User Training With Error Augmentation for sEMG-Based Gesture Classification.

We designed and tested a system for real-time control of a user interface by extracting surface electromyographic (sEMG) activity from eight electrodes in a wristband configuration. sEMG data were streamed into a machine-learning algorithm that classified hand gestures in real-time. After an initial model calibration, participants were presented with one of three types of feedback during a human-learning stage: veridical feedback, in which predicted probabilities from the gesture classification algorithm were displayed without alteration; modified feedback, in which we applied a hidden augmentation of error to these probabilities; and no feedback. User performance was then evaluated in a series of minigames, in which subjects were required to use eight gestures to manipulate their game avatar to complete a task. Experimental results indicated that relative to the baseline, the modified feedback condition led to significantly improved accuracy. Class separation also improved, though this trend was not significant. These findings suggest that real-time feedback in a gamified user interface with manipulation of feedback may enable intuitive, rapid, and accurate task acquisition for sEMG-based gesture recognition applications.

Probabilistic program inference in network-based epidemiological simulations.

Accurate epidemiological models require parameter estimates that account for mobility patterns and social network structure. We demonstrate the effectiveness of probabilistic programming for parameter inference in these models. We consider an agent-based simulation that represents mobility networks as degree-corrected stochastic block models, whose parameters we estimate from cell phone co-location data. We then use probabilistic program inference methods to approximate the distribution over disease transmission parameters conditioned on reported cases and deaths. Our experiments demonstrate that the resulting models improve the quality of fit in multiple geographies relative to baselines that do not model network topology.

AutoTransfer: Subject Transfer Learning with Censored Representations on Biosignals Data.

We investigate a regularization framework for subject transfer learning in which we train an encoder and classifier to minimize classification loss, subject to a penalty measuring independence between the latent representation and the subject label. We introduce three notions of independence and corresponding penalty terms using mutual information or divergence as a proxy for independence. For each penalty term, we provide several concrete estimation algorithms, using analytic methods as well as neural critic functions. We propose a hands-off strategy for applying this diverse family of regularization schemes to a new dataset, which we call "Auto Transfer". We evaluate the performance of these individual regularization strategies under our AutoTransfer framework on EEG, EMG, and ECoG datasets, showing that these approaches can improve subject transfer learning for challenging real-world datasets.

Target-Related Alpha Attenuation in a Brain-Computer Interface Rapid Serial Visual Presentation Calibration.

This study evaluated the feasibility of using occipitoparietal alpha activity to drive target/non-target classification in a brain-computer interface (BCI) for communication. EEG data were collected from 12 participants who completed BCI Rapid Serial Visual Presentation (RSVP) calibrations at two different presentation rates: 1 and 4 Hz. Attention-related changes in posterior alpha activity were compared to two event-related potentials (ERPs): N200 and P300. Machine learning approaches evaluated target/non-target classification accuracy using alpha activity. Results indicated significant alpha attenuation following target letters at both 1 and 4 Hz presentation rates, though this effect was significantly reduced in the 4 Hz condition. Target-related alpha attenuation was not correlated with coincident N200 or P300 target effects. Classification using posterior alpha activity was above chance and benefitted from individualized tuning procedures. These findings suggest that target-related posterior alpha attenuation is detectable in a BCI RSVP calibration and that this signal could be leveraged in machine learning algorithms used for RSVP or comparable attention-based BCI paradigms.

Geometric Analysis of Uncertainty Sampling for Dense Neural Network Layer.

For model adaptation of fully connected neural network layers, we provide an information geometric and sample behavioral active learning uncertainty sampling objective analysis. We identify conditions under which several uncertainty-based methods have the same performance and show that such conditions are more likely to appear in the early stages of learning. We define riskier samples for adaptation, and demonstrate that, as the set of labeled samples increases, margin-based sampling outperforms other uncertainty sampling methods by preferentially selecting these risky samples. We support our derivations and illustrations with experiments using Meta-Dataset, a benchmark for few-shot learning. We compare uncertainty-based active learning objectives using features produced by SimpleCNAPS (a state-of-the-art few-shot classifier) as input for a fully-connected adaptation layer. Our results indicate that margin-based uncertainty sampling achieves similar performance as other uncertainty based sampling methods with fewer labelled samples as discussed in the novel geometric analysis.

De novo ATP1A3 and compound heterozygous NLRP3 mutations in a child with autism spectrum disorder, episodic fatigue and somnolence, and muckle-wells syndrome.

Complex phenotypes may represent novel syndromes that are the composite interaction of several genetic and environmental factors. We describe an 9-year old male with high functioning autism spectrum disorder and Muckle-Wells syndrome who at age 5  years of age manifested perseverations that interfered with his functioning at home and at school. After age 6, he developed intermittent episodes of fatigue and somnolence lasting from hours to weeks that evolved over the course of months to more chronic hypersomnia. Whole exome sequencing showed three mutations in genes potentially involved in his clinical phenotype. The patient has a predicted pathogenic de novo heterozygous p.Ala681Thr mutation in the ATP1A3 gene (chr19:42480621C>T, GRCh37/hg19). Mutations in this gene are known to cause Alternating Hemiplegia of Childhood, Rapid Onset Dystonia Parkinsonism, and CAPOS syndrome, sometimes accompanied by autistic features. The patient also has compound heterozygosity for p.Arg490Lys/p.Val200Met mutations in the NLRP3 gene (chr1:247588214G>A and chr1:247587343G>A, respectively). NLRP3 mutations are associated in an autosomal dominant manner with clinically overlapping auto-inflammatory conditions including Muckle-Wells syndrome. The p.Arg490Lys is a known pathogenic mutation inherited from the patient's father. The p.Val200Met mutation, inherited from his mother, is a variant of unknown significance (VUS). Whether the de novoATP1A3mutation is responsible for or plays a role in the patient's episodes of fatigue and somnolence remains to be determined. The unprecedented combination of two NLRP3 mutations may be responsible for other aspects of his complex phenotype.

A novel de novo mutation in ATP1A3 and childhood-onset schizophrenia.

We describe a child with onset of command auditory hallucinations and behavioral regression at 6 yr of age in the context of longer standing selective mutism, aggression, and mild motor delays. His genetic evaluation included chromosomal microarray analysis and whole-exome sequencing. Sequencing revealed a previously unreported heterozygous de novo mutation c.385G>A in ATP1A3, predicted to result in a p.V129M amino acid change. This gene codes for a neuron-specific isoform of the catalytic α-subunit of the ATP-dependent transmembrane sodium-potassium pump. Heterozygous mutations in this gene have been reported as causing both sporadic and inherited forms of alternating hemiplegia of childhood and rapid-onset dystonia parkinsonism. We discuss the literature on phenotypes associated with known variants in ATP1A3, examine past functional studies of the role of ATP1A3 in neuronal function, and describe a novel clinical presentation associated with mutation of this gene.

Gene expression analysis in Fmr1KO mice identifies an immunological signature in brain tissue and mGluR5-related signaling in primary neuronal cultures.

BACKGROUND: Fragile X syndrome (FXS) is a neurodevelopmental disorder whose biochemical manifestations involve dysregulation of mGluR5-dependent pathways, which are widely modeled using cultured neurons. In vitro phenotypes in cultured neurons using standard morphological, functional, and chemical approaches have demonstrated considerable variability. Here, we study transcriptomes obtained in situ in the intact brain tissues of a murine model of FXS to see how they reflect the in vitro state. METHODS: We used genome-wide mRNA expression profiling as a robust characterization tool for studying differentially expressed pathways in fragile X mental retardation 1 (Fmr1) knockout (KO) and wild-type (WT) murine primary neuronal cultures and in embryonic hippocampal and cortical murine tissue. To study the developmental trajectory and to relate mouse model data to human data, we used an expression map of human development to plot murine differentially expressed genes in KO/WT cultures and brain. RESULTS: We found that transcriptomes from cell cultures showed a stronger signature of Fmr1KO than whole tissue transcriptomes. We observed an over-representation of immunological signaling pathways in embryonic Fmr1KO cortical and hippocampal tissues and over-represented mGluR5-downstream signaling pathways in Fmr1KO cortical and hippocampal primary cultures. Genes whose expression was up-regulated in Fmr1KO murine cultures tended to peak early in human development, whereas differentially expressed genes in embryonic cortical and hippocampal tissues clustered with genes expressed later in human development. CONCLUSIONS: The transcriptional profile in brain tissues primarily centered on immunological mechanisms, whereas the profiles from cell cultures showed defects in neuronal activity. We speculate that the isolation and culturing of neurons caused a shift in neurological transcriptome towards a "juvenile" or "de-differentiated" state. Moreover, cultured neurons lack the close coupling with glia that might be responsible for the immunological phenotype in the intact brain. Our results suggest that cultured cells may recapitulate an early phase of the disease, which is also less obscured with a consequent "immunological" phenotype and in vivo compensatory mechanisms observed in the embryonic brain. Together, these results suggest that the transcriptome of cultured primary neuronal cells, in comparison to whole brain tissue, more robustly demonstrated the difference between Fmr1KO and WT mice and might reveal a molecular phenotype, which is typically hidden by compensatory mechanisms present in vivo. Moreover, cultures might be useful for investigating the perturbed pathways in early human brain development and genes previously implicated in autism.

All-optical electrophysiology in mammalian neurons using engineered microbial rhodopsins.

All-optical electrophysiology-spatially resolved simultaneous optical perturbation and measurement of membrane voltage-would open new vistas in neuroscience research. We evolved two archaerhodopsin-based voltage indicators, QuasAr1 and QuasAr2, which show improved brightness and voltage sensitivity, have microsecond response times and produce no photocurrent. We engineered a channelrhodopsin actuator, CheRiff, which shows high light sensitivity and rapid kinetics and is spectrally orthogonal to the QuasArs. A coexpression vector, Optopatch, enabled cross-talk-free genetically targeted all-optical electrophysiology. In cultured rat neurons, we combined Optopatch with patterned optical excitation to probe back-propagating action potentials (APs) in dendritic spines, synaptic transmission, subcellular microsecond-timescale details of AP propagation, and simultaneous firing of many neurons in a network. Optopatch measurements revealed homeostatic tuning of intrinsic excitability in human stem cell-derived neurons. In rat brain slices, Optopatch induced and reported APs and subthreshold events with high signal-to-noise ratios. The Optopatch platform enables high-throughput, spatially resolved electrophysiology without the use of conventional electrodes.

iPSC-derived neurons as a higher-throughput readout for autism: promises and pitfalls.

The elucidation of disease etiologies and establishment of robust, scalable, high-throughput screening assays for autism spectrum disorders (ASDs) have been impeded by both inaccessibility of disease-relevant neuronal tissue and the genetic heterogeneity of the disorder. Neuronal cells derived from induced pluripotent stem cells (iPSCs) from autism patients may circumvent these obstacles and serve as relevant cell models. To date, derived cells are characterized and screened by assessing their neuronal phenotypes. These characterizations are often etiology-specific or lack reproducibility and stability. In this review, we present an overview of efforts to study iPSC-derived neurons as a model for autism, and we explore the plausibility of gene expression profiling as a reproducible and stable disease marker.

Tools, methods, and applications for optophysiology in neuroscience.

The advent of optogenetics and genetically encoded photosensors has provided neuroscience researchers with a wealth of new tools and methods for examining and manipulating neuronal function in vivo. There exists now a wide range of experimentally validated protein tools capable of modifying cellular function, including light-gated ion channels, recombinant light-gated G protein-coupled receptors, and even neurotransmitter receptors modified with tethered photo-switchable ligands. A large number of genetically encoded protein sensors have also been developed to optically track cellular activity in real time, including membrane-voltage-sensitive fluorophores and fluorescent calcium and pH indicators. The development of techniques for controlled expression of these proteins has also increased their utility by allowing the study of specific populations of cells. Additionally, recent advances in optics technology have enabled both activation and observation of target proteins with high spatiotemporal fidelity. In combination, these methods have great potential in the study of neural circuits and networks, behavior, animal models of disease, as well as in high-throughput ex vivo studies. This review collects some of these new tools and methods and surveys several current and future applications of the evolving field of optophysiology.