and revives them.
Highlight: A PDE9 inhibitor for Alzheimer’s, backed by $6.9M in seed funding. “We don’t start from zero. We start from almost there.” 2. Cycle Pharmaceuticals – Formulation Wizards
Transforms delivery formats (e.g., injectable to inhalable), eliminates cold chains, and redirects drugs to rare diseases.
Example: Reformulated glatiramer acetate for cystic fibrosis. “Formulation is our innovation engine.” 3. Melior Discovery – Phenotypic Pivoters
Screens old drugs on new disease models. Partnered with Pfizer, Merck, and AstraZeneca.
Repositioned an anti-inflammatory for Type 2 diabetes. 4. Recursion Pharmaceuticals – High-Throughput Innovators
Combines AI + cell imaging to screen 1M+ compounds across diseases.
Backed by Bayer & Sanofi. “We’re mapping the druggable universe using data.” 5. Algernon Pharmaceuticals – Rare Pathway Hunters
Repurposes generics for niche conditions.
Example: Ifenprodil, once a neurodrug, now in trials for pulmonary fibrosis and chronic cough. Famous Pharma Revivals * Sildenafil (Pfizer): Failed angina drug → Viagra; now over $2B/year. * Thalidomide (Celgene): Once banned, now approved for multiple myeloma. * Minoxidil (Upjohn): From blood pressure drug to Rogaine for hair growth. * Avastin → Lucentis (Genentech): Cancer biologic → macular degeneration therapy. Signs of the Next Gold Rush Look out for: * Licensing of old Phase II/III assets * Startups with AI repurposing platforms * Announcements of fast-tracked approvals * Novel formulations: sprays, patches, sublinguals, nano-tech Closing Insight
“Every failed drug is a story half-written. These startups are writing its second chapter.” In a high-cost, high-risk R&D world, these companies show that yesterday’s failures may still hold the cures of tomorrow.

AI drug discovery IP is shifting from the molecule to the model, and now to something harder to see: the recursive discovery factory. Xaira Therapeutics broke two years of silence after raising $1 billion with an AI-first generative protein platform, reflecting a broader shift toward methodology IP. Genentech, TCS, and DeepMind lead methodology filings in this space, building patent positions on the computational workflows that power AI-driven drug design. The next layer of enforceable IP goes further, into the "lights-out lab": fully automated, robot-to-robot experimental systems running recursive, AI-driven trials. These systems close the loop from hypothesis to experiment to refinement without human intervention. As these physical workflows become observable and reverse-engineerable, the defensible position moves from model weights to how the system learns, adapts, and connects across each robotic step. The industry is watching the molecule layer and starting to notice the model layer. The companies that file on the recursive discovery factory will hold the IP that matters most in five years.

2025 could be a pillar year for AI-driven lab robotics in pharma and biotech, and here is why 🤖👇 📑 If you missed it, this recent publication from one of our portfolio companies, Insilico Medicine, is worth a close read (link in the comments). It describes how their AI platform was used to discover a novel TNIK inhibitor (INS018_055), a drug candidate with senomorphic activity targeting cellular senescence and SASP. But the core story isn’t just about the discovery of a new aging-related drug candidate, what’s more important—and in my view, a real paradigm shift—is the infrastructure and workflow behind it. ⚙️ Insilico demonstrates a fully integrated AI–robotics loop for drug discovery: 👉 Multi-omic data generation (genome, transcriptome, methylome) 👉 AI-driven target discovery via their PandaOmics platform 👉 Automated compound screening, phenotypic profiling, and wet-lab validation 👉 Real-time feedback and reinforcement learning to refine the models 👉 All executed within a modular six-unit robotics facility, handling everything from cell culture and compound management to high-content imaging and next-gen sequencing This kind of closed-loop system effectively collapses what used to take years into weeks. It's scalable, reproducible, and designed for continuous learning—a foundational shift in how we approach drug development. These end-to-end AI-native infrastructures are still rare. Insilico is one of only a few companies operating at this level, but I believe this is the blueprint for the future of drug discovery. 🚀 Kudos to Alex Zhavoronkov, Alex Aliper and the entire Insilico team for building and executing on a vision that brings true vertical integration to AI-powered drug discovery! 🙏 Image credit: Insilico Medicine

The pharmaceutical industry is gradually reaching an inflection point in AI adoption. As of early 2026, investment is projected to surge from $4 billion in 2025 to $25 billion by 2030. Yet only 5% of pharma companies successfully scale these technologies to deliver measurable value. The challenge is economics where bringing a drug to market costs $2.23 billion with R&D returns at 5.9%. AI is positioned to reverse Eroom's Law which is the observation that discovery becomes slower and costlier over time. While 95% of companies invest in AI, most remain trapped in pilot purgatory. First, agentic AI systems are slowly replacing chatbots. These autonomous agents draft clinical protocols, execute compliance checks, and optimize supply chains with minimal human intervention. Second, EU AI Act enforcement seems to be creating bifurcated strategies such as US hyperscaler models for global operations or sovereign European clouds for compliance. Third, open source models like DeepSeek R1 commoditized biomedical reasoning, cutting costs by 86%. AlphaFold 3 is achieving 50% higher accuracy in protein ligand prediction versus traditional methods. NVIDIA's BioNeMo are becoming the front end foundational model framework for generative biology. Also many drug discovery startups are using AI designed molecules in Phase II trials, claiming to compress drug discovery to range of 18 months from 6 years. Many Pharma players are exhibiting strong results using Biology Foundation Models and LLM framework. AstraZeneca deployed AI lung cancer screening ,creating market demand. Sanofi's Maestro runs real-time supply chain simulations. Novartis built NextGen clinical platforms on AWS, accelerating database lock and Pfizer's PACT initiative enables rapid GenAI prototyping. Digital twins simulation of Pharma Supply Chains are exemplified by few companies . For e.g Siemens' Industrial Copilot reduces repair time significantly where as Merck deploys Golden Batch analytics using bioreactor data to predict yields and prevent batch failures. FDA's Elsa AI creates adversarial review dynamics where companies optimize submissions for AI ingestibility. GenAI automates CTD drafting and pharmacovigilance narratives, transforming compliance into competitive advantage across many Pharma companies. 2027 winners won't be defined by budget size but by organizational rewiring, sovereign infrastructure and cross functional talent. With development costs at $2.23 billion per asset, AI isn't discretionary but it's the only scalable solution to restore innovation economics and competitive positioning. #LifeSciences #PharmaAI #DrugDiscovery #AgenticAI #DigitalPharma #BiopharmaInnovation #AIinHealthcare #SovereignAI #ClinicalTrials #RegulatoryAffairs #PharmaManufacturing #AlphaFold3 #BioNeMo #GenerativeBiology #HealthTech #PharmaLeadership #AIStrategy #DrugDevelopment #PrecisionMedicine Disclaimer: The opinions are mine and not of employer's

Google DeepMind Spinout Isomorphic Labs Nears Human Trials for AI-Designed Cancer Drugs: 💊The company’s platform is powered by AlphaFold3, the latest iteration of DeepMind’s Nobel-winning AI that predicts protein structures and models drug-target interactions 💊Its lead candidates, including cancer drugs, are currently moving through preclinical development, with human trials expected to begin soon 💊The goal isn’t just one breakthrough drug, but a general-purpose AI engine that can be applied across multiple diseases and modalities 💊The company aims to improve speed, cost, and success rates in drug discovery, reducing pharma’s current 10 percent trial success odds. AlphaFold gives scientists a head start by predicting how well a molecule might bind to a disease-relevant protein target, a key early step in drug design 💊Isomorphic ultimately hopes to turn drug discovery into something closer to design automation: “click a button, get a candidate,” with AI doing the heavy lifting. If successful, it could reshape not just timelines, but how pharma allocates resources and defines early-stage risk 💊Isomorphic has raised $600 million (led by Thrive Capital) and signed major R&D deals with Novartis and Eli Lilly and Company, supporting both external and in-house drug programs #DigitalHealth #AI #Pharma

While Silicon Valley fights over politics, Reid Hoffman tackles cancer. His AI startup Manas AI could transform healthcare completely: Manas AI combines generative computational chemistry with lab testing, making drug discovery 100x faster. Their molecular docking compresses development from years to months. Hoffman calls this "the jagged frontier" - AI and humans working together to achieve what neither could alone. With Microsoft's computational firepower and guidance from five Nobel laureates, Hoffman's vision looks increasingly achievable. His ultimate goal? "AI doctors in every home." World-class clinical expertise available to everyone through AI on smartphones. This solves a massive global problem: half the world lacks basic health services. According to WHO data, low-income countries have fewer than 2 doctors per 10,000 people while high-income nations have over 30. AI could provide preventative care and early diagnosis at unprecedented scale. Hoffman believes medical regulation causes more harm than good. "To really drive innovation, you have to allow a certain amount of error," he explained. He advocates for targeted regulations addressing specific harms rather than blanket restrictions. For Hoffman, AI will reduce humanity's existential risks: • Pandemics • Climate change • Nuclear war "We evolve through technology," he stated at Oxford. The key question isn't "can AI diagnose disease?" It's "which 65% of a workflow can AI handle while humans focus on the rest?" When humans and AI team up, they get better results than either would alone. Reid's rare ability to see entire systems - not just individual AI applications - is what truly separates visionaries from the crowd. If you want to build AI-native teams that enhance human creativity rather than replace it, follow me for more insights, AI news, and stories like this. #AI #AINews #ReidHoffman

🔥 AI didn’t just save a life. It’s rewriting the playbook for rare disease treatment.🔥 💎 Joseph Coates, once given days to live due to POEMS syndrome, is now in remission—thanks to an unconventional treatment suggested not by a doctor, but by an AI model. 👉 Developed by Dr. David Fajgenbaum’s team at the University of Pennsylvania, the AI system repurposed a cocktail of existing drugs—chemotherapy, immunotherapy, and steroids—that hadn’t been tried for his condition. Within a week, Joseph responded. Months later, he underwent a successful stem cell transplant. This is the power of AI-driven drug repurposing. 🛑 Drug repurposing isn’t new—think Viagra or Minoxidil—but AI is now enabling researchers to scan thousands of drugs against thousands of rare diseases at scale, uncovering hidden therapeutic opportunities in existing FDA-approved medications. 🔮 The challenge? 90% of rare diseases have no approved treatments, largely due to limited commercial incentives. AI levels that playing field—fast-tracking low-cost, life-saving solutions from existing data and overlooked research. 🔮 Platforms like Every Cure and other AI models from Harvard, Stanford, and UAB have already: 📌 Identified a Parkinson’s drug to treat a rare neurological condition. 📌 Used amphetamines to relieve paralysis in children. 📌 Even found that inhaled isopropyl alcohol can ease chronic nausea. This isn’t about replacing physicians—it’s about augmenting them. Doctors still make the call. But now, they’re armed with faster, smarter insights that may change the trajectory of a life when no other options remain. As Dr. Fajgenbaum says, “Someone had to be the first to try.” And when they do, patients like Joseph get a second chance. 💎 AI won’t solve everything. But in the world of rare disease, it might just be the most human application of technology we’ve seen yet. #HealthcareInnovation #AIMedicine #RareDiseases #DrugRepurposing #LifeSciences #DigitalHealth #HealthTech #AIforGood #PrecisionMedicine #FutureOfMedicine