Skip to content

AI in drug discovery

Structure prediction, generative design and virtual cells: verifiable progress for AI in discovery and the clinic.

141 selectedRelated topicsMethods and resourcesSmall molecules and degradersOriginal research

Latest picks

21–40 of 141
Oct 19Mon
  1. youtube.com59

    JAMA Network Launches JAMA AI, a New Journal Dedicated to AI and Digital Health

    JAMA Network announced the launch of JAMA AI, a new journal focused on AI and digital health. Editor-in-chief Roy Perlis said AI has evolved from a method applied across fields into a discipline in its own right. The journal is the 14th JAMA Network publication and runs alongside JAMA+ AI, which aggregates AI content from 13 journals.

    Why it matters: JAMA AI signals a shift from AI as method to clinical effectiveness evaluation, relevant to how readers assess AI diagnostic papers.

Oct 18Sun
  1. bioRxiv Genomics48

    Benchmarking the AlphaGenome Variant Impact score against SpliceAI on post-release ClinVar classifications and experimental splicing assays

    Researchers benchmarked the AVI score on post-release ClinVar classifications (87,666 rare single-nucleotide variants outside the training chromosomes) and on six experimental splicing assays, both ground truths beyond those used by the AlphaGenome developers.

    Why it matters: Independent benchmark on data not used by the developers

  2. youtube.com68

    HHMI Janelia proposes experiment-computation closed-loop model to predict nuclear condensates from protein sequence

    An HHMI Janelia team proposes an experiment-computation closed-loop model to predict nuclear condensates directly from protein sequences. They first measure condensate propensity at high throughput in cells using pooled protein sequence variant libraries, automatically quantifying condensate size and number, then use the physics-based predictor Finch to predict homotypic and heterotypic interaction strengths and extract sequence features controlling condensate size and number.

    Why it matters: The paper combines pooled-library high-throughput measurement with a physics-based predictor in a closed loop, useful as an experimental-computational reference for studying IDP phase separation.

Oct 16Fri
  1. mp.weixin.qq.com65

    insitro and AstraZeneca Strike Global Exclusive AI Drug Discovery Collaboration for MASH and Cardiomyopathy

    AI drug discovery company insitro has entered a global exclusive research collaboration with AstraZeneca on October 15 to discover new targets and develop candidates for MASH and cardiomyopathy. insitro receives a $40 million upfront payment and is eligible for up to about $1.46 billion in R&D and commercialization milestone payments plus single-digit royalties on future global sales. insitro will use its Virtual Human platform to identify disease-driving mechanisms and advance AstraZeneca-nominated targets to development candidate stage, while AstraZeneca handles clinical development, regulatory filings and commercialization.

    Why it matters: A large AI-driven discovery deal signals how big pharma is paying for machine-learning target discovery.

  2. mittrchina.com83

    David Baker Team Publishes Two Nature Papers on AI-Designed Proteases With Over 10-Billion-Fold Catalytic Acceleration

    David Baker's team published two Nature papers demonstrating AI-based de novo protease design. In the cysteine protease study, 13 of 69 designs cleaved the target substrate, with a maximum rate acceleration of about 30 million-fold. In the metalloprotease study, about 36% of designs showed catalytic activity; after optimizing with 4 amino acid substitutions, rate acceleration exceeded 10 billion-fold, and function was validated in mammalian cells and in cleaving disease-related proteins including TDP-43, Aβ, and SAA.

    Why it matters: The article gives success rates, catalytic acceleration factors, and cell validation results for both designs, allowing readers to compare them directly with natural enzyme catalytic efficiency.

  3. Nat Med69

    An open vision-language model for diverse medical applications.

    MedGemma, a collection of medical vision-language foundation models built on Gemma 3, outperforms similarly sized generative models across medical images and text. On out-of-distribution tasks it improves medical image question answering by 2.6-10%, chest X-ray finding classification by 15.5-18.1%, and agentic evaluations by 10.8% over base models. Its medically tuned vision encoder, MedSigLIP, matches or beats many specialized medical image encoders, and fine-tuning MedGemma beats fine-tuning Gemma 3 directly, especially with limited training data.

    Why it matters: Open foundation models let labs fine-tune medical AI on small private datasets instead of training from scratch.

  4. mp.weixin.qq.com69

    AstraZeneca Signs $1.86 Billion Exclusive AI Drug Discovery Deal with insitro for MASH and Cardiomyopathy

    AstraZeneca and AI drug discovery company insitro have entered a global exclusive research collaboration focused on metabolic dysfunction-associated steatohepatitis (MASH) and cardiomyopathy. insitro receives a $40 million upfront payment and is eligible for up to about $1.46 billion in development and commercial milestone payments plus tiered royalties on global sales. insitro will combine its technology platform with AstraZeneca's proprietary data to identify new druggable targets and advance candidates.

    Why it matters: Large AI drug discovery deal between a top pharma and insitro in MASH and cardiomyopathy.

Oct 15Thu
  1. Nature76

    Artificial intelligence-enabled clinical phenotyping of human brown fat.

    A research team used AI to systematically analyze brown adipose tissue (BAT) in 101,243 patients from a cancer center, finding BAT associated with lower prevalence of cardiovascular and psychiatric diseases. In about 8,700 patients, machine learning semi-automatically quantified BAT activity; higher activity, especially in the axillary depot, was linked to healthier metabolism, with better blood glucose, lipid, inflammatory and echocardiographic measures.

    Why it matters: A 100,000-person cohort with AI quantification in about 8,700 patients, plus losartan as a repurposing candidate, provides a starting point for validation.

  2. Nature77

    A microexon regulatory code reveals autism-linked genetic variation.

    An expanded map of neuronal microexons, built from extensive transcriptome profiling, adds hundreds of new microexons, including subtype-specific ones and others that trigger nonsense-mediated mRNA decay. A machine-learning sequence code predicts neuronally spliced microexons and identifies regulatory elements around them. Applied to autism, the code pinpoints rare splicing-altering variants, mostly in idiopathic individuals, that associate with autism status and converge on synaptic and other autism-linked genes.

    Why it matters: It links autism-associated genetic variation to a splicing code for neuronal microexons.

  3. mp.weixin.qq.com68

    Zealand Pharma Pays $1.2 Billion for Generate Biomedicines AI-Designed Multispecific Proteins to Extend Obesity Pipeline

    Zealand Pharma is paying $1.2 billion to bring Generate Biomedicines' AI-designed multispecific proteins into its obesity pipeline, adding a new modality alongside petrelintide, a long-acting amylin analogue, and survodutide, a glucagon/GLP-1 dual receptor agonist, both in Phase III. The Danish biotech's move comes after recent Phase III data for both candidates fell short of analyst expectations, putting pressure on its shares.

    Why it matters: A $1.2 billion licensing deal adds an AI-designed protein modality to a prominent obesity pipeline after weak Phase III results.

  4. BioPharma Dive61

    Iambic, an AI drug discovery specialist, banks $150M in an IPO

    Iambic Therapeutics, an AI drug discovery startup, raised $150 million in an IPO, selling close to 9.4 million shares at $16 each and trading on the Nasdaq from Thursday under the ticker "IAM." The company has drawn $462 million in private funding since 2019 and will use the proceeds to advance its first drug, IAM1363, an oral tyrosine kinase inhibitor for HER2-positive solid tumors that showed anti-tumor activity and encouraging safety in a Phase 1 study, with a registrational trial possibly beginning in 2027. Behind it are IAM217, targeting KIF18A in ovarian and triple-negative breast cancers, and a CDK2/CDK4-targeting prospect, both potentially entering testing next year.

    Why it matters: The offering adds to a wave of AI drug discovery startups testing whether public investors will back the technology.

  5. Nature71

    Artificial intelligence-enabled clinical phenotyping of human brown fat

    AI-enabled quantification of brown adipose tissue (BAT) activity in humans links higher BAT activity to a healthier metabolic profile and lower rates of psychiatric disease. The same approach identified existing pharmacologic agents that could be repurposed to activate BAT in humans. The work provides a scalable way to phenotype BAT and connect it to clinical outcomes and drug targets.

    Why it matters: It shows AI-based BAT phenotyping at scale, linking brown fat to metabolic and psychiatric outcomes and to repurposable drugs.

  6. Nature80

    Computational design of metalloproteases

    RFdiffusion was used to design proteins de novo with water-activating zinc-based active sites positioned next to a peptide substrate binding pocket. The resulting proteases were highly active and programmable, with the authors pointing to potential applications in therapeutics, biotechnology and bioremediation.

    Why it matters: De novo design of active, programmable metalloproteases extends computational protein design to catalysis.

  7. Nature82

    De novo design of cysteine proteases

    Designed cysteine proteases from scratch using a computational design pipeline built on RFdiffusion2 active-site scaffolding, ProteinMPNN sequence design, and AlphaFold 3 prediction. Candidate enzymes were characterized by crystallography, biochemical kinetic assays, and mammalian-cell imaging, extending the de novo enzyme design approach to the cysteine protease class.

    Why it matters: De novo design reaches cysteine proteases, a class previously outside computationally designed enzymes.

Oct 14Wed
  1. BioSpace63

    Zealand pens up to $1.2B deal with Generate for AI-designed obesity, metabolic drugs

    Zealand Pharma will pay Generate:Biomedicines $30 million upfront, up to $35 million in early preclinical milestones and nearly $1.2 billion in development, regulatory and commercial milestones, plus tiered royalties, to access Generate's generative AI platform for designing protein therapeutics for obesity and metabolic diseases. Generate handles research through candidate selection; Zealand leads clinical development, regulatory and commercialization. The deal follows mixed Zealand readouts for petrelintide and survodutide.

    Why it matters: Large AI drug-discovery alliance in obesity adds Generate's platform to Zealand's metabolic pipeline after mixed trial results.

  2. Nat Med71

    MAGIC: an international network for evaluating generative artificial intelligence in global health.

    The authors establish MAGIC, an international network for evaluating generative artificial intelligence in global health. The network brings together collaborators across institutions in Asia, Europe, Africa and the Americas, including groups at Duke-NUS Medical School, Harvard, Tsinghua and the World Health Organization Regional Office for South-East Asia, with a focus on the ethics, regulation and standardization of AI for health.

    Why it matters: Sets out an international framework for evaluating generative AI in global health, relevant to clinicians and researchers deploying these tools.

  3. mp.weixin.qq.com75

    SPARCS Screen of About 70 Million Cells Recovers Macroautophagy Genes and Golgi Control of STING Signaling

    SPARCS, a spatially resolved CRISPR screening method, identifies abnormal cells from microscopy images, recovers them, and traces the associated genes and protein changes. Applied to roughly 70 million cells, it recovered nearly all core macroautophagy genes and uncovered regulation of STING signaling by Golgi pH state, revealing mechanisms conventional screens tend to miss.

    Why it matters: It shows image-based recovery of complex phenotypes can find mechanisms standard CRISPR screens overlook.

  4. youtube.com76

    HHMI Janelia Releases CellposeSAM, a Cell Segmentation Foundation Model Combining SAM Encoder with Cellpose Flow Fields

    HHMI Janelia has released CellposeSAM, a foundation model for cell segmentation. It replaces the original UNet with the SAM image encoder (256×256, 8×8 patches, global attention at all layers) to predict cell flow fields, and masks are obtained by gradient tracking.

    Why it matters: The source gives the concrete architecture and benchmark data for combining the SAM visual encoder with Cellpose flow fields, which readers can directly transfer to their own segmentation tasks for comparison.

Oct 13Tue
  1. BioSpace56

    Pfizer pens $500M+ deal with Enodia to destroy disease-causing proteins

    Pfizer has signed a deal with Paris-based Enodia Therapeutics worth up to $500 million, covering an upfront payment plus research, development and sales milestones, with tiered royalties on net sales. Enodia will screen, optimize and develop small-molecule inhibitors of the Sec61 membrane protein complex, while Pfizer handles clinical studies, manufacturing, regulatory and commercial work. Target indications are not yet specified. Pfizer's venture arm had co-led Enodia's €20.7 million ($25 million) seed round in January.

    Why it matters: Pfizer commits over $500 million to a Sec61-targeting platform less than a year after the biotech's launch.

Oct 12Mon
  1. mittrchina.com81

    HealthFlux world model keeps 20-year health predictions accurate and simulates drug effects across 1,256 diseases

    HealthFlux, a pan-modal human health world model posted on medRxiv by Stanford postdoc Ying Kejun and colleagues, maintains a continuous 384-value health state trained on UK Biobank data spanning 11 data types including records, blood tests, plasma proteomics, metabolomics and MRI. In about 60,000 held-out participants it predicted 1,010 diseases and death over 5 years with an average AUROC of 0.832 versus 0.805 for RisQ, and after 20 years of autonomous rollout without new data it held 0.837 while Delphi-2M fell to 0.576. Simulated drug effects for 210 drug-disease combinations matched published clinical trials within 18%, and screening 259 drugs across 1,256 diseases suggested metformin cut 10-year mortality risk by about 21% and simvastatin cut major disease risk by about 35%.

    Why it matters: A world model that keeps predicting health trajectories accurately over 20 years without new data, and simulates drug effects consistent with trials, changes how lifelong health and interventions can be studied.

CBACBA Frontier Insights (CFI)Chinese Biopharmaceutical Association (CBA)www.cba-usa.org

Disclaimer CBA Frontier Insights uses AI-assisted methods to aggregate, summarize, and analyze publicly accessible biopharmaceutical information. Content is provided for informational and educational purposes only and does not represent the official views of original sources or affiliated organizations. AI-generated content may contain errors; please verify important information against primary sources. Intellectual property rights remain with their respective owners. For copyright concerns or corrections, please contact us through the website’s feedback page.