Perspectives in 2026

Laboratory robotics is shifting from scripted automation towards autonomous systems that can perceive, decide and act robustly in real experimental environments. Here, the authors introduce the ADePT framework, comprising adaptability and learning, dexterity, perception and task complexity, to benchmark robotic capability, expose key bottlenecks and chart practical routes towards truly self-driving laboratories.

Conventional homogeneous strategies for the synthesis of deuterated alcohols — valuable synthetic targets due to their roles in pharmaceuticals, materials, and mechanistic studies — present limitations in terms of cost and catalyst recovery, with heterogeneous catalysis providing recyclability and enhanced scalability. In this Perspective, the authors discuss the evolution of heterogeneous catalytic systems into scalable and efficient platforms for deuterated alcohol synthesis, highlighting new directions for sustainable isotope incorporation.

Deep learning models are pivotal in drug design, yet they often fail to capture crucial molecular interactions for binding affinity and specificity. Here, the authors present an overview of the methods to integrate molecular representations and propose that future work should develop biochemical foundation models that jointly encode diverse molecular modalities.