Search for: All records

Abstract We describe the discovery of an unspecific peroxygenase (UPO) variant that catalyzes the remote‐site functionalization of halogenated and unsaturated hydrocarbons with high catalytic site‐specificity. UPOs are fungal heme‐thiolate biocatalysts with wide‐ranging oxidative activities, including C─H bond oxygenation, usually with limited regioselectivity. We describe here a wild‐type MroUPO, newly isolated in high yield from a previously uncharacterized strain ofMarasmius rotula. This variant, MroUPO‐TN, catalyzes the selective oxygenation of a range of haloalkanes, cyclic haloalkanes and cyclic olefins to generate useful remote‐site haloketones. The regioselectivity for eight‐membered rings reaches 99% with significant enantiomeric excess. Mechanistic studies performed with deuterated substrates and¹⁸O‐labeling experiments have revealed a synergy between intrinsic substrate properties and the highly aliphatic, heme active site. The observed selectivity offers routes to new and useful, bifunctional synthons and pharmacophores, thus providing practical ways to employ these natural and environmentally benign biocatalysts. 

Indoor location services often use Bluetooth low energy (BLE) devices for their low energy consumption and easy implementation. Applications like device monitoring, ranging, and asset tracking utilize the received signal strength (RSS) of the BLE signal to estimate the proximity of a device from the receiver. However, in multipath environments, RSS-based solutions may not provide an accurate estimation. In such environments, receivers with antenna arrays are used to calculate the difference in time of flight (ToF) and therefore calculate the direction of arrival (DoA) of the Bluetooth signal. Other techniques like triangulation have also been used, such as having multiple transmitters or receivers as a network of sensors. To find a lost item, devices like Tile© use an onboard beeper to notify users of their presence. In this paper, we present a system that uses a single-measurement device and describe the method of measurement to estimate the location of a BLE device using RSS. A BLE device is configured as an Eddystone beacon for periodic transmission of advertising packets with RSS information. We developed a smartphone application to read RSS information from the beacon, designed an algorithm to estimate the DoA, and used the phone’s internal sensors to evaluate the DoA with respect to true north. The proposed measurement method allows for asset tracking by iterative measurements that provide the direction of the beacon and take the user closer at every step. The receiver application is easily deployable on a smartphone, and the algorithm provides direction of the beacon within a 30° range, as suggested by the provided results. 

We study (differentially) private federated learning (FL) of language models. The language models in cross-device FL are relatively small, which can be trained with meaningful formal user-level differential privacy (DP) guarantees when massive parallelism in training is enabled by the participation of a moderate size of users. Recently, public data has been used to improve privacy-utility trade-offs for both large and small language models. In this work, we provide a systematic study of using large-scale public data and LLMs to help differentially private training of on-device FL models, and further improve the privacy-utility tradeoff by techniques of distillation. Moreover, we propose a novel distribution matching algorithm with theoretical grounding to sample public data close to private data distribution, which significantly improves the sample efficiency of (pre-) training on public data. The proposed method is efficient and effective for training private models by taking advantage of public data, especially for customized on-device architectures that do not have ready-touse pre-trained models. 

Self‐assembled materials with complex nanoscale and mesoscale architecture attract considerable attention in energy and sustainability technologies. Their high performance can be attributed to high surface area, quantum effects, and hierarchical organization. Delineation of these contributions is, however, difficult because complex materials display stochastic structural patterns combining both order and disorder, which is difficult to be consistently reproduced yet being important for materials' functionality. Their compositional variability make systematic studies even harder. Here, a model system of FeSe₂“hedgehog” particles (HPs) was selected  to gain insight into the mechanisms of charge storage n complex nanostructured materials common for batteries and supercapacitors. Specifically, HPs represent self‐assembled biomimetic nanomaterials with a medium level of complexity; they display an organizational pattern of spiky colloids with considerable disorder yet non‐random; this patternt is consistently reproduced from particle to particle. . It was found that HPs can accommodate ≈70× greater charge density than spheroidal nano‐ and microparticles. Besides expanded surface area, the enhanced charge storage capacity was enabled by improved hole transport and reversible atomic conformations of FeSe₂layers in the blade‐like spikes associated with the rotatory motion of the Se atoms around Fe center. The dispersibility of HPs also enables their easy integration into energy storage devices. HPs quadruple stored electrochemical energy and double the storage modulus of structural supercapacitors.