Computational noise in reward-guided learning drives behavioral variability in volatile environments, bioRxiv, 2018-10-11
AbstractWhen learning the value of actions in volatile environments, humans often make seemingly irrational decisions which fail to maximize expected value. We reasoned that these ‘non-greedy’ decisions, instead of reflecting information seeking during choice, may be caused by computational noise in the learning of action values. Here, using reinforcement learning (RL) models of behavior and multimodal neurophysiological data, we show that the majority of non-greedy decisions stems from this learning noise. The trial-to-trial variability of sequential learning steps and their impact on behavior could be predicted both by BOLD responses to obtained rewards in the dorsal anterior cingulate cortex (dACC) and by phasic pupillary dilation – suggestive of neuromodulatory fluctuations driven by the locus coeruleus-norepinephrine (LC-NE) system. Together, these findings indicate that most of behavioral variability, rather than reflecting human exploration, is due to the limited computational precision of reward-guided learning.
biorxiv neuroscience 100-200-users 2018Functional clustering of dendritic activity during decision-making, bioRxiv, 2018-10-11
SummaryThe active properties of dendrites support local nonlinear operations, but previous imaging and electrophysiological measurements have produced conflicting views regarding the prevalence of local nonlinearities in vivo. We imaged calcium signals in pyramidal cell dendrites in the motor cortex of mice performing a tactile decision task. A custom microscope allowed us to image the soma and up to 300 μm of contiguous dendrite at 15 Hz, while resolving individual spines. New analysis methods were used to estimate the frequency and spatial scales of activity in dendritic branches and spines. The majority of dendritic calcium transients were coincident with global events. However, task-associated calcium signals in dendrites and spines were compartmentalized by dendritic branching and clustered within branches over approximately 10 μm. Diverse behavior-related signals were intermingled and distributed throughout the dendritic arbor, potentially supporting a large computational repertoire and learning capacity in individual neurons.
biorxiv neuroscience 100-200-users 2018Minimal phenotyping yields GWAS hits of low specificity for major depression, bioRxiv, 2018-10-11
AbstractMinimal phenotyping refers to the reliance on self-reported responses to one or two questions for disease case identification. This strategy has been applied to genome-wide association studies (GWAS) of major depressive disorder (MDD). Here we report that the genotype derived heritability (h2SNP) of depression defined by minimal phenotyping (14%, SE = 0.8%) is lower than strictly defined MDD (26%, SE = 2.2%), and that it shares as much genetic liability with strictly defined MDD (0.81, SE = 0.03) as it does with neuroticism (0.84, SE = 0.05), a trait not defined by the cardinal symptoms of depression. While they both show similar shared genetic liability with the personality trait neuroticism, a greater proportion of the genome contribute to the minimal phenotyping definitions of depression (80.2%, SE = 0.6%) than to strictly defined MDD (65.8%, SE = 0.6%). We find that GWAS loci identified in minimal phenotyping definitions of depression are not specific to MDD they also predispose to other psychiatric conditions. Finally, genetic predictors based on minimal phenotyping definitions are not predictive of strictly defined MDD in independent cohorts. Our results reveal that genetic analysis of minimal phenotyping definitions of depression identifies non-specific genetic factors shared between MDD and other psychiatric conditions. Reliance on results from minimal phenotyping for MDD may thus bias views of the genetic architecture of MDD and impedes ability to identify pathways specific to MDD.
biorxiv genetics 100-200-users 2018Minimal phenotyping yields GWAS hits of reduced specificity for major depression, bioRxiv, 2018-10-11
AbstractMinimal phenotyping refers to the reliance on the use of a small number of self-report items for disease case identification. This strategy has been applied to genome-wide association studies (GWAS) of major depressive disorder (MDD). Here we report that the genotype derived heritability (h2SNP) of depression defined by minimal phenotyping (14%, SE = 0.8%) is lower than strictly defined MDD (26%, SE = 2.2%). This cannot be explained by differences in prevalence between definitions or including cases of lower liability to MDD in minimal phenotyping definitions of depression, but can be explained by misdiagnosis of those without depression or with related conditions as cases of depression. Depression defined by minimal phenotyping is as genetically correlated with strictly defined MDD (rG = 0.81, SE = 0.03) as it is with the personality trait neuroticism (rG = 0.84, SE = 0.05), a trait not defined by the cardinal symptoms of depression. While they both show similar shared genetic liability with neuroticism, a greater proportion of the genome contributes to the minimal phenotyping definitions of depression (80.2%, SE = 0.6%) than to strictly defined MDD (65.8%, SE = 0.6%). We find that GWAS loci identified in minimal phenotyping definitions of depression are not specific to MDD they also predispose to other psychiatric conditions. Finally, while highly predictive polygenic risk scores can be generated from minimal phenotyping definitions of MDD, the predictive power can be explained entirely by the sample size used to generate the polygenic risk score, rather than specificity for MDD. Our results reveal that genetic analysis of minimal phenotyping definitions of depression identifies non-specific genetic factors shared between MDD and other psychiatric conditions. Reliance on results from minimal phenotyping for MDD may thus bias views of the genetic architecture of MDD and may impede our ability to identify pathways specific to MDD.
biorxiv genetics 100-200-users 2018Selene a PyTorch-based deep learning library for biological sequence-level data, bioRxiv, 2018-10-10
AbstractTo enable the application of deep learning in biology, we present Selene (<jatsext-link xmlnsxlink=httpwww.w3.org1999xlink ext-link-type=uri xlinkhref=httpsselene.flatironinstitute.org>httpsselene.flatironinstitute.org<jatsext-link>), a PyTorch-based deep learning library for fast and easy development, training, and application of deep learning model architectures for any biological sequences. We demonstrate how Selene allows researchers to easily train a published architecture on new data, develop and evaluate a new architecture, and use a trained model to answer biological questions of interest.
biorxiv bioinformatics 100-200-users 2018Single-cell virus sequencing of influenza infections that trigger innate immunity, bioRxiv, 2018-10-07
SUMMARYThe outcome of viral infection is extremely heterogeneous, with infected cells only sometimes activating innate immunity. Here we develop a new approach to assess how the genetic variation inherent in viral populations contributes to this heterogeneity. We do this by determining both the transcriptome and full-length sequences of all viral genes in single influenza-infected cells. Most cells are infected by virions with defects such as amino-acid mutations, internal deletions, or failure to express a gene. We identify instances of each type of defect that increase the likelihood that a cell activates an innate-immune response. However, immune activation remains stochastic in cells infected by virions with these defects, and sometimes occurs even when a cell is infected by a virion that expresses unmutated copies of all genes. Our work shows that viral genetic variation substantially contributes to but does not fully explain the heterogeneity in single influenza-infected cells.
biorxiv microbiology 100-200-users 2018