The FDA approved Rasonque (daraxonrasib) on August 26, 2026 for metastatic pancreatic adenocarcinoma in patients previously treated with at least one systemic therapy. It is the first direct RAS-targeted oral drug approved in pancreatic cancer, a disease where more than 90% of patients carry KRAS mutations. The approval comes 44 years after RAS was identified as the first human oncogene in 1982.
On September 23, an FDA advisory panel of 10 experts voted 6-4 on Grail's multi-cancer early detection test Galleri. Galleri has been offered in the US as an LDT since mid-2021, with nearly 600,000 tests sold by the end of June 2026 and screening revenue of $42.6 million in the second quarter of 2026 alone. The panel's divided vote precedes further FDA consideration of the test.
Single-cell multi-omics of small-cell lung cancer resolved distinct cell states along the neuroendocrine-to-nonneuroendocrine axis, including a previously unrecognised Intermediate state of low-identity immunogenic cells that serves as an obligatory route in lineage transition. A Boolean Bayes model of transcription factor networks (BoBa-T) predicted state regulators, and RORB was validated as a gatekeeper of the neuroendocrine state: its inactivation pushes NE cells into the Intermediate state, upregulates immunogenic programs, and inhibits tumour growth in immunocompetent hosts.
A METTL3–NUP93 interaction couples m6A-mediated nuclear mRNA export to cholesterol biosynthesis, driving castration-resistant prostate cancer. Overexpression of wild-type METTL3 or NUP93, but not enzymatically dead METTL3 or the METTL3-interaction-deficient NUP93 mutant, raised intracellular androgens, activated AR signaling under castrate conditions, and promoted androgen-independent growth in vitro and in vivo. Pharmacological inhibition of METTL3 or targeted demethylation of mRNAs encoding cholesterol biosynthesis enzymes suppressed CRPC malignancy.
Disrupting the SWI/SNF chromatin remodeling complex in tumor-associated macrophages reprograms them and improves checkpoint blockade immunotherapy. Pharmacologic inhibition of SWI/SNF ATPase activity slowed tumor growth, and deleting the SWI/SNF subunit Arid1a in myeloid cells was sufficient to suppress tumor progression and enhance checkpoint blockade response. The reprogramming involved reduced enhancer accessibility at poor-prognosis genes such as Spp1, increased accessibility at interferon-stimulated gene promoters, elevated CD86, and required CD86 costimulation and CD8+ T cells.
Cyclin B3, previously thought to act only in female meiosis, is expressed in mitotic cells and partners with CDK1 and CDK2 to drive timely mitotic progression through proper chromosome alignment and stable kinetochore-microtubule interactions. Depleting cyclin B3 disrupted mitosis in most cancer cell lines tested, including colorectal, osteosarcoma, and breast cancer models, while non-transformed cells were largely unaffected. Phospho-mass spectrometry identified kinetochore and mitotic spindle components as substrates, pointing to a preferential cancer-cell dependency.
García-Gómez et al. show that brain-seeding cancer cells enter an MXD4-dependent proliferative pause that restrains MYC and activates protective stress responses, allowing survival under stress before metastatic outgrowth. The pause represents a vulnerability that could be targeted to eradicate micrometastases before clinical relapse. The work appears in Cancer Cell.
Eight recurrent malignant epithelial states in non-metastatic HR+/HER2- breast cancer show distinct ctDNA release potential, defined by scMASTER, a framework combining single-cell transcriptomes and mutations, spatial transcriptomics and matched tumor and plasma DNA sequencing across ten patients. Antigen-presentation, stress, mitosis and mesenchymal-like states released most ctDNA, as did EMT-enriched regions, intraductal luminal progenitors and avascular hypoxic zones. PIK3CA and ESR1 mutations tied to distinct functional states altered ctDNA detectability, and a stratification model from these features identified high-risk, immunosuppressed patients with poor prognosis in external cohorts.
Non-canonical HLA class I peptides could widen the target pool for T-cell immunotherapy in medulloblastoma, a tumor with low mutational burden. Integrated immunopeptidomics and transcriptomics across 21 samples covering the four molecular subgroups identified 2,298 non-canonical HLA-I peptides, 598 of them cancer-restricted, derived mainly from retained introns, out-of-frame open reading frames, and long non-coding RNAs. More than 70% were predicted to bind a patient-matched HLA allotype and about 87% to be immunogenic.
Platinum chemotherapy drives DNA damage response gene-mutant clonal hematopoiesis more strongly than PARP inhibitors, an analysis of treatment-associated clone behavior shows. Carrying germline mutations that confer homologous recombination deficiency slowed the expansion of these clones during treatment. The findings separate the mutagenic pressure of the two drug classes and identify germline HR status as a modifier of TP53-mutant clone fitness.
Why it matters: It distinguishes platinum from PARP inhibitor effects on clonal hematopoiesis and adds germline HR status as a modifier of clone expansion.
A common germline missense variant in BLTP3A, M1098T, impairs lysosomal stress resolution in high-grade serous ovarian cancer, prolonging STING activity and tumor-derived IFN-λ. Patients carrying M1098T had longer overall survival and tumors enriched for cytotoxic T cell-myeloid neighborhoods. In syngeneic tumors, Bltp3a loss or the orthologous variant delayed progression and generated Cxcl9+ antigen-presenting macrophage niches, with tumor-cell STING and IFN-λ required for immune remodeling and tumor control.
Base editor scanning across 22 MAPK genes, performed under pathway hyperactivation or inhibition at various nodes, produced a chemical-genetic interaction map of the RAS-RAF-MEK-ERK cascade. Drug resistance mutations often arose in proteins other than the direct drug target, catalytically impaired CRAF mutants gave bidirectional resistance and sensitization to structurally similar MEK inhibitors, and an allosteric site was found in the KRAS N-terminus. The work establishes BE scanning as a scalable method for network-level chemical-genetic mapping in a major oncogenic pathway.
Suppressive neutrophils (PMN-MDSCs) depend on mSWI/SNF chromatin remodeling to maintain their identity, and two orally bioavailable mSWI/SNF ATPase antagonists, a degrader and a catalytic inhibitor, produced immune-dependent tumor control in syngeneic models, including tumors resistant to PD-1 blockade. Treatment depleted intratumoral PMN-MDSCs, restored functional CD8+ T cells, enhanced checkpoint therapy without overt toxicity, and blocked PMN-MDSC expansion in mouse and human cells. Mechanistically, antagonism collapsed chromatin accessibility and PU.1/C/EBPβ occupancy at regulatory elements sustaining PMN-MDSC identity.
Spatial transcriptomics of 49 primary DCIS specimens shows stromal cell gradients, with TREM2+ macrophages and myofibroblastic cancer-associated fibroblasts enriched in tumor-proximal niches. TREM2+ macrophage abundance near tumor nests is associated with recurrence as invasive breast cancer, and ligand-receptor analysis implicates GRN-SORT1, SPP1-CD44, and LGALS9-CD44 signaling between these macrophages and epithelial cells. Findings were validated by multiplex immunostaining.
China now accounts for 82% of the 140 in vivo CAR-T-cell therapies in development globally, according to an analysis by John Wu at Boston Consulting Group and colleagues published last month. Most foundational preclinical work came from the United States and Germany, but China has moved fastest to the clinic, aided by government support and investigator-initiated trials that bypass drug-regulator oversight. The FDA this month approved the first China-developed in vivo CAR-T-cell therapy, for blood cancers, for clinical testing.
LENTRA, an antigen-targeted lentiviral platform, genetically reprograms tumor cells in vivo to produce therapeutic proteins directly within the tumor after intravenous administration. In multiple murine solid tumor models, tumor-restricted expression of anti-CTLA-4 or IL-12 triggered potent antitumor responses without systemic toxicities, while sialidase secretion drove dendritic cell and T cell infiltration, macrophage reprogramming, tumor regression and durable immune memory. LENTRA also mediated efficient transduction, immune remodeling and reduced tumor viability in patient-derived tumor explants.
Glutaminase-1 acts as a metabolic checkpoint limiting T cell infiltration and function in head and neck squamous cell carcinoma (HNSCC). Pharmacologic inhibition or genetic deletion of glutaminase-1 increased tumor susceptibility to immune cell-mediated cytotoxicity and sensitized patient-derived tumor fragments to immune checkpoint blockade, via disruption of glutathione-dependent ROS defense, impaired tumor proliferation, and increased MHC expression. Breast cancer fragments, which show low glutaminase-1 expression, were not responsive, supporting patient stratification for metabolism-targeted immunotherapy.
FRANK, a neural network classifier for pan-cancer pediatric tumors, distinguishes 181 childhood tumor and normal tissue types from RNA-seq data. Trained on 11,467 curated transcriptomes with data augmentation and a hierarchical penalty, it detected tumor types from as few as three training examples and reached 98.3% overall accuracy across nine validation datasets totaling 16,398 samples. It outperformed other published pediatric pan-cancer RNA-seq classifiers on an external dataset and remained accurate on long-read, noisy and low-quality data.
Why it matters: It sets a new accuracy benchmark for RNA-seq-based classification across 181 pediatric tumor types, including with sparse and low-quality data.
An inducible CRISPR interference platform with a custom sgRNA library enabled gene-specific repression of RNA polymerase III-transcribed genes and Pol III machinery. Genome-wide screens in diploid fibroblasts and HEK293T cells uncovered several Pol III dependencies, with multiple initiator methionine tRNA genes among the strongest fitness dependencies. Glioblastoma models, however, remained largely insensitive to repression of both individual Pol III genes and core Pol III components despite efficient target knockdown, challenging the assumption that cancer cells uniformly depend on elevated Pol III activity.
Overall survival extrapolations in Swedish TLV oncology appraisals were frequently conservative: reapplying original assumptions to longer follow-up data raised the average area under the curve of extrapolated OS by 8.8% (95% CI 5.2% to 12.6%) under TLV assumptions and by 10.9% (95% CI 6.9% to 15.1%) under manufacturer assumptions. The findings argue for revisiting survival model fit and treatment-waning assumptions when longer-term data become available.
