Biology subjects

Walter, M.

Publications and source records attributed to Walter, M..

6 recordsLinked to original sources

bioRxiv2018

Dynamic enhancer partitioning instructs activation of a growth regulator during exit from naïve pluripotency

During early mammalian development, the genome undergoes profound transitions in chromatin states, topological organization and recruitment of cis regulatory factors involved in transcriptional control. How these three layers of gene regulation interact is the matter of intense research. The Zdbf2 gene--which is involved in growth control--provides a valuable model to study this question: upon exit from naive pluripotency and prior to tissue differentiation, it undergoes a switch in usage from a distal to a proximal promoter, along with a switch in chromatin states, from polycomb to DNA methylation occupancy. Using an embryonic stem cell (ESC) culture system to mimic this period, we show here that four enhancers contribute to the Zdbf2 promoter switch, concomitantly with dynamic changes in chromosome architecture. Indeed, CTCF plays a key role in partitioning the locus in ESCs, to facilitate enhancer contact with the distal Zdbf2 promoter only. Partition relieving enhances proximal Zdbf2 promoter activity, as observed during differentiation or with mutants that lack local CTCF-based partition. Importantly, we show that CTCF-based regulation occurs independently of the polycomb and DNA methylation pathways. Our study reveals the importance of multi-layered regulatory frameworks to ensure proper spatio-temporal activation of developmentally important genes.

molecular biology↗

bioRxiv2018

Deep Learning for Quality Control of Subcortical Brain 3D Shape Models

We present several deep learning models for assessing the morphometric fidelity of deep grey matter region models extracted from brain MRI. We test three different convolutional neural net architectures (VGGNet, ResNet and Inception) over 2D maps of geometric features. Further, we present a novel geometry feature augmentation technique based on parametric spherical mapping. Finally, we present an approach for model decision visualization, allowing human raters to see the areas of subcortical shapes most likely to be deemed of failing quality by the machine. Our training data is comprised of 5200 subjects from the ENIGMA Schizophrenia MRI cohorts, and our test dataset contains 1500 subjects from the ENIGMA Major Depressive Disorder cohorts. Our final models reduce human rater time by 46-70%. ResNet outperforms VGGNet and Inception for all of our predictive tasks.

neuroscience↗

bioRxiv2018

Motion and physiological noise effects on amygdala real-time fMRI neurofeedback learning

Real-time fMRI neurofeedback allows to learn control over activity in a localized brain region. However, with fMRI, physiological factors such as the cardiac cycle and respiration interfere with the measurement of brain activation. In conventional fMRI studies this is usually mitigated by inclusion of motion parameters and/or physiological parameters as nuisance regressors at the analysis stage, allowing to correct for and filter out such confounders. In real-time fMRI, however, such an approach is not routinely feasible due to the necessity to process all signals during the runtime of an experiment. The absence of on-line correction can therefore compromise real-time fMRI study outcomes reporting volitional self-regulation capability as BOLD signal changes. This is especially true for BOLD signal changes in subcortical regions situated close to blood vessels or air cavities, such as the amygdala. We therefore aimed to establish the effects of motion, heart rate, heart rate variability, and respiratory volume on learning effects, which means here an increase in BOLD signal change over NF training, in an amygdala neurofeedback experiment. Specifically, we investigate motion parameters from two emotion regulation studies - performed at 3T and 7T scanners - and additionally acquired physiological variance for the latter one. Our results revealed differences in these parameters between groups and especially between regulation and resting periods within each participant. However, strictly considering these parameters as nuisance regressors in data analysis revealed that the learning of volitional self-regulation of the amygdala is not driven by motion and physiological changes. As validation of our real-time findings, we compare them to the gold standard of assessment of motion and physiology from the Human Connectome Project. Based on this, we recommend to carefully report neurofeedback study results including physiological nuisance regression. To our knowledge, this is the first study investigating the effects of motion and physiological noise correction on neurofeedback BOLD effects.

neuroscience↗

bioRxiv2017

Automated microuidic cell culture of stem cell derived dopaminergic neurons in Parkinson’s disease

Parkinsons disease is a slowly progressive neurodegenerative disease characterised by dysfunction and death of selectively vulnerable midbrain dopaminergic neurons leading mainly to motor dysfunction, but also other non-motor symptoms. The development of human in vitro cellular models with similar phenotypic characteristics to selectively vulnerable neurons is a major challenge in Parkinsons disease research. We constructed a fully automated cell culture platform optimised for long-term maintenance and monitoring of induced pluripotent stem cell derived neurons in three dimensional microfluidic cell culture devices. The system can be flexibly adapted to various experimental protocols and features time-lapse imaging microscopy for quality control and electrophysiology monitoring to assess neuronal activity. Using this system, we continuously monitored the differentiation of Parkinsons disease patient derived human neuroepithelial stem cells into midbrain specific dopaminergic neurons. Calcium imaging confirmed the electrophysiological activity of differentiated neurons and immunostaining confirmed the efficiency of the differentiation protocol. This system is the first example of a fully automated Organ-on-a-Chip culture and enables a versatile array of in vitro experiments for patient-specific disease modelling.

bioengineering↗

bioRxiv2017

Machine Learning for Large-Scale Quality Control of 3D Shape Models in Neuroimaging

As very large studies of complex neuroimaging phenotypes become more common, human quality assessment of MRI-derived data remains one of the last major bottlenecks. Few attempts have so far been made to address this issue with machine learning. In this work, we optimize predictive models of quality for meshes representing deep brain structure shapes. We use standard vertex-wise and global shape features computed homologously across 19 cohorts and over 7500 human-rated subjects, training kernelized Support Vector Machine and Gradient Boosted Decision Trees classifiers to detect meshes of failing quality. Our models generalize across datasets and diseases, reducing human workload by 30-70%, or equivalently hundreds of human rater hours for datasets of comparable size, with recall rates approaching inter-rater reliability.

neuroscience↗

bioRxiv2017

Preexisting antibodies can protect against congenital cytomegalovirus infection in monkeys

Human cytomegalovirus (HCMV) is the most common congenital infection and a known cause of microcephaly, sensorineural hearing loss, and cognitive impairment among newborns worldwide. Natural maternal HCMV immunity reduces the incidence of congenital infection, but does not prevent the disease altogether. We employed a nonhuman primate model of congenital CMV infection to investigate the ability of preexisting antibodies to protect against placental CMV transmission. Pregnant, CD4+ T cell-depleted, rhesus CMV (RhCMV)-seronegative rhesus monkeys were treated with either standardly-produced hyperimmune globulin (HIG) from RhCMV-seropositive macaques or dose-optimized, potently RhCMV-neutralizing HIG prior to intravenous challenge with an RhCMV swarm. HIG passive infusion provided complete protection against fetal loss in both groups, and the potently-neutralizing HIG additionally inhibited placental transmission of RhCMV. Our findings suggest that antibody alone at the time of primary infection can prevent congenital CMV and therefore could be a primary target of vaccines to eliminate this neonatal infection.

immunology↗