DeepBlastoid: A deep learning model for automated and efficient evaluation of human blastoids.
Peer-Reviewed Publication
Updates every hour. Last Updated: 2-Apr-2026 04:16 ET (2-Apr-2026 08:16 GMT/UTC)
Researchers from King Abdullah University of Science and Technology (KAUST) have developed deepBlastoid, the first deep-learning platform specifically designed for the high-throughput, automated classification of human stem cell-derived embryo models (blastoids). By leveraging a ResNet-18 architecture and a novel Confidence Rate metric, the model achieves up to 97% accuracy and processes images 1,000 times faster than human experts. This tool facilitates large-scale drug screening and basic research into early human development by providing a standardized, objective evaluation framework.
Engineering multifunctional extracellular vesicles that can selectively and effectively kill lung cancer cells with minimal off‑target toxicity, offers a promising platform for next‑generation, precision‑guided cancer therapeutics.
The research team of Xumu Zhang and Genqiang Chen at Southern University of Science and Technology has achieved a new breakthrough in the synthesis and application of benzoxoxetine ligands. They efficiently and modularly constructed benzoxoxetine ligands through a tandem nucleophilic addition SNAr reaction. The related findings, titled "Redox-Free and Modular Access to Oxacyclic Phosphines Enabled by a Robust Ambiphilic Phosphine Reagent," were recently published as an open access Reserach Article in CCS Chemistry.
Precocial animals, the ones that move autonomously within hours after hatching or birth, have many biases they are born with that help them survive, finds a new Royal Society paper led by Queen Mary University of London. The new model proposed by the researchers suggest that naïve animals like newborn turtles and chicks are not blank slates but are supported by the presence of multiple biases that interact.