Biology subjects

Choi, K.

Publications and source records attributed to Choi, K..

At least 37 records · Page 2Linked to original sources

bioRxiv2024

Infralimbic activity during REM sleep facilitates fear extinction memory

Rapid eye movement (REM) sleep is known to facilitate fear extinction and play a protective role against fearful memories. Consequently, disruption of REM sleep after a traumatic event may increase the risk for developing PTSD. However, the underlying mechanisms by which REM sleep promotes extinction of aversive memories remain largely unknown. The infralimbic cortex (IL) is a key brain structure for the consolidation of extinction memory. Using calcium imaging, we found in mice that most IL pyramidal neurons are intensively activated during REM sleep. Optogenetically suppressing the IL activity during REM sleep within a 4-hour window after auditory-cued fear conditioning impaired extinction memory consolidation. In contrast, REM-specific inhibition of the IL cortex after extinction learning did not affect the extinction memory. Whole-cell patch-clamp recordings demonstrated that inactivating IL neurons during REM sleep depresses their excitability. Together, our findings demonstrate that REM sleep after fear conditioning facilitates fear extinction by enhancing IL excitability, and highlight the importance of REM sleep in the aftermath of traumatic events for protecting against traumatic memories.

neuroscience↗

bioRxiv2023

Arrested Agonist Paradigm For Selective Radiosensitization of Prostate Cancer

As a prototypical nuclear hormone receptor, the androgen receptor (AR) signals via a sequential cascade triggered by binding to androgenic ligands such as testosterone and dihydrotestosterone (DHT). This cascade includes dimerization of the ligand-receptor complex, nuclear translocation, chromatin binding to response elements, recruitment of TOP2B and co-activator complexes, and induction of an effector transcriptional program. In prostate cancers, this AR signaling cascade is an essential driver of growth and survival, yet its activity confers potential vulnerabilities through transient TOP2B-mediated DNA double strand breaks. We investigated the ability of non-steroidal AR ligands to activate initial steps of the AR signaling cascade up to the point of AR- and TOP2B-mediated double strand breaks, with subsequent arrest of the signaling cascade to prevent induction of pro-growth/survival transcriptional programs in prostate cancer cells. We identified hydroxyflutamide (FLU) as such an androgen receptor arrested agonist; in androgen-deprived conditions, FLU induced AR nuclear translocation, chromatin binding, and TOP2B-mediated double strand breaks, but failed to induce AR target gene expression and prostate cancer cell growth. The FLU-mediated arrest in the signaling cascade could be attributed to the inability of FLU to allow association of AR with SMARCD2, a critical component of the BAF chromatin remodeling complex required for androgen induced AR co-activation. Interestingly, the FLU-induced, AR- and TOP2B-mediated double strand breaks could be used to selectively sensitize AR-positive prostate cancer cells to ionizing radiation in vitro and in vivo. These findings support a novel arrested agonist paradigm for selective radiosensitization of prostate cancer cells without inducing AR-mediated pro-growth and survival transcriptional programs.

cancer biology↗

bioRxiv2023

Identification of Genetic and Genomic Influences on Progressive Ethanol Consumption in Diversity Outbred Mice

Genetic factors play a significant role in the risk for development of alcohol use disorder (AUD). Using 3-bottle choice intermittent access ethanol (IEA), we have employed the Diversity Outbred (DO) mouse panel as a model of alcohol use disorder in a genetically diverse population. Through use of gene expression network analysis techniques, in combination with expression quantitative trait loci (eQTL) mapping, we have completed an extensive analysis of the influence of genetic background on gene expression changes in the prefrontal cortex (PFC). This approach revealed that, in DO mice, genes whose expression was significantly disrupted by intermittent ethanol in the PFC also tended to be those whose expression correlated to intake. This finding is in contrast to previous studies of both mice and nonhuman primates. Importantly, these analyses identified genes involved in myelination in the PFC as significantly disrupted by IEA, correlated to ethanol intake, and having significant eQTLs. Genes that code for canonical components of the myelin sheath, such as Mbp, also emerged as key drivers of the gene expression response to intermittent ethanol drinking. Several regulators of myelination were also key drivers of gene expression, and had significant QTLs, indicating that genetic background may play an important role in regulation of brain myelination. These findings underscore the importance of disruption of normal myelination in the PFC in response to prolonged ethanol exposure, that genetic variation plays an important role in this response, and that this interaction between genetics and myelin disruption in the presence of ethanol may underlie previously observed behavioral changes under intermittent access ethanol drinking such as escalation of consumption.

genomics↗

bioRxiv2023

Unveiling the Odor Representation in the Inner Brain of Drosophila through Compressed Sensing

The putative dimension of a space spanned by chemical stimuli is deemed enormous; however, when odorant molecules are bound to a finite number of receptor types and their information is transmitted and projected to a perceptual odor space in the brain, a substantial reduction in dimensionality is made. Compressed sensing (CS) is an algorithm that enables recovery of high-dimensional signals from the data compressed in a lower dimension when the representation of such signals is sufficiently sparse. By analyzing the recent Drosophila connectomics data, we find that the Drosophila olfactory system effectively meets the prerequisites for CS to work. The neural activity profile of projection neurons (PNs) can be faithfully recovered from a low-dimensional response profile of mushroom body output neurons (MBONs) which can be reconstructed using the electro-physiological recordings to a wide range of odorants. By leveraging the residuals calculated between the measured and the predicted MBON responses, we visualize the perceptual odor space by means of residual spectrum and discuss the differentiability of an odor from others. Our study highlights the sparse coding of odor to the receptor space as an essential component for odor identifiability, clarifying the concentration-dependent odor percept. Further, a simultaneous exposure of the olfactory system to many different odorants saturates the neural activity profile of PNs, significantly degrading the capacity of signal recovery, resulting in a perceptual state analogous to "olfactory white." Our study applying the CS to the connectomics data provides novel and quantitative insights into the odor representation in the inner brain of Drosophila.

neuroscience↗

bioRxiv2023

Schwann cells modulate nociception in neurofibromatosis 1

Pain of unknown etiology is frequent in individuals with the tumor predisposition syndrome Neurofibromatosis 1 (NF1), even when tumors are absent. Schwann cells (SC) were recently shown to play roles in nociceptive processing, and we find that chemogenetic activation of SCs is sufficient to induce afferent and behavioral mechanical hypersensitivity in mice. In mouse models, animals show afferent and behavioral hypersensitivity when SC, but not neurons, lack Nf1. Importantly, hypersensitivity corresponds with SC-specific upregulation of mRNA encoding glial cell line derived neurotrophic factor (GDNF), independent of the presence of tumors. Neuropathic pain-like behaviors in the NF1 mice were inhibited by either chemogenetic silencing of SC calcium or by systemic delivery of GDNF targeting antibodies. Together, these findings suggest that Nf1 loss in SCs causes mechanical pain by influencing adjacent neurons and, data may identify cell-specific treatment strategies to ameliorate pain in individuals with NF1. Graphical AbstractGDNF released from Schwann cells acts on sensory neurons leading to mechanical hypersensitivity and pain-like behaviors in preclinical models of NF1. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=181 SRC="FIGDIR/small/533004v1_ufig1.gif" ALT="Figure 1"> View larger version (23K): org.highwire.dtl.DTLVardef@d17177org.highwire.dtl.DTLVardef@706f56org.highwire.dtl.DTLVardef@19db1ccorg.highwire.dtl.DTLVardef@3e28d_HPS_FORMAT_FIGEXP M_FIG C_FIG

neuroscience↗

bioRxiv2022

Multimodal Imaging-Based Classification of PTSD Using Data-Driven Computational Approaches: A Multisite Big Data Study from the ENIGMA-PGC PTSD Consortium

BackgroundCurrent clinical assessments of Posttraumatic stress disorder (PTSD) rely solely on subjective symptoms and experiences reported by the patient, rather than objective biomarkers of the illness. Recent advances in data-driven computational approaches have been helpful in devising tools to objectively diagnose psychiatric disorders. Here we aimed to classify individuals with PTSD versus controls using heterogeneous brain datasets from the ENIGMA-PGC PTSD Working group. MethodsWe analyzed brain MRI data from 3,527 structural-MRI; 2,502 resting state-fMRI; and 1,953 diffusion-MRI. First, we identified the brain features that best distinguish individuals with PTSD from controls (TEHC and HC) using traditional machine learning methods. Second, we assessed the utility of the denoising variational autoencoder (DVAE) and evaluated its classification performance. Third, we assessed the generalizability and reproducibility of both models using leave-one-site-out cross-validation procedure for each modality. ResultsWe found lower performance in classifying PTSD vs. controls with data from over 20 sites (60% test AUC for s-MRI, 59% for rs-fMRI and 56% for d-MRI), as compared to other studies run on single-site data. The performance increased when classifying PTSD from HC without trauma history across all three modalities (75% AUC). The classification performance remained intact when applying the DVAE framework, which reduced the number of features. Finally, we found that the DVAE framework achieved better generalization to unseen datasets compared with the traditional machine learning frameworks, albeit performance was slightly above chance. ConclusionOur findings highlight the promise offered by machine learning methods for the diagnosis of patients with PTSD. The utility of brain biomarkers across three MRI modalities and the contribution of DVAE models for improving generalizability offers new insights into neural mechanisms involved in PTSD. Significance Classifying PTSD from trauma-unexposed healthy controls (HC) using three imaging modalities performed well ([~]75% AUC), but performance suffered markedly when classifying PTSD from trauma-exposed healthy controls (TEHC) using three imaging modalities ([~]60% AUC). Using deep learning for feature reduction (denoising variational auto-encoder; DVAE) dramatically reduced the number of features with no concomitant performance degradation. Utilizing denoising variational autoencoder (DVAE) models improves generalizability across heterogeneous multi-site data compared with the traditional machine learning frameworks

neuroscience↗

bioRxiv2022

Sniper2L, a high-fidelity Cas9 variant with high activity

Although several high-fidelity SpCas9 variants that have reduced activities at mismatched target sequences have been reported, it has been observed that this increased specificity is associated with reduced on-target activity, limiting the applications of the high-fidelity variants when efficient genome editing is required. Here, we developed an improved version of Sniper-Cas9, Sniper2L, which represents an exception to this trade-off trend as it showed higher specificity with retained high activity. We evaluated Sniper2L activities at a large number of target sequences, and developed DeepSniper, a deep-learning model that can predict the activity of Sniper2L. We also confirmed that Sniper2L can induce highly efficient and specific editing at a large number of target sequences when it is delivered as a ribonucleoprotein complex. Mechanically, the high specificity of Sniper2L originates from its superior ability to avoid unwinding a target DNA containing even a single mismatch. We envision that Sniper2L will be useful when efficient and specific genome editing is required.

bioengineering↗

bioRxiv2022

Detecting isoform-level allelic imbalance accounting for inferential uncertainty

Allelic imbalance (AI) of gene expression in heterozygous individuals is a hallmark of cis-genetic regulation, revealing mechanisms underlying the association of non-coding genetic variation with downstream traits, as in GWAS. Most methods for detecting AI from RNA-sequencing (RNA-seq) data examine allelic expression per exonic SNP, which may obscure imbalance in expression of individual isoforms. Detecting AI at the isoform level requires accounting for inferential uncertainty (IU) of expression estimates, caused by multi-mapping of RNA-seq reads to isoforms and alleles. Swish, a method developed previously for differential transcript expression accounting for IU, can be applied in a paired setting to detect AI. However, in AI analysis, most transcripts will have high IU across alleles such that even methods like Swish will lose power. Our proposed method, SEESAW, offers AI analysis at various level of resolution, including gene level, isoform level, and optionally aggregating isoforms within a gene based on their transcription start site (TSS). This TSS-based aggregation strategy strengthens the signal for transcripts that may have high IU with respect to allelic quantification. SEESAW is primarily designed for experiments with multiple replicates or conditions of organisms with the same genotype, as in an F1 cross or time course experiments of cell lines. Additionally, we introduce a new test for detecting AI that changes across a continuous covariate, as in a time course experiment. The SEESAW suite of methods is evaluated both on simulated data and applied to an RNA-seq dataset of differentiating F1 mouse osteoblasts.

bioinformatics↗

bioRxiv2022

Interleukin-1α and Leukemia Inhibitory Factor Promote Extramedullary Hematopoiesis

Extramedullary hematopoiesis (EMH) expands hematopoietic capacity outside of the bone marrow in response to inflammatory conditions, including infections and cancer. Because of its inducible nature, EMH offers a unique opportunity to study the interaction between hematopoietic stem and progenitor cells (HSPCs) and their niche. In cancer patients, the spleen frequently serves as an EMH organ and provides myeloid cells that may worsen pathology. Here, we examined the relationship between HSPCs and their splenic niche in EMH secondary to solid tumor. We identify an inflammatory gene signature characterized by TNF expression in HSPCs. We show a role for IL-1 in producing this gene signature and TNF expression in HSPCs in activating splenic niche activity. We also demonstrate that tumor-derived Leukemia Inhibitory Factor (LIF) induces proliferation of splenic niche cells. IL-1 and LIF display cooperative effects in activating EMH and are both upregulated in some human cancers. Together, these data expand avenues for developing niche-directed therapies and further exploring EMH accompanying inflammatory pathologies like cancer.

immunology↗

bioRxiv2022

Chronic inflammation suppresses del(5q)-like MDS HSCs via p53

Inflammation is associated with the pathogenesis of Myelodysplastic syndromes (MDS). Emerging evidence suggests that MDS hematopoietic stem and progenitor cells (HSPCs) exhibit an altered response to systemic low-grade inflammation, which contributes to their competitive advantage over wild-type HSPCs and ensuing hematopoietic defects. Deletion of the long arm of chromosome 5 (del(5q)) is the most common chromosomal abnormality in patients with MDS. Although this subtype of MDS contains several haploinsufficient genes that directly impact innate immune signaling, the effects of an inflammatory milieu on del(5q) MDS HSPCs remains poorly defined. Utilizing a model of del(5q)-like MDS, wherein two 5q genes, miR-146a and TIFAB, are deleted, we found that chronic low-grade inflammation impaired the function of del(5q)-like MDS HSPCs and contributed to a more severe disease. The del(5q)-like MDS HSPCs exposed to chronic inflammation became less quiescent, but without changes in cell viability. In response to inflammation, mouse and human del(5q) MDS HSPCs activated a partial p53 response. The impaired function and reduced cellular quiescence of del(5q) MDS HSPCs exposed to inflammation could be restored by deletion of p53. Since TP53 mutations are highly enriched in del(5q) AML patients following an initial MDS diagnosis, increased p53 activation in del(5q) MDS HSPCs due to inflammation may create a selective pressure for genetic inactivation of p53. These findings uncover the contribution of systemic inflammation on dyshematopoiesis in del(5q) MDS and provide a potential explanation for acquired p53 mutations in myeloid malignancies with del(5q).

cancer biology↗

bioRxiv2022

Mouse NEMP1 is required for erythropoiesis, nuclear envelope openings and enucleation of erythroblasts

Nuclear Envelope Membrane Proteins (NEMP) are a conserved family of nuclear envelope proteins that reside within the inner nuclear membrane. Even though Nemp1 knockout (KO) mice are overtly normal, they display a pronounced splenomegaly. This phenotype and recent reports describing a requirement for nuclear envelope openings during erythroblasts terminal maturation led us to examine a potential role for Nemp1 in erythropoiesis. Here, we report that Nemp1 knockout (KO) mice show peripheral blood defects, anemia in neonates, ineffective erythropoiesis, splenomegaly and stress erythropoiesis. The erythroid lineage of Nemp1 KO mice is overrepresented until the pronounced apoptosis of polychromatophilic erythroblasts. We show that NEMP1 localizes to the nuclear envelope of erythroblasts and their progenitors. Mechanistically, we discovered that NEMP1 accumulates into aggregates that localize near or at the edge of nuclear envelope openings and Nemp1 deficiency leads to a marked decrease of both nuclear envelope openings and ensuing enucleation. Together, our results for the first time demonstrate that NEMP1 is essential for nuclear envelope openings and erythropoietic maturation in vivo and provide the first mouse model of defective erythropoiesis directly linked to the loss of an inner nuclear membrane protein.

cell biology↗

bioRxiv2022

Polymer physics-based classification of neurons

Recognizing that diverse morphologies of neurons are reminiscent of structures of branched polymers, we put forward a principled and systematic way of classifying neurons that employs the ideas of polymer physics. In particular, we use 3D coordinates of individual neurons, which are accessible in recent neuron reconstruction datasets from electron microscope images. We numerically calculate the form factor, F (q), a Fourier transform of the distance distribution of particles comprising an object of interest, which is routinely measured in scattering experiments to quantitatively characterize the structure of materials. For a polymer-like object consisting of n monomers spanning over a length scale of r, F (q) scales with the wavenumber q(= 2{pi}/r) as F (q) [~] q-[D] at an intermediate range of q, where [D] is the fractal dimension or the inverse scaling exponent ([D] ={nu} -1) characterizing the geometrical feature (r [~] n{nu}) of the object. F (q) can be used to describe a neuron morphology in terms of its size (Rn) and the extent of branching quantified by [D]. By defining the distance between F (q)s as a measure of similarity between two neuronal morphologies, we tackle the neuron classification problem. In comparison with other existing classification methods for neuronal morphologies, our F (q)-based classification rests solely on 3D coordinates of neurons with no prior knowledge of morphological features. When applied to publicly available neuron datasets from three different organisms, our method not only complements other methods but also offers a physical picture of how the dendritic and axonal branches of an individual neuron fill the space of dense neural networks inside the brain.

neuroscience↗

bioRxiv2022

Olfactory responses of Drosophila are encoded in the organization of projection neurons

The projection neurons (PNs), reconstructed from electron microscope (EM) images of the Drosophila olfactory system, offer a detailed view of neuronal anatomy, providing glimpses into information flow in the brain. About 150 uPNs constituting 58 glomeruli in the antennal lobe (AL) are bundled together in the axonal extension, routing the olfactory signal received at AL to mushroom body (MB) calyx and lateral horn (LH). Here we quantify the neuronal organization by inter-PN distances and examine its relationship with the odor types sensed by Drosophila. The homotypic uPNs that constitute glomeruli are tightly bundled and stereotyped in position throughout the neuropils, even though the glomerular PN organization in AL is no longer sustained in the higher brain center. Instead, odor-type dependent clusters consisting of multiple homotypes innervate the MB calyx and LH. Pheromone-encoding and hygro/thermo-sensing homotypes are spatially segregated in MB calyx, whereas two distinct clusters of food-related homotypes are found in LH in addition to the segregation of pheromone-encoding and hygro/thermo-sensing homotypes. We find that there are statistically significant associations between the spatial organization among a group of homotypic uPNs and certain stereotyped olfactory responses. Additionally, the signals from some of the tightly bundled homotypes converge to a specific group of lateral horn neurons (LHNs), which indicates that homotype (or odor type) specific integration of signals occurs at the synaptic interface between PNs and LHNs. Our findings suggest that before neural computation in the inner brain, some of the olfactory information are already encoded in the spatial organization of uPNs, illuminating that a certain degree of labeled-line strategy is at work in the Drosophila olfactory system.

neuroscience↗

bioRxiv2021

Conserved angio-immune subtypes of the cancer microenvironment predict response to immune checkpoint blockade therapy

Tumor microenvironment (TME) shapes the tumor progression and therapy outcome. Particularly, tumor angiogenesis and immunity impact the effect of immune checkpoint blockade (ICB) therapy. Here, we analyzed the transcriptome from 11,069 patients from The Cancer Genome Atlas (TCGA) to assess 91 functional gene sets corresponding to endothelial and T-cell activity. Intriguingly, TME across 30 non-hematological tumors can be classified into three distinct conserved angio-immune subtypes: high angiogenesis with low immune activity, low angiogenesis with high immune activity, and the one in-between. Remarkably, patients displaying TME with poor angiogenic activity with corresponding high immune activity show the most significant responses to ICB therapy in many cancer types. Notably, re-evaluation of the Javelin Renal 101, renal cell carcinoma clinical trial, provided compelling evidence that the baseline angiogenic state is critical in determining responses to checkpoint blockade. This study offers a clear rationale for incorporating baseline angiogenic state for ICB treatment decision-making.

cancer biology↗

bioRxiv2021

Distributed processing for action control by prelimbic circuits targeting anterior-posterior dorsal striatal subregions.

Fronto-striatal circuits have been extensively implicated in the cognitive control of behavioral output for both social and appetitive rewards. The functional diversity of prefrontal cortical populations is strongly dependent on their synaptic targets, with control of motor output strongly mediated by connectivity to the dorsal striatum. Despite evidence for functional diversity along the anterior-posterior axis of the dorsomedial striatum (DMS), it is unclear how distinct fronto- striatal sub-circuits support neural computations essential for action selection. Here we identify prefrontal populations targeting distinct DMS subregions and characterize their functional roles. We first performed neural circuit tracing to reveal segregated prefrontal populations defined by anterior/posterior dorsomedial striatal target. We then probed the functional relevance of these parallel circuits via in vivo calcium imaging and temporally precise causal manipulations during a feedback-based 2-alternative choice task. Single-photon imaging revealed circuit-specific representations of task-relevant information with prelimbic neurons targeting anterior DMS (PL::A- DMS) uniquely encoded choices and responses to negative outcomes, while prelimbic neurons targeting posterior DMS (PL::P-DMS) encoded internal representations of value and positive outcomes contingent on prior choice. Consistent with this distributed coding, optogenetic inhibition of PL::A-DMS circuits strongly impacted choice monitoring and behavioral control in response to negative outcomes while perturbation of PL::P-DMS signals impaired task engagement and strategies following positive outcomes. Di-synaptic retrograde tracing uncovered differences in afferent connectivity that may underlie these pathways functional divergence. Together our data uncover novel PL populations engaged in distributed processing for action control. SUMMARYO_LIPrelimbic cortex engages A- and P-DMS via distinct circuits C_LIO_LIPL::A-DMS and PL::P-DMS pathways encode divergent aspects of a simple goal-directed task C_LIO_LIPL::A-DMS pathways shape responding to negative outcomes via multiple mechanisms C_LIO_LIPL::P-DMS pathways guide engagement and choices in response to positive outcomes C_LIO_LIAfferent connectomes of PL neurons defined by A-P DMS target are distinct C_LI

neuroscience↗

bioRxiv2021

HEAT SHOCK FACTOR BINDING PROTEIN limits meiotic crossovers by repressing HEI10 transcription

The number of meiotic crossovers is tightly controlled and most depend on pro-crossover ZMM proteins, such as the E3 ligase HEI10. Despite the importance of HEI10 dosage for crossover formation, how HEI10 transcription is controlled remains unexplored. In a forward genetic screen using a sensitive fluorescent seed crossover reporter in Arabidopsis thaliana we identify heat shock factor binding protein (HSBP) as a repressor of HEI10 transcription and crossover numbers. Using genome-wide crossover mapping and cytogenetics, we show that hsbp mutations or meiotic HSBP knockdowns increase ZMM-dependent crossovers towards the telomeres, mirroring the effects of HEI10 overexpression. Through RNA sequencing, DNA methylome and chromatin immunoprecipitation analysis, we reveal that HSBP directly represses HEI10 transcription by binding with heat shock factors (HSFs) at the HEI10 promoter and maintaining DNA methylation over the HEI10 5' untranslated region. Our findings provide insights into how the temperature response regulator HSBP restricts meiotic HEI10 transcription and crossover number by attenuating HSF activity.

plant biology↗

bioRxiv2021

Airpart: Interpretable statistical models for analyzing allelic imbalance in single-cell datasets

MotivationAllelic expression analysis aids in detection of cis-regulatory mechanisms of genetic variation which produce allelic imbalance (AI) in heterozygotes. Measuring AI in bulk data lacking time or spatial resolution has the limitation that cell-type-specific (CTS), spatial-, or time-dependent AI signals may be dampened or not detected. ResultsWe introduce a statistical method airpart for identifying differential CTS AI from single-cell RNA-sequencing (scRNA-seq) data, or other spatially- or time-resolved datasets. airpart outputs discrete partitions of data, pointing to groups of genes and cells under common mechanisms of cis-genetic regulation. In order to account for low counts in single-cell data, our method uses a Generalized Fused Lasso with Binomial likelihood for partitioning groups of cells by AI signal, and a hierarchical Bayesian model for AI statistical inference. In simulation, airpart accurately detected partitions of cell types by their AI and had lower RMSE of allelic ratio estimates than existing methods. In real data, airpart identified differential AI patterns across cell states and could be used to define trends of AI signal over spatial or time axes. AvailabilityThe airpart package is available as an R/Bioconductor package at https://bioconductor.org/packages/airpart.

bioinformatics↗

bioRxiv2021

A Comparison of Methods to Harmonize Cortical Thickness Measurements Across Scanners and Sites

Results of neuroimaging datasets aggregated from multiple sites may be biased by site- specific profiles in participants demographic and clinical characteristics, as well as MRI acquisition protocols and scanning platforms. We compared the impact of four different harmonization methods on results obtained from analyses of cortical thickness data: (1) linear mixed-effects model (LME) that models site-specific random intercepts (LMEINT), (2) LME that models both site-specific random intercepts and age-related random slopes (LMEINT+SLP), (3) ComBat, and (4) ComBat with a generalized additive model (ComBat-GAM). Our test case for comparing harmonization methods was cortical thickness data aggregated from 29 sites, which included 1,343 cases with posttraumatic stress disorder (PTSD) (6.2-81.8 years old) and 2,067 trauma-exposed controls without PTSD (6.3-85.2 years old). We found that, compared to the other data harmonization methods, data processed with ComBat-GAM were more sensitive to the detection of significant case-control differences in regional cortical thickness (X2(3) = 34.339, p < 0.001), and case-control differences in age-related cortical thinning (X2(3) = 15.128, p = 0.002). Specifically, ComBat-GAM led to larger effect size estimates of cortical thickness reductions (corrected p-values < 0.001), smaller age-appropriate declines (corrected p-values < 0.001), and lower female to male contrast (corrected p-values < 0.001) in cases compared to controls relative to other harmonization methods. Harmonization with ComBat-GAM also led to greater estimates of age-related declines in cortical thickness (corrected p-values < 0.001) in both cases and controls compared to other harmonization methods. Our results support the use of ComBat-GAM for harmonizing cortical thickness data aggregated from multiple sites and scanners to minimize confounds and increase statistical power.

neuroscience↗