Search for: All records

A popular line of recent research incorporates ML advice in the design of online algorithms to improve their performance in typical instances. These papers treat the ML algorithm as a blackbox, and redesign online algorithms to take advantage of ML predictions. In this paper, we ask the complementary question: can we redesign ML algorithms to provide better predictions for online algorithms? We explore this question in the context of the classic rent-or-buy problem, and show that incorporating optimization benchmarks directly in ML loss functions leads to significantly better performance, while maintaining a worst-case adversarial result when the advice is completely wrong. We support this finding both through theoretical bounds and numerical simulations, and posit that “learning for optimization” is a fertile area for future research. 

Abstract Niche signals maintain stem cells in a prolonged quiescence or transiently activate them for proper regeneration 1 . Altering balanced niche signalling can lead to regenerative disorders. Melanocytic skin nevi in human often display excessive hair growth, suggesting hair stem cell hyperactivity. Here, using genetic mouse models of nevi 2,3 , we show that dermal clusters of senescent melanocytes drive epithelial hair stem cells to exit quiescence and change their transcriptome and composition, potently enhancing hair renewal. Nevus melanocytes activate a distinct secretome, enriched for signalling factors. Osteopontin, the leading nevus signalling factor, is both necessary and sufficient to induce hair growth. Injection of osteopontin or its genetic overexpression is sufficient to induce robust hair growth in mice, whereas germline and conditional deletions of either osteopontin or CD44, its cognate receptor on epithelial hair cells, rescue enhanced hair growth induced by dermal nevus melanocytes. Osteopontin is overexpressed in human hairy nevi, and it stimulates new growth of human hair follicles. Although broad accumulation of senescent cells, such as upon ageing or genotoxic stress, is detrimental for the regenerative capacity of tissue 4 , we show that signalling by senescent cell clusters can potently enhance the activity of adjacent intact stem cells and stimulate tissue renewal. This finding identifies senescent cells and their secretome as an attractive therapeutic target in regenerative disorders. 

The DNA-binding protein H-NS is a pleiotropic gene regulator in gram-negative bacteria. Through its capacity to sense temperature and other environmental factors, H-NS allows pathogens like Salmonella to adapt their gene expression to their presence inside or outside warm-blooded hosts. To investigate how this sensing mechanism may have evolved to fit different bacterial lifestyles, we compared H-NS orthologs from bacteria that infect humans, plants, and insects, and from bacteria that live on a deep-sea hypothermal vent. The combination of biophysical characterization, high-resolution proton-less nuclear magnetic resonance spectroscopy, and molecular simulations revealed, at an atomistic level, how the same general mechanism was adapted to specific habitats and lifestyles. In particular, we demonstrate how environment-sensing characteristics arise from specifically positioned intra- or intermolecular electrostatic interactions. Our integrative approach clarified the exact modus operandi for H-NS-mediated environmental sensing and suggested that this sensing mechanism resulted from the exaptation of an ancestral protein feature. 

The motion of cells orthogonal to the direction of main flow is of importance in natural and engineered systems. The lateral movement of red blood cells (RBCs) distal to sudden expansion is considered to influence the formation and progression of thrombosis in venous valves, aortic aneurysms, and blood-circulating devices and is also a determining parameter for cell separation applications in flow-focusing microfluidic devices. Although it is known that the unique geometry of venous valves alters the blood flow patterns and cell distribution in venous valve sinuses, the interactions between fluid flow and RBCs have not been elucidated. Here, using a dilute cell suspension in an in vitro microfluidic model of a venous valve, we quantified the spatial distribution of RBCs by microscopy and image analysis, and using micro-particle image velocimetry and 3D computational fluid dynamics simulations, we analyzed the complex flow patterns. The results show that the local hematocrit in the valve pockets is spatially heterogeneous and is significantly different from the feed hematocrit. Above a threshold shear rate, the inertial separation of streamlines and lift forces contribute to an uneven distribution of RBCs in the vortices, the entrapment of RBCs in the vortices, and non-monotonic wall shear stresses in the valve pockets. Our experimental and computational characterization provides insights into the complex interactions between fluid flow, RBC distribution, and wall shear rates in venous valve mimics, which is of relevance to understanding the pathophysiology of thrombosis and improving cell separation efficiency.

 

In both biological and engineered systems, functioning at peak power output for prolonged periods of time requires thermoregulation. Here, we report a soft hydrogel-based actuator that can maintain stable body temperatures via autonomic perspiration. Using multimaterial stereolithography, we three-dimensionally print finger-like fluidic elastomer actuators having a poly- N -isopropylacrylamide (PNIPAm) body capped with a microporous (~200 micrometers) polyacrylamide (PAAm) dorsal layer. The chemomechanical response of these hydrogel materials is such that, at low temperatures (<30°C), the pores are sufficiently closed to allow for pressurization and actuation, whereas at elevated temperatures (>30°C), the pores dilate to enable localized perspiration in the hydraulic actuator. Such sweating actuators exhibit a 600% enhancement in cooling rate (i.e., 39.1°C minute −1 ) over similar non-sweating devices. Combining multiple finger actuators into a single device yields soft robotic grippers capable of both mechanically and thermally manipulating various heated objects. The measured thermoregulatory performance of these sweating actuators (~107 watts kilogram −1 ) greatly exceeds the evaporative cooling capacity found in the best animal systems (~35 watts kilogram −1 ) at the cost of a temporary decrease in actuation efficiency.