Evaluation of data imputation strategies in complex, deeply-phenotyped data sets: the case of the EU-AIMS Longitudinal European Autism Project.

An increasing number of large-scale multi-modal research initiatives has been conducted in the typically developing population, e.g. Dev. Cogn. Neur. 32:43-54, 2018; PLoS Med. 12(3):e1001779, 2015; Elam and Van Essen, Enc. Comp. Neur., 2013, as well as in psychiatric cohorts, e.g. Trans. Psych. 10(1):100, 2020; Mol. Psych. 19:659-667, 2014; Mol. Aut. 8:24, 2017; Eur. Child and Adol. Psych. 24(3):265-281, 2015. Missing data is a common problem in such datasets due to the difficulty of assessing multiple measures on a large number of participants. The consequences of missing data accumulate when researchers aim to integrate relationships across multiple measures. Here we aim to evaluate different imputation strategies to fill in missing values in clinical data from a large (total N = 764) and deeply phenotyped (i.e. range of clinical and cognitive instruments administered) sample of N = 453 autistic individuals and N = 311 control individuals recruited as part of the EU-AIMS Longitudinal European Autism Project (LEAP) consortium. In particular, we consider a total of 160 clinical measures divided in 15 overlapping subsets of participants. We use two simple but common univariate strategies-mean and median imputation-as well as a Round Robin regression approach involving four independent multivariate regression models including Bayesian Ridge regression, as well as several non-linear models: Decision Trees (Extra Trees., and Nearest Neighbours regression. We evaluate the models using the traditional mean square error towards removed available data, and also consider the Kullback-Leibler divergence between the observed and the imputed distributions. We show that all of the multivariate approaches tested provide a substantial improvement compared to typical univariate approaches. Further, our analyses reveal that across all 15 data-subsets tested, an Extra Trees regression approach provided the best global results. This not only allows the selection of a unique model to impute missing data for the LEAP project and delivers a fixed set of imputed clinical data to be used by researchers working with the LEAP dataset in the future, but provides more general guidelines for data imputation in large scale epidemiological studies.

Preference for biological motion is reduced in ASD: implications for clinical trials and the search for biomarkers.

BACKGROUND: The neurocognitive mechanisms underlying autism spectrum disorder (ASD) remain unclear. Progress has been largely hampered by small sample sizes, variable age ranges and resulting inconsistent findings. There is a pressing need for large definitive studies to delineate the nature and extent of key case/control differences to direct research towards fruitful areas for future investigation. Here we focus on perception of biological motion, a promising index of social brain function which may be altered in ASD. In a large sample ranging from childhood to adulthood, we assess whether biological motion preference differs in ASD compared to neurotypical participants (NT), how differences are modulated by age and sex and whether they are associated with dimensional variation in concurrent or later symptomatology. METHODS: Eye-tracking data were collected from 486 6-to-30-year-old autistic (N = 282) and non-autistic control (N = 204) participants whilst they viewed 28 trials pairing biological (BM) and control (non-biological, CTRL) motion. Preference for the biological motion stimulus was calculated as (1) proportion looking time difference (BM-CTRL) and (2) peak look duration difference (BM-CTRL). RESULTS: The ASD group showed a present but weaker preference for biological motion than the NT group. The nature of the control stimulus modulated preference for biological motion in both groups. Biological motion preference did not vary with age, gender, or concurrent or prospective social communicative skill within the ASD group, although a lack of clear preference for either stimulus was associated with higher social-communicative symptoms at baseline. LIMITATIONS: The paired visual preference we used may underestimate preference for a stimulus in younger and lower IQ individuals. Our ASD group had a lower average IQ by approximately seven points. 18% of our sample was not analysed for various technical and behavioural reasons. CONCLUSIONS: Biological motion preference elicits small-to-medium-sized case-control effects, but individual differences do not strongly relate to core social autism associated symptomatology. We interpret this as an autistic difference (as opposed to a deficit) likely manifest in social brain regions. The extent to which this is an innate difference present from birth and central to the autistic phenotype, or the consequence of a life lived with ASD, is unclear.

Effects of BDNF Val66Met genotype and schizophrenia familial risk on a neural functional network for cognitive control in humans.

Cognitive control represents an essential neuropsychological characteristic that allows for the rapid adaption of a changing environment by constant re-allocation of cognitive resources. This finely tuned mechanism is impaired in psychiatric disorders such as schizophrenia and contributes to cognitive deficits. Neuroimaging has highlighted the contribution of the anterior cingulate cortex (ACC) and prefrontal regions (PFC) on cognitive control and demonstrated the impact of genetic variation, as well as genetic liability for schizophrenia. In this study, we aimed to examine the influence of the functional single-nucleotide polymorphism (SNP) rs6265 of a plasticity-related neurotrophic factor gene, BDNF (Val66Met), on cognitive control. Strong evidence implicates BDNF Val66Met in neural plasticity in humans. Furthermore, several studies suggest that although the variant is not convincingly associated with schizophrenia risk, it seems to be a modifier of the clinical presentation and course of the disease. In order to clarify the underlying mechanisms using functional magnetic resonance imaging (fMRI), we studied the effects of this SNP on ACC and PFC activation, and the connectivity between these regions in a discovery sample of 85 healthy individuals and sought to replicate this effect in an independent sample of 253 individuals. Additionally, we tested the identified imaging phenotype in relation to schizophrenia familial risk in a sample of 58 unaffected first-degree relatives of schizophrenia patients. We found a significant increase in interregional connectivity between ACC and PFC in the risk-associated BDNF 66Met allele carriers. Furthermore, we replicated this effect in an independent sample and demonstrated its independence of structural confounds, as well as task specificity. A similar coupling increase was detectable in individuals with increased familial risk for schizophrenia. Our results show that a key neural circuit for cognitive control is influenced by a plasticity-related genetic variant, which may render this circuit particular susceptible to genetic and environmental risk factors for schizophrenia.

The scent of salience--is there olfactory-trigeminal conditioning in humans?

Pavlovian fear conditioning has been thoroughly studied in the visual, auditory and somatosensory domain, but evidence is scarce with regard to the chemosensory modality. Under the assumption that Pavlovian conditioning relies on the supra-modal mechanism of salience attribution, the present study was set out to attest the existence of chemosensory aversive conditioning in humans as a specific instance of salience attribution. fMRI was performed in 29 healthy subjects during a differential aversive conditioning paradigm. Two odors (rose, vanillin) served as conditioned stimuli (CS), one of which (CS+) was intermittently coupled with intranasally administered CO2. On the neural level, a robust differential response to the CS+ emerged in frontal, temporal, occipito-parietal and subcortical brain regions, including the amygdala. These changes were paralleled by the development of a CS+-specific connectivity profile of the anterior midcingulate cortex (aMCC), which is a key structure for processing salience information in order to guide adaptive response selection. Increased coupling could be found between key nodes of the salience network (anterior insula, neo-cerebellum) and sensorimotor areas, representing putative input and output structures of the aMCC for exerting adaptive motor control. In contrast, behavioral and skin conductance responses did not show significant effects of conditioning, which has been attributed to contingency unawareness. These findings imply substantial similarities of conditioning involving chemosensory and other sensory modalities, and suggest that salience attribution and adaptive control represent a general, modality-independent principle underlying Pavlovian conditioning.