IMPORTANCE Large-scale neuroimaging studies have revealed group differences in cortical thickness across many psychiatric disorders. The underlying neurobiology behind these differences is not well understood.OBJECTIVE To determine neurobiologic correlates of group differences in cortical thickness between cases and controls in 6 disorders: attention-deficit/hyperactivity disorder (ADHD), autism spectrum disorder (ASD), bipolar disorder (BD), major depressive disorder (MDD), obsessive-compulsive disorder (OCD), and schizophrenia (SCZ). DESIGN, SETTING, AND PARTICIPANTSProfiles of group differences in cortical thickness between cases and controls were generated using T1-weighted magnetic resonance images. Similarity between interregional profiles of cell-specific gene expression and those in the group differences in cortical thickness were investigated in each disorder. Next, principal component analysis was used to reveal a shared profile of group difference in thickness across the disorders. Gene coexpression, clustering, and enrichment for genes associated with these disorders were conducted. Data analysis was conducted between June and December 2019. The analysis included 145 cohorts across 6 psychiatric disorders drawn from the ENIGMA consortium. The number of cases and controls in each of the 6 disorders were as follows:

No abstract

No abstract

Reduced fractional anisotropy (FA) associated with Major Depressive Disorder (MDD) overlaps anatomically with effects of childhood maltreatment experiences. The aim of this study was, therefore, to replicate the negative effect of childhood maltreatment on white matter fiber structure and to demonstrate, that alterations in MDD might be partially attributed to the higher occurrence of childhood maltreatment in MDD. Two independent cohorts (total N = 1 256) were investigated in a diffusion tensor imaging study: The Münster Neuroimaging Cohort (MNC, N = 186 MDD, N = 210 healthy controls, HC) as discovery sample and the Marburg-Münster Affective Disorders Cohort Study (MACS, N = 397 MDD, N = 462 HC) as replication sample. The effects of diagnosis (HC vs. MDD) and Childhood Trauma Questionnaire (CTQ) scores on FA were analyzed. A main effect of diagnosis with higher FA in MDD patients compared with HC was found in the MNC (p FWE = 0.021), but not in the MACS (p FWE = 0.52) before correcting for CTQ. A significant negative correlation of FA with CTQ emerged in both cohorts (MNC: p FWE = 0.006, MACS: p FWE = 0.012) in several tracts previously described in the literature. No CTQ × diagnosis interaction could be detected. Any main effect of diagnosis was abolished after correcting for CTQ (MNC: p FWE = 0.562, MACS: p FWE = 0.115). No differences in FA between MDD and HC could be found after correcting for childhood maltreatment, suggesting that previously reported group differences might be attributed partially to higher levels of maltreatment experiences in MDD rather than diagnosis itself. Furthermore, a well-established finding of reduced FA following childhood maltreatment experiences was replicated.

Numerous studies have implicated involvement of the hippocampus in the etiology and expression of schizophrenia-spectrum psychopathology, and reduced hippocampal volume is one of the most robust brain abnormalities reported in schizophrenia. Recent studies indicate that early stages of schizophrenia are specifically characterized by reductions in anterior hippocampal volume; however, studies have not examined hippocampal volume reductions in subclinical schizotypy. The present study was the first to examine the associations of positive, negative, and disorganized schizotypy dimensions with hippocampal subfield volumes in a large sample (n = 195) of nonclinically ascertained young adults, phenotyped using the Multidimensional Schizotypy Scale (MSS). Hippocampal subfields were analyzed from high-resolution 3 Tesla structural magnetic resonance imaging scans testing anatomical models, including anterior vs posterior regions and the cornu ammonis (CA), dentate gyrus (DG), and subiculum subfields separately for the left and right hemispheres. We demonstrate differential spatial effects across anterior vs posterior hippocampus segments across different dimensions of the schizotypy risk phenotype. The interaction of negative and disorganized schizotypy robustly predicted left hemisphere volumetric reductions for the anterior and total hippocampus, and anterior CA and DG, and the largest reductions were seen in participants high in negative and disorganized schizotypy. These findings extend previous early psychosis studies and together with behavioral studies of hippocampal-related memory impairments provide the basis for a dimensional neurobiological hippocampal model of schizophrenia risk. Subtle hippocampal subfield volume reductions may be prevalent prior to the onset of detectable prodromal clinical symptoms of psychosis and play a role in the etiology and development of such conditions.

We present a method for using previously-trained 'teacher' agents to kickstart the training of a new 'student' agent. To this end, we leverage ideas from policy distillation (Rusu et al., 2015;Parisotto et al., 2015) and population based training (Jaderberg et al., 2017). Our method places no constraints on the architecture of the teacher or student agents, and it regulates itself to allow the students to surpass their teachers in performance. We show that, on a challenging and computationally-intensive multi-task benchmark (Beattie et al., 2016), kickstarted training improves the data efficiency of new agents, making it significantly easier to iterate on their design. We also show that the same kickstarting pipeline can allow a single student agent to leverage multiple 'expert' teachers which specialise on individual tasks. In this setting kickstarting yields surprisingly large gains, with the kickstarted agent matching the performance of an agent trained from scratch in almost 10× fewer steps, and surpassing its final performance by 42%. Kickstarting is conceptually simple and can easily be incorporated into reinforcement learning experiments.

No abstract

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.