Using deep autoencoders to identify abnormal brain structural patterns in neuropsychiatric disorders: A large-scale multi-sample study.
Ontology highlight
ABSTRACT:
SUBMITTER: Pinaya WHL
PROVIDER: S-EPMC6492107 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:PurposeTo visualize and quantify structural patterns of optic nerve edema encountered in papilledema during treatment.MethodsA novel bi-channel deep-learning variational autoencoder (biVAE) model was trained using 1498 optical coherence tomography (OCT) scans of 125 subjects over time from the Idiopathic Intracranial Hypertension Treatment Trial (IIHTT) and 791 OCT scans of 96 control subjects from the University of Iowa. An independent test dataset of 70 eyes from 70 papilledema subjects was used to evaluate the ability of the biVAE model to quantify and reconstruct the papilledema spatial patterns from input OCT scans using only two variables.ResultsThe montage color maps of the retinal nerve fiber layer (RNFL) and total retinal thickness (TRT) produced by the biVAE model provided an organized visualization of the variety of morphological patterns of optic disc edema (including differing patterns at similar thickness levels). Treatment effects of acetazolamide versus placebo in the IIHTT were also demonstrated in the latent space. In image reconstruction, the mean signed peripapillary retinal nerve fiber layer thickness (pRNFLT) difference ± SD was -0.12 ± 17.34 µm, the absolute pRNFLT difference was 13.68 ± 10.65 µm, and the RNFL structural similarity index reached 0.91 ± 0.05.ConclusionsA wide array of structural patterns of papilledema, integrating the magnitude of disc edema with underlying disc and retinal morphology, can be quantified by just two latent variables.Translational relevanceA biVAE model encodes structural patterns, as well as the correlation between channels, and may be applied to visualize individuals or populations with papilledema throughout treatment.
| S-EPMC10795546 | biostudies-literature
Project description:OBJECTIVE:Neurodevelopmental disorders presumably involve events that occur during brain development. The authors hypothesized that neuropsychiatric disorders considered to be developmental in etiology are associated with susceptibility genes that are relatively upregulated during fetal life (i.e., differentially expressed). METHOD:The authors investigated the presence of prenatal expression enrichment of susceptibility genes systematically, as composite gene sets associated with six neuropsychiatric disorders in the microarray-based "BrainCloud" dorsolateral prefrontal cortex transcriptome. RESULTS:Using a fetal/postnatal log2-fold change threshold of 0.5, genes associated with syndromic neurodevelopmental disorders (N=31 genes, p=3.37×10-3), intellectual disability (N=88 genes, p=5.53×10-3), and autism spectrum disorder (N=242 genes, p=3.45×10-4) were relatively enriched in prenatal transcript abundance, compared with the overall transcriptome. Genes associated with schizophrenia by genome-wide association studies were not preferentially fetally expressed (N=106 genes, p=0.46), nor were genes associated with schizophrenia by exome sequencing (N=212 genes, p=0.21), but specific genes within copy-number variant regions associated with schizophrenia were relatively enriched in prenatal transcript abundance, and genes associated with schizophrenia by meta-analysis were functionally enriched for some neurodevelopmental processes. In contrast, genes associated with neurodegenerative disorders were significantly underexpressed during fetal life (N=46 genes, p=1.67×10-3). CONCLUSIONS:The authors found evidence for relative prenatal enrichment of putative susceptibility genes for syndromic neurodevelopmental disorders, intellectual disability, and autism spectrum disorder. Future transcriptome-level association studies should evaluate regions other than the dorsolateral prefrontal cortex, at other time points, and incorporate further RNA sequencing analyses.
| S-EPMC4140692 | biostudies-literature
Project description:The genetic architecture of psychiatric disorders is characterized by a large number of small-effect variants1 located primarily in non-coding regions, suggesting that the underlying causal effects may influence disease risk by modulating gene expression2-4. We provide comprehensive analyses using transcriptome data from an unprecedented collection of tissues to gain pathophysiological insights into the role of the brain, neuroendocrine factors (adrenal gland) and gastrointestinal systems (colon) in psychiatric disorders. In each tissue, we perform PrediXcan analysis and identify trait-associated genes for schizophrenia (n associations = 499; n unique genes = 275), bipolar disorder (n associations = 17; n unique genes = 13), attention deficit hyperactivity disorder (n associations = 19; n unique genes = 12) and broad depression (n associations = 41; n unique genes = 31). Importantly, both PrediXcan and summary-data-based Mendelian randomization/heterogeneity in dependent instruments analyses suggest potentially causal genes in non-brain tissues, showing the utility of these tissues for mapping psychiatric disease genetic predisposition. Our analyses further highlight the importance of joint tissue approaches as 76% of the genes were detected only in difficult-to-acquire tissues.
| S-EPMC6590703 | biostudies-literature
Project description:Neuropsychiatric disease has polygenic determinants but is often precipitated by environmental pressures, including adverse perinatal events. However, the way in which genetic vulnerability and early-life adversity interact remains obscure. We hypothesised that the extreme environmental stress of prematurity would promote neuroanatomic abnormality in individuals genetically vulnerable to psychiatric disorders. In 194 unrelated infants (104 males, 90 females), born before 33 weeks of gestation (mean gestational age 29.7 weeks), we combined Magnetic Resonance Imaging with a polygenic risk score (PRS) for five psychiatric pathologies to test the prediction that: deep grey matter abnormalities frequently seen in preterm infants are associated with increased polygenic risk for psychiatric illness. The variance explained by the PRS in the relative volumes of four deep grey matter structures (caudate nucleus, thalamus, subthalamic nucleus and lentiform nucleus) was estimated using linear regression both for the full, mixed ancestral, cohort and a subsample of European infants. Psychiatric PRS was negatively associated with lentiform volume in the full cohort (??=?-0.24, p?=?8?×?10-4) and a European subsample (??=?-0.24, p?=?8?×?10-3). Genetic variants associated with neuropsychiatric disease increase vulnerability to abnormal lentiform development after perinatal stress and are associated with neuroanatomic changes in the perinatal period.
| S-EPMC6374514 | biostudies-literature
Project description:A heterogeneous disease such as cancer is activated through multiple pathways and different perturbations. Depending upon the activated pathway(s), the survival of the patients varies significantly and shows different efficacy to various drugs. Therefore, cancer subtype detection using genomics level data is a significant research problem. Subtype detection is often a complex problem, and in most cases, needs multi-omics data fusion to achieve accurate subtyping. Different data fusion and subtyping approaches have been proposed over the years, such as kernel-based fusion, matrix factorization, and deep learning autoencoders. In this paper, we compared the performance of different deep learning autoencoders for cancer subtype detection. We performed cancer subtype detection on four different cancer types from The Cancer Genome Atlas (TCGA) datasets using four autoencoder implementations. We also predicted the optimal number of subtypes in a cancer type using the silhouette score and found that the detected subtypes exhibit significant differences in survival profiles. Furthermore, we compared the effect of feature selection and similarity measures for subtype detection. For further evaluation, we used the Glioblastoma multiforme (GBM) dataset and identified the differentially expressed genes in each of the subtypes. The results obtained are consistent with other genomic studies and can be corroborated with the involved pathways and biological functions. Thus, it shows that the results from the autoencoders, obtained through the interaction of different datatypes of cancer, can be used for the prediction and characterization of patient subgroups and survival profiles.
| S-EPMC8122584 | biostudies-literature
Project description:Proper neuronal migration and laminar formation during corticogenesis is essential for normal brain function. Disruption of these developmental processes is thought to be involved in the pathogenesis of some neuropsychiatric conditions. Especially, Reelin, a glycoprotein mainly secreted by the Cajal-Retzius cells and a subpopulation of GABAergic interneurons, has been shown to play a critical role, both during embryonic and postnatal periods. Indeed, animal studies have clearly revealed that Reelin is an essential molecule for proper migration of cortical neurons and finally regulates the cell positioning in the cortex during embryonic and early postnatal stages; by contrast, Reelin signaling is closely involved in synaptic function in adulthood. In humans, genetic studies have shown that the reelin gene (RELN) is associated with a number of psychiatric diseases, including Schizophrenia (SZ), bipolar disorder (BP) and autistic spectrum disorder. Indeed, Reln haploinsufficiency has been shown to cause cognitive impairment in rodents, suggesting the expression level of the Reelin protein is closely related to the higher brain functions. However, the molecular abnormalities in the Reelin pathway involved in the pathogenesis of psychiatric disorders are not yet fully understood. In this article, we review the current progress in the understanding of the Reelin functions that could be related to the pathogenesis of psychiatric disorders. Furthermore, we discuss the basis for selecting Reelin and molecules in its downstream signaling pathway as potential therapeutic targets for psychiatric illnesses.
| S-EPMC5067484 | biostudies-literature
Project description:BackgroundDrug-target interaction (DTI) plays a vital role in drug discovery. Identifying drug-target interactions related to wet-lab experiments are costly, laborious, and time-consuming. Therefore, computational methods to predict drug-target interactions are an essential task in the drug discovery process. Meanwhile, computational methods can reduce search space by proposing potential drugs already validated on wet-lab experiments. Recently, deep learning-based methods in drug-target interaction prediction have gotten more attention. Traditionally, DTI prediction methods' performance heavily depends on additional information, such as protein sequence and molecular structure of the drug, as well as deep supervised learning.ResultsThis paper proposes a method based on deep unsupervised learning for drug-target interaction prediction called AutoDTI++. The proposed method includes three steps. The first step is to pre-process the interaction matrix. Since the interaction matrix is sparse, we solved the sparsity of the interaction matrix with drug fingerprints. Then, in the second step, the AutoDTI approach is introduced. In the third step, we post-preprocess the output of the AutoDTI model.ConclusionsExperimental results have shown that we were able to improve the prediction performance. To this end, the proposed method has been compared to other algorithms using the same reference datasets. The proposed method indicates that the experimental results of running five repetitions of tenfold cross-validation on golden standard datasets (Nuclear Receptors, GPCRs, Ion channels, and Enzymes) achieve good performance with high accuracy.
| S-EPMC8056558 | biostudies-literature
Project description:The accumulation of senescent cells behaves as a key driver of several age-related cancers and degenerative diseases of the general population. Whether cellular senescence also contributes to psychiatric diseases is still elusive. We report here the characterization of a zebrafish model of psychiatric disorder resulting from the invalidation of the ppp2r2c gene known to be involved in various psychiatric pathologies in human and encoding a neural specific regulatory subunit of the Protein Phosphatase 2A (PP2A). We found that the behavioral abnormalities of adult ppp2r2c-defficient fish were associated to neural cell senescence and that they can be reversed by invalidating the tp53 gene or by treating the fish with a drug that eliminates senescent cells. A molecule commonly used to treat ADHD patients, the methylphenidate, decreases the number of senescent neural cells both in adult PPP2R2C-defficient fish and in wild-type old fish. At the molecular level, methylphenidate attenuates the DNA damage response. We propose that some of the drugs commonly used in neuropsychiatry such as methylphenidate can prevent the appearance of senescent neural cells and that general strategies to clear senescent cells are promising therapies for common psychiatric disorders.
2022-12-22 | GSE151307 | GEO
Project description:The Scale of Body Connection (SBC) was created to address the need for a self-report measure to examine body awareness and bodily dissociation in mind-body research. Developed in the U.S.A., it has been translated into many languages and tested for validity of scale translation. The burgeoning of mind-body research and the widespread use of the SBC scale underscored the need for critical assessment of the instrument's measurement properties. Thus, a broader evaluation of the SBC was designed using large samples from eight international, cross-sectional studies drawn from community (i.e., non-clinical) populations. Specifically, we assessed scale distribution properties and internal consistency reliabity, and using confirmatory factory analysis we evaluated scale contruct validity and compared male/female measurement models. The results indicated acceptable reliability for both the body awareness and bodily dissociation scales, and a good fit between the proposed theoretic model and the data, providing evidence of construct validity across all samples. Mean differences in body awareness were observed for males vs. females in most samples, with females generally showing higher body awareness compared to males. Multi-group structural equation modeling demonstrated a stable latent factor structure and factor loadings, indicating equivalent measurement models for males and females. In summary, this multi-sample study demonstrated SBC construct validity that supports its use in clinical research as a brief, readily translated, easy to administer measure of body awareness and bodily dissociation.
| S-EPMC5640211 | biostudies-literature
Project description:Sample size estimation is a crucial step in experimental design but is understudied in the context of deep learning. Currently, estimating the quantity of labeled data needed to train a classifier to a desired performance, is largely based on prior experience with similar models and problems or on untested heuristics. In many supervised machine learning applications, data labeling can be expensive and time-consuming and would benefit from a more rigorous means of estimating labeling requirements. Here, we study the problem of estimating the minimum sample size of labeled training data necessary for training computer vision models as an exemplar for other deep learning problems. We consider the problem of identifying the minimal number of labeled data points to achieve a generalizable representation of the data, a minimum converging sample (MCS). We use autoencoder loss to estimate the MCS for fully connected neural network classifiers. At sample sizes smaller than the MCS estimate, fully connected networks fail to distinguish classes, and at sample sizes above the MCS estimate, generalizability strongly correlates with the loss function of the autoencoder. We provide an easily accessible, code-free, and dataset-agnostic tool to estimate sample sizes for fully connected networks. Taken together, our findings suggest that MCS and convergence estimation are promising methods to guide sample size estimates for data collection and labeling prior to training deep learning models in computer vision.
| S-EPMC9747810 | biostudies-literature