In a paper published in Mycology, a research team from Southwest University identified multiple genes in the lysine biosynthesis pathway of the pathogenic fungus Cryptococcus neoformans. These genes were demonstrated to play significant roles in the growth, stress response, and virulence expression of C. neoformans. This provides important insights and ideas for deepening the understanding of the relationship between fungal metabolism and pathogenicity, as well as for developing novel antifungal drug targets.

Klebsiella pneumoniae can cause severe infections, and its protective capsule is a key marker for tracking virulence, drug-resistance risks, and transmission. Yet capsule typing remains slow or limited when it relies on antisera, PCR panels, or genome sequencing. Researchers developed a modular bioluminescent reporter phage platform that converts capsule recognition into a rapid light signal. By inserting NanoLuc into a K2-specific phage and swapping receptor-binding proteins, the team generated reporter phages for clinically important K1, K2, K47, K57, and K64 capsular types. The system rapidly detected target strains in culture, synthetic urine, and polymicrobial urine-like samples, while directed evolution further improved phage adsorption and detection sensitivity. In synthetic urine, the assay produced signals at clinically relevant bacterial loads, and it retained specificity in mixtures containing other common uropathogens. The platform could support faster surveillance and help guide more precise antimicrobial or phage-based interventions.

Autonomous AI robots require enhanced visual intelligence to accurately assess their local environment and identify objects before deciding which action to take and acting on that decision. Recently, a group of researchers outlined the current autonomous robot state of the art from the perspective of 2D and 3D visual intelligence and summarized current obstacles and future directions in the field.

A new study published in the Chinese Journal of Aeronautics （https://doi.org/10.1016/j.cja.2025.103943） tries to break this deadlock. Researchers from Northwestern Polytechnical University and Tsinghua University have developed a novel modeling framework called differential equation model long-short term memory neural network (DEM-LSTM). This hybrid approach integrates the theoretical rigor of differential equations with the nonlinear mapping capabilities of LSTM neural network, offering a solution that is both accurate and physically interpretable.