We describe a methodology for rapid experimentation in statistical machine translation which we use to add a large number of features to a baseline system exploiting features from a wide range of levels of syntactic representation. Feature values were combined in a log-linear model to select the highest scoring candidate translation from an n-best list. Feature weights were optimized directly against the BLEU evaluation metric on held-out data. We present results for a small selection of features at each level of syntactic representation.

Animals can store information about experiences by activating specific neuronal populations, and subsequent reactivation of these neural ensembles can lead to recall of salient experiences. In the hippocampus, granule cells of the dentate gyrus participate in such memory engrams; however, whether there is an underlying logic to granule cell participation has not been examined. Here, we find that a range of novel experiences preferentially activates granule cells of the suprapyramidal blade relative to the infrapyramidal blade. Motivated by this, we identify a suprapyramidal-blade-enriched population of granule cells with distinct spatial, morphological, physiological, and developmental properties. Via transcriptomics, we map these traits onto a sparse and discrete granule cell subtype that is recruited at a 10-fold greater frequency than expected by subtype prevalence, constituting the majority of all recruited granule cells. Thus, in behaviors known to involve hippocampal-dependent memory formation, a rare and spatially localized subtype dominates overall granule cell recruitment.

UNLABELLED: Midbrain dopamine (DA) neurons are thought to be a critical node in the circuitry that mediates reward learning. DA neurons receive diverse inputs from regions distributed throughout the neuraxis from frontal neocortex to the mesencephalon. While a great deal is known about changes in the activity of individual DA neurons during learning, much less is known about the functional changes in the microcircuits in which DA neurons are embedded. Here we used local field potentials recorded from the midbrain of behaving mice to show that the midbrain evoked potential (mEP) faithfully reflects the temporal and spatial structure of the phasic response of midbrain neuron populations during conditioning. By comparing the mEP to simultaneously recorded single units, we identified specific components of the mEP that corresponded to phasic DA and non-DA responses to salient stimuli. The DA component of the mEP emerged with the acquisition of a conditioned stimulus, was extinguished following changes in reinforcement contingency, and could be inhibited by pharmacological manipulations that attenuate the phasic responses of DA neurons. In contrast to single-unit recordings, the mEP permitted relatively dense sampling of the midbrain circuit during conditioning and thus could be used to reveal the spatiotemporal structure of multiple intermingled midbrain circuits. Finally, the mEP response was stable for months and thus provides a new approach to study long-term changes in the organization of ventral midbrain microcircuits during learning.

SIGNIFICANCE STATEMENT: Neurons that synthesize and release the neurotransmitter dopamine play a critical role in voluntary reward-seeking behavior. Much of our insight into the function of dopamine neurons comes from recordings of individual cells in behaving animals; however, it is notoriously difficult to record from dopamine neurons due to their sparsity and depth, as well as the presence of intermingled non-dopaminergic neurons. Here we show that much of the information that can be learned from recordings of individual dopamine and non-dopamine neurons is also revealed by changes in specific components of the local field potential. This technique provides an accessible measurement that could prove critical to our burgeoning understanding of the molecular, functional, and anatomical diversity of neuron populations in the midbrain.

TFIID-a complex of TATA-binding protein (TBP) and TBP-associated factors (TAFs)-is a central component of the Pol II promoter recognition apparatus. Recent studies have revealed significant downregulation of TFIID subunits in terminally differentiated myocytes, hepatocytes and adipocytes. Here, we report that TBP protein levels are tightly regulated by the ubiquitin-proteasome system. Using an in vitro ubiquitination assay coupled with biochemical fractionation, we identified Huwe1 as an E3 ligase targeting TBP for K48-linked ubiquitination and proteasome-mediated degradation. Upregulation of Huwe1 expression during myogenesis induces TBP degradation and myotube differentiation. We found that Huwe1 activity on TBP is antagonized by the deubiquitinase USP10, which protects TBP from degradation. Thus, modulating the levels of both Huwe1 and USP10 appears to fine-tune the requisite degradation of TBP during myogenesis. Together, our study unmasks a previously unknown interplay between an E3 ligase and a deubiquitinating enzyme regulating TBP levels during cellular differentiation.

We discovered a bimodal behavior in the genetically tractable organism Drosophila melanogaster that allowed us to directly probe the neural mechanisms of an action selection process. When confronted by a predator-mimicking looming stimulus, a fly responds with either a long-duration escape behavior sequence that initiates stable flight or a distinct, short-duration sequence that sacrifices flight stability for speed. Intracellular recording of the descending giant fiber (GF) interneuron during head-fixed escape revealed that GF spike timing relative to parallel circuits for escape actions determined which of the two behavioral responses was elicited. The process was well described by a simple model in which the GF circuit has a higher activation threshold than the parallel circuits, but can override ongoing behavior to force a short takeoff. Our findings suggest a neural mechanism for action selection in which relative activation timing of parallel circuits creates the appropriate motor output.

Interactions between proteins play an essential role in metabolic and signaling pathways, cellular processes and organismal systems. We report the development of splitFAST, a fluorescence complementation system for the visualization of transient protein-protein interactions in living cells. Engineered from the fluorogenic reporter FAST (Fluorescence-Activating and absorption-Shifting Tag), which specifically and reversibly binds fluorogenic hydroxybenzylidene rhodanine (HBR) analogs, splitFAST displays rapid and reversible complementation, allowing the real-time visualization of both the formation and the dissociation of a protein assembly.

Techniques that enable precise manipulations of subsets of neurons in the fly central nervous system have greatly facilitated our understanding of the neural basis of behavior. Split-GAL4 driver lines allow specific targeting of cell types in Drosophila melanogaster and other species. We describe here a collection of 3060 lines targeting a range of cell types in the adult Drosophila central nervous system and 1373 lines characterized in third-instar larvae. These tools enable functional, transcriptomic, and proteomic studies based on precise anatomical targeting. NeuronBridge and other search tools relate light microscopy images of these split-GAL4 lines to connectomes reconstructed from electron microscopy images. The collections are the result of screening over 77,000 split hemidriver combinations. In addition to images and fly stocks for these well-characterized lines, we make available 300,000 new 3D images of other split-GAL4 lines.

Realistic computational models of single neurons require component ion channels that reproduce experimental findings. Here, a topology-mutating genetic algorithm that searches for the best state diagram and transition-rate parameters to model macroscopic ion-channel behavior is described. Important features of the algorithm include a topology-altering strategy, automatic satisfaction of equilibrium constraints (microscopic reversibility), and multiple-protocol fitting using sequential goal programming rather than explicit weighting. Application of this genetic algorithm to design a sodium-channel model exhibiting both fast and prolonged inactivation yields a six-state model that produces realistic activity-dependent attenuation of action-potential backpropagation in current-clamp simulations of a CA1 pyramidal neuron.

Many functional RNAs have an evolutionarily conserved secondary structure. Conservation of RNA base pairing induces pairwise covariations in sequence alignments. We developed a computational method, R-scape (RNA Structural Covariation Above Phylogenetic Expectation), that quantitatively tests whether covariation analysis supports the presence of a conserved RNA secondary structure. R-scape analysis finds no statistically significant support for proposed secondary structures of the long noncoding RNAs HOTAIR, SRA, and Xist.