Biology subjects

James, A.

Publications and source records attributed to James, A..

36 records · Page 2Linked to original sources

bioRxiv2023

Genetic tools for the study of the mangrove killifish, Kryptolebias marmoratus, an emerging vertebrate model for phenotypic plasticity

Kryptolebias marmoratus (Kmar), a teleost fish of the order Cyprinodontiformes, has a suite of unique phenotypes and behaviors not observed in other fishes. Many of these phenotypes are discrete and highly plastic -varying over time within an individual, and in some cases reversible. Kmar and its interfertile sister species, K. hermaphroditus, are the only known self-fertile vertebrates. This unusual sexual mode has the potential to provide unique insights into the regulation of vertebrate sexual development, and also lends itself to genetics. Kmar is easily adapted to the lab and requires little maintenance. However, its internal fertilization and small clutch size limits its experimental use. To support Kmar as a genetic model, we compared alternative husbandry techniques to maximize recovery of early cleavage-stage embryos. We find that frequent egg collection enhances yield, and that protease treatment promotes the greatest hatching successes from diapause. We completed a forward mutagenesis screen and recovered several mutant lines that serve as important tools for genetics in this model. Several will serve as useful viable recessive markers for marking crosses. Importantly, the mutant kissylips lays embryos at twice the rate of wild-type. Combining frequent egg collection with the kissylips mutant background allows for a substantial enhancement of early embryo yield. These improvements were sufficient to allow experimental analysis of early development and the successful mono- and bi-allelic targeted knockout of an endogenous tyrosinase gene with CRISPR/Cas9 nucleases. Collectively, these tools will facilitate modern developmental genetics in this fascinating fish, leading to future insights into the regulation of plasticity.

developmental biology↗

bioRxiv2022

Genomic and ecologic characteristics of the airway microbial-mucosal complex

SO_SCPLOWUMMARYC_SCPLOWO_SCPCAP C_SCPCAPO_SCPLOWPARAGRAPHC_SCPLOWLung diseases due to infection and dysbiosis affect hundreds of millions of people world-wide1-4. Microbial communities at the airway mucosal barrier are conserved and highly ordered5, reflecting symbiosis and co-evolution with human host factors6. Freed of selection to digest nutrients for the host, the airway microbiome underpins cognate management of mucosal immunity and pathogen resistance. We show here the results of the first systematic culture and whole-genome sequencing of the principal airway bacterial species, identifying abundant novel organisms within the genera Streptococcus, Pauljensenia, Neisseria and Gemella. Bacterial genomes were enriched for genes encoding antimicrobial synthesis, adhesion and biofilm formation, immune modulation, iron utilisation, nitrous oxide (NO) metabolism and sphingolipid signalling. RNA-targeting CRISPR elements in some taxa suggest the potential to prevent or treat specific viral infections. Homologues of human RO60 present in Neisseria spp. provide a possible respiratory primer for autoimmunity in systemic lupus erythematosus (SLE) and Sjogren syndrome. We interpret the structure and biogeography of airway microbial communities from clinical surveys in the context of whole-genome content, identifying features of airway dysbiosis that may presage breakdown of homeostasis during acute attacks of asthma and chronic obstructive pulmonary disease (COPD). We match the gene content of isolates to human transcripts and metabolites expressed late in airway epithelial differentiation, identifying pathways that can sustain host interactions with the microbiota. Our results provide a systematic basis for decrypting interactions between commensals, pathogens, and mucosal immunity in lung diseases of global significance.

microbiology↗

bioRxiv2022

CNTNAP4 signaling regulates osteosarcoma disease progression

Improved treatment strategies for sarcoma rely on clarification of the molecular mediators of disease progression. Recently, we reported that the secreted glycoprotein NELL-1 modulates osteosarcoma (OS) disease progression in part via altering the sarcomatous extracellular matrix (ECM) and cell-ECM interactions. Of known NELL-1 interactor proteins, Contactin-associated protein-like 4 (Cntnap4) encodes a member of the neurexin superfamily of transmembrane molecules best known for its presynaptic functions in the central nervous system. Here, CRISPR/Cas9 gene deletion of CNTNAP4 reduced OS tumor growth, sarcoma-associated angiogenesis, and pulmonary metastases. CNTNAP4 knockout (KO) in OS tumor cells largely phenocopied the effects of NELL-1 KO, including reductions in sarcoma cell attachment, migration, and invasion. Further, CNTNAP4 KO cells were found to be unresponsive to the effects of NELL-1 treatment. Transcriptomic analysis combined with protein phospho-array demonstrated notable reductions in the MAPK/ERK signaling cascade with CNTNAP4 deletion, and the ERK1/2 agonist isoproterenol restored cell functions among CNTNAP4 KO tumor cells. Finally, human primary cells and tissues in combination with sequencing datasets confirmed the significance of CNTNAP4 signaling in human sarcomas. In summary, our findings demonstrate the biological importance of NELL-1/CNTNAP4 signaling axis in disease progression of human sarcomas and suggest that targeting the NELL-1/CNTNAP4 signaling pathway represents a strategy with potential therapeutic benefit in sarcoma patients.

cancer biology↗

bioRxiv2021

Neuron-like function of the nephron central command

Interoceptive neurons that sense and regulate our internal milieu have been identified in several organs except in the kidney cortex despite its major importance in maintaining body homeostasis. Here we report that the chief kidney cell type of the macula densa (MD) forms coordinated neural networks in each nephron that resemble peripheral ganglia. A combined in vivo single-cell 4D physiology (sc4DP) and scRNA sequencing approach identified the MD mechanisms of neuronal differentiation, heterogeneity (pacemaker MD cells), sensing of the local and systemic environment via multi-organ crosstalk, and regulation of organ functions by acting as the nephron central command. Consistent with their neuron-like nature, MD cells express the molecular fingerprint of neurodegeneration. Here we put forth the single-cell MD model and concept of local neural networks that control organ and body functions via interoception in normal physiological state and use an integrated mechanism of neurodegeneration in disease.

physiology↗

bioRxiv2021

Macrophage proliferation machinery leads to PDAC progression, but susceptibility to innate immunotherapy.

Tumor-associated macrophages (TAMs) are involved in many aspects of cancer progression and correlate with poor clinical outcomes in many cancer types, including pancreatic ductal adenocarcinomas (PDACs). Previous studies have shown that TAMs can populate PDAC tumors not only by monocyte recruitment but also by local proliferation. However, the impact local proliferation might have on macrophage phenotype and cancer progression is unknown. Here, we utilized genetically engineered cancer models, single-cell RNA-sequencing data, and in vitro systems to show that proliferation of TAMs was driven by colony stimulating factor-1 (CSF1) produced by cancer-associated fibroblasts. CSF1 induced high levels of p21 in macrophages, which regulated both TAM proliferation and phenotype. TAMs in human and mouse PDACs with high levels of p21 had more inflammatory and immunosuppressive phenotypes. The p21 expression in TAMs was induced by both stromal interaction and/or chemotherapy treatment. Finally, by modeling p21 expression levels in TAMs, we found that p21-driven macrophage immunosuppression in vivo drove tumor progression. Serendipitously, the same p21-driven pathways that drive tumor progression, also drive response to CD40 agonist. These data suggest that stromal or therapy-induced regulation of cell cycle machinery can regulate both macrophage-mediated immune suppression and susceptibility to innate immunotherapy. SummaryTAMs are indicative of poor clinical outcomes and in PDAC their number is sustained in part by local proliferation. This study shows that stromal desmoplasia drives local proliferation of TAMs, and induces their immunosuppressive ability through altering cell cycle machinery, including p21 expression. Serendipitously, these changes in p21 in TAMs also potentially render tumors more sensitive to CD40 agonist therapy.

cancer biology↗

bioRxiv2021

Obtaining Spatially Resolved Tumor Purity Maps Using Deep Multiple Instance Learning In A Pan-cancer Study

Tumor purity is the proportion of cancer cells in the tumor tissue. An accurate tumor purity estimation is crucial for accurate pathologic evaluation and for sample selection to minimize normal cell contamination in high throughput genomic analysis. We developed a novel deep multiple instance learning model predicting tumor purity from H&E stained digital histopathology slides. Our model successfully predicted tumor purity from slides of fresh-frozen sections in eight different TCGA cohorts and formalin-fixed paraffin-embedded sections in a local Singapore cohort. The predictions were highly consistent with genomic tumor purity values, which were inferred from genomic data and accepted as the golden standard. Besides, we obtained spatially resolved tumor purity maps and showed that tumor purity varies spatially within a sample. Our analyses on tumor purity maps also suggested that pathologists might have chosen high tumor content regions inside the slides during tumor purity estimation in the TCGA cohorts, which resulted in higher values than genomic tumor purity values. In short, our model can be utilized for high throughput sample selection for genomic analysis, which will help reduce pathologists workload and decrease inter-observer variability. Moreover, spatial tumor purity maps can help better understand the tumor microenvironment as a key determinant in tumor formation and therapeutic response.

cancer biology↗

bioRxiv2021

Genetic determinism of phage-bacteria coevolution in natural populations

Coevolution between bacteriophage (or phage) and their bacterial host is thought to be key for the coexistence of these antagonists. Recent studies have revealed the major role of mobile genetic elements in the emergence of phage resistant hosts but how phage escape these defenses in the wild remained to be explored. Here we show a striking parallel in phage evolving counter defenses to host defenses in natural population. We established a large collection of phages and their bacterial hosts and we explored the genetic structure of their interaction. We find that clearly delineated genomic clusters of phage are specific for distinct clades within a bacterial species, Vibrio crassostreae, yet while all phages can adsorb, only a subset of hosts are killed due to intracellular defense mechanisms. Host genomes contain multiple mobile defense genes and susceptibility to phage is negatively correlated with genome size. Phages also display extensive gene content variation, but their genome size remains conserved. We show that this gene content variation in hosts and phage is due to rapid turnover of genes involved in defense and escape, and that by exchanging anti-defense genes, phages irreversibly switch host. This could be indicative of co-evolution following the matching-allele-model of specificity and the spatial and temporal variability of phage infectivity further suggests that negative-frequency dependent selection drives phage-vibrio coevolutionary dynamics. We propose a "pan-escape system" that can be shared among phages by homologous recombination within a population that infects a bacterial host.

microbiology↗

bioRxiv2021

The multidimensional nutritional niche of fungus-cultivar provisioning in free-ranging colonies of a neotropical leafcutter ant

The foraging trails of Atta leafcutter colonies are among the most iconic scenes in Neotropical ecosystems, with thousands of ants carrying freshly cut plant fragments back to their nests where they are used to provision a fungal food crop. We tested a hypothesis that the fungal cultivars multidimensional requirements for macronutrients (protein and carbohydrates) and minerals (Al, Ca, Cu, Fe, K, Mg, Mn, Na, P and Zn) govern the foraging breadth of Atta colombica leafcutter ants in a Panamanian rainforest. Analyses of freshly cut plant fragments carried by leafcutter foragers showed that the combination of fruits, flowers, and leaves provide for a broad realized nutritional niche that can maximize cultivars performance. And, while the leaves that comprised the most harvested resource also delivered an intake target containing protein in excess of the amounts that can maximize cultivar growth, in vitro experiments showed that the minerals P, Al, and Fe can enhance the cultivars tolerance to protein-biased substrates, and potentially expand the ants foraging niche. Yet, the cultivar also exhibits narrow margins between mineral limitation and toxicity that may render plant fragments with seemingly optimal blends of macronutrients unsuitable for provisioning. Our approach highlights that optimal foraging is inherently multidimensional and links the foraging behavior of a generalist insect herbivore to the fundamental nutritional niche of its microbial symbiont. Significance StatementColonies of Atta colombica leafcutter ants can contain millions of specialized workers exhibiting large-scale generalist herbivory. Yet, this generalist foraging niche also depends on the poorly understood physiological needs of the ants domesticated fungal cultivar. We show the cultivars fundamental nutritional niche is broad for carbohydrates but narrower for protein and a suite of minerals, but that the cultivars sensitivity to excess protein is also mediated by Al, Fe, and P. More generally, this study decouples the multidimensional foraging strategies that enable a generalist herbivore to navigate a complex nutritional landscape and mix many imbalanced foods to achieve balanced cultivar provisioning.

ecology↗

bioRxiv2020

Exploring the limits of ComBat method for multi-site diffusion MRI harmonization

The findings from diffusion-weighted magnetic resonance imaging (dMRI) studies often show inconsistent and sometimes contradictory results due to small sample sizes as well as differences in acquisition parameters and pre-/post-processing methods. To address these challenges, collaborative multi-site initiatives have provided an opportunity to collect larger and more diverse groups of subjects, including those with neuropsychiatric disorders, leading to increased power and findings that may be more representative at the group and individual level. With the availability of these datasets openly, the ability of joint analysis of multi-site dMRI data has become more important than ever. However, intrinsic- or acquisition-related variability in scanner models, acquisition protocols, and reconstruction settings hinder pooling multi-site dMRI directly. One powerful and fast statistical harmonization method called ComBat (https://github.com/Jfortin1/ComBatHarmonization) was developed to mitigate the "batch effect" in gene expression microarray data and was adapted for multi-site dMRI harmonization to reduce scanner/site effect. Our goal is to evaluate this commonly used harmonization approach using a large diffusion MRI dataset involving 542 individuals from 5 sites. We investigated two important aspects of using ComBat for harmonization of fractional anisotropy (FA) across sites: First, we assessed how well ComBat preserves the inter-subject biological variability (measured by the effect sizes of between-group FA differences) after harmonization. Second, we evaluated the effect of minor differences in pre-processing on ComBats performance. While the majority of effect sizes are mostly preserved in some sites after harmonization, they are not well-preserved at other sites where non-linear scanner contributions exist. Further, even minor differences in pre-processing can yield unwanted effects during ComBat harmonization. Thus, our findings suggest paying careful attention to the data being harmonized as well as using the same processing pipeline while using ComBat for data harmonization.

neuroscience↗

bioRxiv2020

Adolescent-onset and adult-onset schizophrenia: reduced ribosomal protein expression via mTOR signalling in patient-derived olfactory cells

Schizophrenia is a heterogeneous disorder associated with many genetic and environmental risk factors that could affect brain development. It is unknown whether adolescent-onset and adult-onset schizophrenia have similar aetiology. To address this we used discovery-based proteomics to find proteins differentially expressed in olfactory neurosphere-derived cells from adolescents with schizophrenia compared to age- and gender-matched healthy controls. Of 1638 proteins identified, 241 were differentially expressed in patient cells, with significant down-regulation of ribosomal and cytoskeletal proteins, and dysregulation of protein synthesis pathways. We then re-analysed our previous adult-onset proteomic data to compare directly with adolescent-onset protein expression. Schizophrenia-associated protein expression in adult-onset patients was remarkably similar to adolescent-onset patients. To increase sample size and power we combined the two datasets for a bioinformatic meta-analysis. Schizophrenia-associated protein expression indicated significant downregulation of the mTOR signalling pathway, which regulates protein synthesis, indicated by the reduced expression of all ribosomal proteins and other mTOR-dependent proteins: RPS6, VIM, LDHB and PPP2R1A. A protein-protein interaction network built from differentially expressed proteins in the combined dataset was significantly associated with schizophrenia-associated risk genes and with proteins regulating neural stem cell differentiation, cell adhesion and growth cones in the developing brain. This study demonstrates that despite the divergent age of onset, the proteomes of olfactory neural stem cells of adolescent- and adult-onset patients are remarkably similar. The dysregulated proteins in patient cells form a tightly interconnected protein-protein interaction network associated with mTOR signalling, protein translation, neurogenesis and axon growth - all key components of brain development.

neuroscience↗

bioRxiv2020

Hidden features of the malaria vectormosquito, Anopheles stephensi ,revealed by a high-quality referencegenome

BackgroundThe mosquito Anopheles stephensi is a vector of urban malaria in Asia that recently invaded Africa. Studying the genetic basis of vectorial capacity and engineering genetic interventions are both impeded by limitations of a vectors genome assembly. The existing assemblies of An. stephensi are draft-quality and contain thousands of sequence gaps, potentially missing genetic elements important for its biology and evolution. ResultsTo access previously intractable genomic regions, we generated a reference-grade genome assembly and full transcript annotations that achieve a new standard for reference genomes of disease vectors. Here, we report novel species-specific transposable element families and insertions in functional genetic elements, demonstrating the widespread role of TEs in genome evolution and phenotypic variation. We discovered 29 previously hidden members of insecticide resistance genes, uncovering new candidate genetic elements for the widespread insecticide resistance observed in An. stephensi. We identified 2.4 Mb of the Y-chromosome and seven new male-linked gene candidates, representing the most extensive coverage of the Y-chromosome in any mosquito. By tracking full length mRNA for >15 days following blood feeding, we discover distinct roles of previously uncharacterized genes in blood metabolism and female reproduction. The Y-linked heterochromatin landscape reveals extensive accumulation of long-terminal repeat retrotransposons throughout the evolution and degeneration of this chromosome. Finally, we identify a novel Y-linked putative transcription factor that is expressed constitutively through male development and adulthood, suggesting an important role throughout male development. ConclusionCollectively, these results and resources underscore the significance of previously hidden genomic elements in the biology of malaria mosquitoes and will accelerate development of genetic control strategies of malaria transmission.

genomics↗

bioRxiv2020

Cooperative and non-cooperative behaviour in the exploitation of a common renewable resource with environmental stochasticity

Classical fisheries biology aims to optimise fisheries-level outcomes, such as yield or profit, by controlling the fishing effort. This can be adjusted to allow for the effects of environmental stochasticity, or noise, in the population dynamics. However, when multiple fishing entities, which could represent countries, commercial organisations, or individual vessels, can autonomously determine their own fishing effort, the the optimal action for one fishing entity depends on the actions of others. Coupled with noise in the population dynamics, and with decisions about fishing effort made repeatedly, this becomes an iterated stochastic game. We tackle this problem using the tools of stochastic optimisation, first for the monopolists problem and then for the duopolists problem. In each case, we derive optimal policies that specify the best level of fishing effort for a given stock biomass. Under these optimal policies, we can calculate the equilibrium stock biomass, the expected long-term return from fishing and the probability of stock collapse. We also show that there is a threshold stock biomass below which it is optimal to stop fishing until the stock recovers. We then develop an agent-based model to test the effectiveness of simple strategies for responding to deviations by an opponent from a cooperative fishing level. Our results show that the economic value of the fishery to a monopolist, or to a consortium of fishing agents, is robust to a certain level of noise. However, without the means of making agreements about fishing effort, even low levels of noise make sustained cooperation between autonomous fishing agents difficult.

ecology↗

bioRxiv2019

Regulation of heterotopic ossification through local inflammatory monocytes in a mouse model of aberrant wound healing

Heterotopic ossification (HO) is an aberrant regenerative process with ectopic bone induction in response to musculoskeletal trauma, in which mesenchymal stem cells (MSC) differentiate into osteochondrogenic cells instead of myocytes or tenocytes. Despite frequent cases of hospitalized musculoskeletal trauma, the inflammatory responses and cell population dynamics that regulate subsequent wound healing and tissue regeneration are still unclear. Here we examine, using a mouse model of trauma-induced HO, the local microenvironment of the initial post-injury inflammatory response. Single cell transcriptome analyses identify distinct monocyte/macrophage populations at the injury site, with their dynamic changes over time elucidated using trajectory analyses. Mechanistically, transforming growth factor beta-1 (TGF{beta}1)-producing monocytes/macrophages are associated with HO and aberrant chondrogenic progenitor cell differentiation, while CD47-activating peptides that reduce systemic TGF{beta} levels help ameliorate HO. Our data thus implicate CD47 activation as a novel therapeutic approach for modulating monocyte/macrophage phenotypes, MSC differentiation and HO formation during wound healing.

immunology↗

bioRxiv2019

White Matter Microstructure and its Relation to Clinical Features of Obsessive-Compulsive Disorder: Findings from the ENIGMA OCD Working Group

ImportanceMicrostructural alterations in cortico-subcortical connections are thought to be present in Obsessive-Compulsive Disorder (OCD). However, prior studies have yielded inconsistent findings, perhaps because small sample sizes provided insufficient power to detect subtle abnormalities. ObjectiveTo investigate microstructural white matter alterations and their relation to clinical features in the largest dataset of adult and pediatric OCD to date. Design, Setting, and ParticipantsIn this cross-sectional case-control magnetic resonance study, we investigated diffusion tensor imaging metrics from 700 adult patients and 645 adult controls, as well as 174 pediatric patients and 144 pediatric controls across 19 sites participating in the ENIGMA-OCD Working Group. Main Outcomes and MeasuresWe extracted measures of fractional anisotropy (FA) as main outcome, and mean diffusivity, radial diffusivity, and axial diffusivity as secondary outcomes for 25 white matter regions. We meta-analyzed patient-control group differences (Cohens d) across sites, after adjusting for age and sex, and investigated associations with clinical characteristics. ResultsAdult OCD patients showed significant FA reduction in the sagittal stratum (d=-0.21, z=-3.21, p=0.001) and posterior thalamic radiation (d=-0.26, z=-4.57, p<0.0001). In the sagittal stratum only, lower FA was associated with a younger age of onset (z=2.71, p=0.006), longer duration of illness (z=-2.086, p=0.036) and a higher percentage of medicated patients in the cohorts studied (z=-1.98, p=0.047). No significant association with symptom severity was found. Pediatric OCD patients did not show any detectable microstructural abnormalities compared to matched controls. Conclusions and RelevanceMicrostructural alterations in projection and association fibers to posterior brain regions were found in adult OCD, and related to disease course and medication status. Such results are relevant to models positing deficits in connectivity as a crucial mechanism in OCD. KEY POINTSO_ST_ABSQuestionC_ST_ABSDo patients with Obsessive-Compulsive Disorder (OCD) show white matter microstructural alterations, and are these alterations related to clinical features? FindingsData from 19 sites of the ENIGMA-OCD Consortium were included, involving 700 adult patients and 645 adult controls, 174 pediatric patients and 144 pediatric controls. Diffusion tensor imaging data were meta-analyzed using a harmonized data processing and analysis protocol. Adult, but not pediatric, patients showed alterations in the sagittal stratum and posterior thalamic radiation; sagittal stratum differences were associated with clinical features. MeaningMicrostructural abnormalities found in adult but not in the pediatric cohort, are related to illness duration and medication status.

neuroscience↗

bioRxiv2019

Kia kaua te reo e rite ki te moa, ka ngaro: Do not let the language suffer the same fate as the moa

More than a third of the worlds languages are currently classified as endangered and more than half are expected to go extinct by 2100. Strategies aimed at revitalising endangered languages have been implemented in numerous countries, with varying degrees of success. Here, we develop a new model regarding language transmission by dividing the population into defined proficiency categories and dynamically quantifying transition rates between categories. The model can predict changes in proficiency levels over time and, ultimately, whether a given endangered language is on a long-term trajectory towards extinction or recovery. We calibrate the model using data from Wales and show that the model predicts that the Welsh language will thrive in the long term. We then apply the model to te reo M[a]ori, the Indigenous language of New Zealand, as a case study. Initial conditions for this model are estimated using New Zealand census data. We modify the model to describe a country, such as New Zealand, where the endangered language is associated with a particular subpopulation representing the Indigenous People. We conclude that, with current learning rates, te reo M[a]ori is on a pathway towards extinction, but identify strategies that could help restore it to an upward trajectory.

ecology↗

bioRxiv2019

Distinctive airway microbial ecology of smokers and asthmatics

BackgroundNormal airway microbial communities play a central role in respiratory health but are poorly characterized. Cigarette smoking is the dominant global environmental influence on lung function, and asthma has become the most prevalent chronic respiratory disease worldwide. Both conditions have major microbial components that are also poorly defined. MethodsWe investigated airway bacterial communities in a general population sample of 529 Australian adults. Posterior oropharyngeal swabs were analysed by sequencing of the 16S rRNA and methionine aminopeptidase genes. The microbiota were characterised according to their prevalence, abundance, and network memberships. FindingsMicrobial communities were similar across the population and were strongly organized into co-abundance networks. Smoking associated with diversity loss, negative effects on abundant taxa, profound alterations to network structure and expansion of Streptococcus spp. By contrast, the asthmatic microbiota were selectively affected by an increase in Neisseria spp. and by reduced numbers of low abundance but prevalent organisms. InterpretationOur study shows healthy airway microbiota are contained within a highly structured ecosystem, indicating balanced relationships between the microbiome and human host factors. The marked abnormalities in smokers may be pathogenic for chronic obstructive pulmonary disease (COPD) and lung cancer. The narrow spectrum of abnormalities in asthmatics encourages investigation of damaging and protective effects of specific bacteria. FundingThe study was funded by the Asmarley Trust and a Wellcome Senior Investigator Award to WOCC and MFM (P46009). The Busselton Healthy Ageing Study is supported by the Government of Western Australia (Office of Science, Department of Health) the City of Busselton, and private donations.

ecology↗

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↗

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↗