At the JPMorgan conference BioNTech unveiled its next‑generation antibody‑drug conjugate (ADC) DB‑1311 (BNT324). The company said it will launch a pivotal Phase III trial as a first‑line therapy for metastatic castration‑resistant prostate cancer (mCRPC) after Phase I/II data showed a 42.3% overall response rate and a 90.4% disease‑control rate in heavily pre‑treated patients. BioNTech believes DB‑1311 could become a best‑in‑class, pan‑cancer asset and has already started Phase II studies in other solid tumours, both as a single agent and combined with BNT327. In parallel, Gilead Sciences completed a multibillion‑dollar acquisition of Ouro Medicines, securing the BCMA×CD3 bispecific antibody CM336 originally from Kymon Biosciences. Early trials in autoimmune hemolytic anemia and primary immune thrombocytopenia showed complete peripheral B‑cell clearance for up to six months without the cytokine‑release‑syndrome or neurotoxicity that often limit immunotherapies. Chinese biotech firms are also leading the global PD‑1/VEGF bispecific market, with more than half of the pipeline coming from China. Recent deals have triggered nearly 20 milestones and over $1.5 billion in payments in the past six months, underscoring the rapid rise of Chinese innovative drugs on the world stage.
Read moreLung cancer kills more people than any other cancer, and CT scans are the main tool doctors use to find early signs. But the sheer number of scans means radiologists are rushed, and their diagnoses can vary widely depending on experience. To tackle this, a team of Chinese researchers created an artificial‑intelligence system called DeepFAN that looks at both the big picture and the tiny details of each nodule. DeepFAN combines three neural‑network modules: a visual transformer that captures the overall shape and surrounding tissue, a 3‑D residual network that picks up fine features like density and spiculation, and a graph‑convolutional network that merges these insights into a single, comprehensive view. The model was trained on more than 11,000 nodules from nine hospitals and then tested in a blind trial with 463 nodules across three independent centers. The results are striking: on its own, DeepFAN achieved an AUC of 0.954, with 95% sensitivity and 85% specificity. When used to assist junior radiologists, it boosted their AUC by nearly 11% and raised accuracy, sensitivity, and specificity by 10%‑13% each. Importantly, the AI also made doctors’ assessments more consistent, moving agreement from “fair” to “moderate.” Heat‑map visualizations show that DeepFAN bases its judgments on clinically recognized cues such as nodule size, shape, and density, making its decisions transparent. Approved by China’s medical regulator, DeepFAN is now ready for real‑world use, promising faster, more reliable lung‑cancer screening for patients everywhere.
Read more