Home › Research

Research

Overview of research programs

My long-term research objective is to employ multimodal and multiscale modeling approaches to delineate the pathological brain dynamics and identify the pathophysiological mechanisms underlying neurological and psychiatric disorders including Alzheimer’s disease (AD), major depressive disorder (MDD) and epilepsy. Specifically, I will utilize dynamic brain modeling to elucidate the pathological network dynamics and dysfunctional network connectivity underlying specific brain disorders, and employ sophisticated and biophysically realistic models to understand the cellular, synaptic and circuit mechanisms giving rise to pathological dynamic states. My ultimate goal is to develop optimal closed-loop brain stimulation techniques to restore pathological brain dynamics to the normal state in collaboration with experimental and clinical neuroscientists. To achieve these long-term objectives, I will focus my research over the next five years on four projects: (1) Mechanistic identification of excitation-inhibition imbalance in AD; (2) Large-scale mapping of disrupted circuit interactions in MDD; (3) accurate localization of seizure onset zone via multiscale neural model inversion of resting-state functional MRI; and (4) Charting Excitation-inhibition balance over the human lifespan. These research topics are well grounded within my research experiences to date, and I believe that my extensive research experiences in computational neuroscience and neuroimaging will enable me to achieve my research objectives.

Mechanistic identification of excitation-inhibition imbalance in AD

Alzheimer’s disease (AD) is a neurodegenerative disorder that causes severe cognitive malfunction in individuals and tremendous financial burden to the society. It is the leading cause of dementia affecting more than 47 million people worldwide and this number is expected to increase to 131 million by 2050. The healthcare cost for patients with AD and other dementias is enormous and is estimated to be 236 billion in the US for 2016 alone and predicted to quadruple by 2050. Despite decades of intensive research, there are still no clear mechanistic understanding and effective treatment options for this devastating disease. Current treatments are only symptomatic without slowing down the progression of the disease. The lack of effective treatment highlights the paramount importance of identifying new pathophysiological and therapeutic targets. Excitation-inhibition balance represents a promising pathophysiological and therapeutic target for AD due to its critical role in controlling neuronal circuit functions and significant disruption in AD pathology. The goal of this project is to mechanistically identify excitation-inhibition (E-I) imbalance and its causal relationship with network dysfunction in AD using multiscale modeling of resting-state functional MRI.

Figure 1. Overview of the MEMI framework.

  • Accurate and individualized prediction of E-I imbalance in AD using data-driven neural model (NIH-R21 project)

The goal of this project is to validate and refine a Multiscale nEural Model Inversion (MEMI) framework for accurate and individualized estimation of excitation-inhibition (E-I) imbalance in Alzheimer’s Disease (AD) based on resting-state functional magnetic resonance imaging (rs-fMRI). To achieve this goal, we will pursue two specific aims. In Aim 1, we will predict disrupted E-I balance in an AD mouse model using MEMI of rs-fMRI. We will first perform ZTE-fMRI and dMRI on wild-type (WT) control and 3xTg-AD (TG) mice. We will then apply the MEMI model to predict regional E-I balance based on rs-fMRI and dMRI and derive areas with E-I impairments in AD mice. Based on MEMI predictions we will select four brain regions (three with the most significant E-I impairments in TG mice plus one control region) for in vivo optical measurements. In Aim 2, we will validate the MEMI model predictions using in vivo optical E-I measurements and behavioral testing. We will first perform simultaneous ZTE-fMRI and fiber photometry (at the four selected sites) in a different set of age-matched WT and TG mice as Aim 1. We will then validate the model predictions at both individual subject and group levels and improve the MEMI framework if model predictions deviate from empirical E-I measures. Lastly, we will examine if the E-I imbalance in TG mice is associated with cognitive impairments. The overarching goal of our research is to combine computational modeling, fMRI, and cutting-edge neuromodulation and recording tools to delineate pathological network activity, elucidate the underlying circuit mechanisms, and develop more effective treatment modalities for AD.

  • Multiscale modeling of E-I imbalance and network dysfunction in AD

The goal of this project is to mechanistically identify excitation-inhibition (E-I) imbalance and its causal relationship with network dysfunction in AD using multiscale modeling of resting-state functional MRI. We will test the central hypothesis that amyloid-β (Aβ) induced E-I imbalance underlies large-scale network dysfunction in AD. To achieve this goal, we have previously developed a Multiscale nEural Model Inversion (MEMI) framework based on functional MRI (fMRI) to estimate excitatory and inhibitory connection strengths in small- and medium-size neuronal networks. In Aim 1, we will first expand the small-scale MEMI to large-scale MEMI (LMEMI) to enable whole-brain estimation of effective connectivity and validate its accuracy and robustness. In Aim 2, we will identify impaired E-I balance in AD using the validated LMEMI on a large-cohort fMRI dataset, identify a core network for dynamical analysis of pathological state transition, and elucidate how E-I imbalance impacts large-scale network dynamics using a spiking neuron model. In Aim 3, we will develop a detailed biophysical model of the core AD network to delineate the cellular and synaptic mechanisms of E-I imbalance and identify its causal relationship with network dysfunction. The overarching goal of our research is to combine computational neuronal modeling with macroscale neuroimaging analysis to delineate  pathological network activity, elucidate the underlying cellular and circuit mechanisms, and develop more effective treatment modalities for AD

Large-scale mapping of disrupted circuit interactions in MDD

Major depressive disorder (MDD) is a serious mental illness that is characterized by depressed mood, diminished interests, and impaired cognitive function. It is a leading cause of chronic disability worldwide with a lifetime prevalence of up to 17%. Despite decades of extensive research, the etiology and pathophysiology of MDD remain not well understood. Functional magnetic resonance imaging (fMRI) remains a core method for studying the pathophysiological mechanisms of MDD, but existing analytic approaches focus primarily on macroscale systemic modeling of inter-regional interactions such as undirected functional connectivity (FC) and graph theory, which cannot provide a mechanistic understanding of MDD at the cellular and circuit levels. Also, the small sample size of most MDD neuroimaging studies causes low sensitivity and reliability. The goal of this project is to develop a multiscale modeling framework based on large-cohort resting-state functional MRI (rs-fMRI) to identify disrupted circuit interactions and delineate the underlying neurochemical basis in MDD. In Aim 1 of this proposal, I will first develop a new Large-scale nNeural Model Inversion (LEMI) framework that enables large-scale estimation of effective connectivity in whole-brain networks (up to 200 brain regions). The LEMI framework will then be applied to detect impaired circuit interactions in MDD using a large-cohort REST-meta-MDD dataset that consists of 1300 MDDs and 1128 normal control subjects. In Aim 2, we will construct biophysical neural mass models of the core and extended MDD networks to elucidate the neurochemical mechanisms of impaired circuit interactions. We will test the central hypothesis that limbic-cingulate-executive (triple system) malfunction caused by stress-induced aberrant glutamatergic and GABAergic neurotransmission underlies the core mechanism of MDD. The successful implementation of this proposal will lead to a unified computational framework that serves to identify pathophysiological mechanisms, formulate quantitative working hypotheses, and make testable predictions to guide experimental and clinical investigations for better diagnosis and treatment of MDD. The proposed work is also of broader significance to the fields of neuroscience, neuroimaging, psychiatry, and psychology since novel tools will be developed to enable the examination of cellular and circuit mechanisms of normal and aberrant cognitive functions.

Accurate localization of seizure onset zone via multiscale neural model inversion of resting-state functional MRI

Epilepsy is a common and devastating disorder that causes severe personal disability and tremendous financial burden to society. Despite pharmacotherapy, about one-third of patients suffer from persistent seizures leading to drug resistant epilepsy (DRE). As the most effective treatment option for DRE, resective surgery critically relies on the precise identification of seizure onset zone (SOZ) for its success. However, accurate localization of SOZ has been a challenging and resource-consuming task which often requires the deployment of invasive intracranial electroencephalograph (iEEG) imposing substantial risk to DRE patients. Deep brain stimulation (DBS) is an alternative treatment to DRE, but the selection of stimulation target remains largely a trial-and-error process. The goal of this study is to mechanistically identify SOZ for surgical planning and the optimal neuromodulation target for DBS in epilepsy through multi-level network modeling of resting-state functional MRI. We will test the central hypothesis that E-I imbalance serves as an important localizing biomarker for both the SOZ and the optimal DBS modulation target in DRE. To achieve this goal, we will pursue three aims. In Aim 1, we will derive SOZ candidates using a newly developed Multiscale Neural Model Inversion (MNM) framework and verify the predictions with surgical outcome. Specifically, we will apply the MNMI model to a high-quality rs-fMRI from DRE patients with hypothalamic hamartoma (HH) to estimate both intra-regional and inter-regional excitatory and inhibitory connection strengths. We hypothesize that the SOZ is associated with the highest regional E-I ratio, compared to pZs, and that when the predicted SOZ aligns/misaligns with the surgically-targeted SOZ, the patient will have a successful/failed surgical outcome by seizure frequency. In Aim 2, we will identify pathological E-I imbalance in the HH epileptic network and delineate its impact on seizure initiation and propagation. We will apply the MNMI model to a rs-fMRI dataset of age- and gender- matched healthy control (HC) subjects from the Baby Connectome Project (BCP) and Human Connectome Project (HCP) and identify disrupted E-I balance in DRE using statistical comparison. Next, we will develop a biophysical spiking model to reveal how the identified pathological E-I imbalance impairs epileptic network dynamics leading to seizure generation and propagation. We hypothesize that disrupted E-I balance in the epileptic network underlies pathological network dynamics enabling the initiation and propagation of epileptic discharges. In Aim 3, we will derive the optimal stimulation target for seizure abortion. We will apply periodic current pulses resembling DBS to each brain region separately in the epileptic network and identify the optimal stimulation site that best suppresses seizure activities. We hypothesize that the optimal stimulation target consists of the pZ region that exerts the largest inhibition on the SOZ, and the optimal stimulation aborts epileptic discharges by eliminating network bistability.

Charting excitation-inhibition balance over the human lifespan

Excitation-inhibition (E-I) balance plays a crucial role in normal circuit function and stability. Abnormal E-I balance has been hypothesized to be a key driver of impaired neural circuit functions leading to the pathogenesis of multiple neurological and psychiatric disorders. For example, it has been hypothesized that Alzheimer’s disease, the most common form of dementia, results from neuronal hyperactivation-driven amyloid-β pathology. However, till today there are still no normative reference charts for E-I balance that can be used to benchmark individual growth trajectories and to predict aberrant E-I balance for early detection of Alzheimer’s disease and other brain disorders. The goal of this project is to chart the normative developmental trajectories of E-I balance over the human lifespan using multiscale modeling of resting-state functional magnetic resonance imaging (rs-fMRI). To achieve this goal, we have previously developed a Multiscale Neural Model Inversion (MNMI) framework based on fMRI and diffusion MRI to estimate excitatory and inhibitory connection strengths for E-I inference in small- and medium-size neuronal networks. We have also applied the MNMI model to identify E-I imbalance in Alzheimer’s disease and major depressive disorder. In Aim 1, we will develop a new Large-scale nEural Model Inversion (LEMI) framework that enables E-I inference in whole-brain networks and establish its validity in E-I measurement using both ground-truth simulations and in vivo ground-truth E-I measurement in mice. We will also benchmark LEMI against other major E-I estimation approaches in fMRI. In Aim 2, we will apply the validated LEMI model to large-cohort high-quality BCP/HCP fMRI and diffusion MRI datasets to chart developmental E-I balance over the human lifespan. In addition, we will develop a large-scale biophysical spiking network model to elucidate the synaptic and circuit mechanisms of E-I balance changes over the life cycle. The overarching goal of our research is to combine computational neuronal modeling with macroscale neuroimaging analysis to delineate developmental quantitative E-I balance and functional network dynamics over the human lifespan.