Fabricius Crowder (suedeplow32)

Protein domains without the usual distribution of amino acids, called low complexity (LC) domains, can be prone to self-assembly into amyloid-like fibrils. Self-assembly of LC domains that are nearly devoid of hydrophobic residues, such as the 214-residue LC domain of the RNA-binding protein FUS, is particularly intriguing from the biophysical perspective and is biomedically relevant due to its occurrence within neurons in amyotrophic lateral sclerosis, frontotemporal dementia, and other neurodegenerative diseases. We report a high-resolution molecular structural model for fibrils formed by the C-terminal half of the FUS LC domain (FUS-LC-C, residues 111-214), based on a density map with 2.62 Å resolution from cryo-electron microscopy (cryo-EM). In the FUS-LC-C fibril core, residues 112-150 adopt U-shaped conformations and form two subunits with in-register, parallel cross-β structures, arranged with quasi-21 symmetry. All-atom molecular dynamics simulations indicate that the FUS-LC-C fibril core is stabilized by a plethora of hydrogen bonds involving sidechains of Gln, Asn, Ser, and Tyr residues, both along and transverse to the fibril growth direction, including diverse sidechain-to-backbone, sidechain-to-sidechain, and sidechain-to-water interactions. Nuclear magnetic resonance measurements additionally show that portions of disordered residues 151-214 remain highly dynamic in FUS-LC-C fibrils and that fibrils formed by the N-terminal half of the FUS LC domain (FUS-LC-N, residues 2-108) have the same core structure as fibrils formed by the full-length LC domain. These results contribute to our understanding of the molecular structural basis for amyloid formation by FUS and by LC domains in general.Deep neural networks (DNNs) excel at visual recognition tasks and are increasingly used as a modeling framework for neural computations in the primate brain. Just like individual brains, each DNN has a unique connectivity and representational profile. Here, we investigate individual differences among DNN instances that arise from varying only the random initialization of the network weights. Using tools typically employed in systems neuroscience, we show that this minimal change in initial conditions prior to training leads to substantial differences in intermediate and higher-level network representations despite similar network-level classification performance. We locate the origins of the effects in an under-constrained alignment of category exemplars, rather than misaligned category centroids. These results call into question the common practice of using single networks to derive insights into neural information processing and rather suggest that computational neuroscientists working with DNNs may need to base their inferences on groups of multiple network instances.Human skin is a self-healing mechanosensory system that detects various mechanical contact forces efficiently through three-dimensional innervations. Here, we propose a biomimetic artificially innervated foam by embedding three-dimensional electrodes within a new low-modulus self-healing foam material. The foam material is synthesized from a one-step self-foaming process. By tuning the concentration of conductive metal particles in the foam at near-percolation, we demonstrate that it can operate as a piezo-impedance sensor in both piezoresistive and piezocapacitive sensing modes without the need for an encapsulation layer. The sensor is sensitive to an object's contact force directions as well as to human proximity. Moreover, the foam material self-heals autonomously with immediate function restoration despite mechanical damage. It further recovers from mechanical bifurcations with gentle heating (70 °C). We anticipate that this material will be useful as damage robust human-machine interfaces.The world is combating an ongoing COVID-19 pandemic with health-care systems, society and economies impacted in an unprecedented way. It is unclear how many people have con