Introduction

To lay the foundation for this review, it is first necessary to clarify its scope and understand the meaning of the term "mental illness". The term "mental illness" and its related term "mental health" are frequently used, but their definitions are not always clear. In the context of neuroscience research, there is a great deal of interest in understanding the neural biological basis of cognitive, behavioral, and emotional phenomena in both healthy and ill states. Here, we will focus on phenomena related to illness. However, for mental illnesses, especially mood and anxiety disorders, the boundaries between health and illness are particularly unclear, and the boundaries between different illnesses are even harder to distinguish. In practice, psychiatrists and other relevant professionals use a fairly broad standard to differentiate between health and illness, which states that "mental health symptoms must cause clinically significant distress or impairment in social, occupational, or other important functioning areas" [1]. An individual's disease classification is more based on their symptoms than on laboratory tests or causal findings. For neuroscientists studying human subjects, this classification system may not be applicable in all areas, especially in identifying robust associations between general and specific diseases. Therefore, there has been a recent trend towards specific functional domain neuroscience research - for example, using the Research Domain Criteria (RDoC) framework proposed by the US National Institute of Mental Health [2] to elucidate neural pathways that may be disrupted in disease. These methods are particularly well-suited to neuroscience, as cross-species research into functional domains is more direct than studying multifaceted and heterogeneous diseases. Therefore, this review will pay less attention to models of specific diseases and focus more on how advances in understanding functional domains can lead to better care methods. In this context, it is also crucial to understand the definition of neuroscience. While symptoms of mental illness manifest in the brain, other organ systems also play a crucial role. For example, the immune system and the gastrointestinal system are believed to be associated with the risk of mental illness, and they transmit signals to the brain through hormones, neurotransmitters, or even cells, especially during neural development when critical neural circuits are being established and modified [3~6]. Humoral stress hormones serve as a symbol of the collaboration between the central nervous system and peripheral organs, and also directly or indirectly affect brain function, exerting profound effects on functional domains relevant to mental illnesses [7]. Although these domains have played an important role in the understanding and treatment of mental illnesses, this review will primarily focus on the progress made based on the brain itself. Understanding that mental illnesses originate from the brain and that the neuroscience of mental illness includes studying the brain basis of behavior, cognition, and emotional functions, this review will first discuss how past neuroscience findings have led to the current available clinical approaches for mental illnesses. Next, we will focus on significant recent advances in the fields of genetics, molecular neuroscience, neural circuits, and computational methods. In these areas, technological and conceptual innovations have greatly enhanced our understanding of the neural biology of functions related to mental illnesses. We will discuss the latest advances in each field and analyze the potential for these advances to lead to new clinical approaches.

The Past Forward: Neuroscience's Contributions to Current Therapies

When discussing the contribution of neuroscience to the current treatment of mental illness, we have to acknowledge a widely cited point: many of the major discoveries are due to chance.In fact, as Hyman pointed out [8], many drugs are introduced into psychiatry "by accident to a considerable extent".Lithium, for example, has been used to treat mood disorders since the end of the 19th century, long before its rediscovery in the 1940s [9].However, initial attempts and subsequent studies to validate its efficacy in the treatment of mania were based on erroneous pathophysiological theories [10].Despite the serendipitous nature of the discovery of these drug therapies, neuroscience has played an important supporting role. Chlorpromazine, the first antipsychotic drug, was developed from the benzothiazide class of antihistamines that have long been considered sedative [9].Originally, chlorpromazine was synthesized and studied as an adjunct to anesthetics; after failing in this use, it was attempted to sedate patients with manic-type psychosis and was found to reduce psychotic symptoms [11]. Similarly, the first antidepressant, imipramine, was developed from antihistamines; it was also used in psychiatric patients, inspired by the effects of chlorpromazine.Although it failed to alleviate psychotic symptoms, its effects suggest that it may be effective in reducing depressive symptoms.Further testing in patients with non-psychotic depression confirmed its therapeutic potential [12].Thus, the discovery of both drugs was based on an understanding of the role of histamine receptors, although their actual therapeutic potential differed from what was expected by neuroscience at the time. Neuroscience has undoubtedly played an important role in expanding the range of therapeutic options based on initial explorations in the field of psychopharmacology, most of which were the result of accidental discoveries.For example, following the discovery of chlorpromazine and the development of other phenothiazines, combined rodent behavioral studies and receptor binding experiments have shown that the efficacy of antipsychotics is related to the binding of D2 receptors (see Fig. 1) [13,14].This discovery paves the way for further development of more antipsychotic drugs. There are many similar discoveries: the mechanism of removing neurotransmitters through reuptake pumps has inspired the various new antidepressant drugs in use today.Most famously, Nobel laureate Julius Axelrod and his colleagues demonstrated that imipramine and other psychotropic drugs inhibit the reuptake of norepinephrine [15].

How to distinguish "bad guy" gallbladder polyps? Due to the fact that pseudopolyps do not pose a significant threat to the human body, not all gallbladder polyps need to be removed. In general, the examination and judgment of gallbladder polyps in medicine are mainly based on ultrasound, and in some special cases, magnetic resonance imaging can be used for further diagnosis. Therefore, whether gallbladder polyps can be removed is also based on the results of these two examinations.

An important contribution of basic neuroscience to existing therapies for mental illness has been the development of brain stimulation techniques.Direct electrical stimulation was first attempted to answer basic scientific questions, such as the role of the brain in the generation of motor and sensory perception [16], and later to solve problems of the mind [17].The functional maps generated by these studies inspired people to think that cognitive and emotional functions may also be localized in the brain, so they began to try to treat mental illness by stimulating specific brain areas more accurately than electroconvulsive therapy [18]. With the development of neuroimaging research, more support has been given to the hypothesis of dysfunctional localization of mental illness, and these early attempts have been greatly improved by methods that can accurately stimulate deep structures.Today, deep brain stimulation is being used in studies of major depressive disorder and obsessive-compulsive disorder [19, 20], and is expected to be further improved by a personalized approach based on patient-specific neurophysiological characteristics, inspired by basic science [21].

Building the Foundation for the Future: Genetics and Molecular Neuroscience

Although much has been achieved, the future of mental illness care will benefit greatly as our understanding of brain mechanisms becomes clearer.However, the progress of mechanistic research is still hindered by a major problem: it is impossible to identify the specific causes of mental illness and the specific effects of these causes on the brain. For a long time, researchers have believed that mental illness is the result of a combination of family heredity and environmental factors.Over the past few decades, these early inferences have been further refined: familial factors are mainly caused by genetic inheritance, but not entirely, while environmental factors include a series of external influences such as prenatal infection, famine, adverse childhood experiences, chronic stressors and acute traumatic events [22-27].Translating these risk factors into actionable biological information is challenging.However, the rapid progress in molecular neuroscience and genetics in recent years is likely to accelerate the discovery and application of new therapeutic targets.

1. Advances in understanding the genetics of mental illness Only 15 years ago, the field of the genetics of mental illness was still littered with unreproducible results and baffling ineffectual explorations.This has changed with the understanding of genome-wide association studies (GWAS), which have borrowed from other branches of medicine to reveal the underlying genetic structure of diseases with clear inheritance, such as schizophrenia and bipolar disorder, although these diseases do not follow Mendelian inheritance laws. Earlier efforts, represented by the Psychiatric Genomics Consortium (Psychiatric Genomics Consortium), have collected samples of hundreds of thousands or even hundreds of thousands of people, although these samples are mostly composed of individuals of European descent.These studies have revealed hundreds of genome-wide loci that are clearly associated with psychiatric disorders such as schizophrenia, bipolar disorder, and major depression [28].Each gene locus is a biological clue, which may reveal the mechanism information of the origin of mental illness, and then lead to the biological mechanism that can be targeted. The limitation of these early efforts is the lack of genetic ancestry diversity in current data sets.In order to ensure that the results of genetic research are applicable to all populations, concerted efforts must be made to increase the diversity of subjects in ongoing research.Therefore, the National Institute of Mental Health initiated the Ancestral Population Network (Ancestral Populations Network) [29,30], which aims to accelerate gene discovery in non-European populations.Increasing the diversity of genetic samples is not only expected to improve the applicability and fairness of genetic results in clinical practice, but also to enhance the understanding of the potential biological role of existing gene variants [31,32].

2. From genes to biology

As an example, in schizophrenia, the most critical GWAS locus currently identified is associated with a gene in the complement system [33].This risk allele appears to contain a duplication of the complement component 4A (C4A) gene, which may lead to increased protein expression [33]. Simultaneous basic neuroscience studies have shown that complement proteins in the central nervous system can bind to synapses and be degraded by microglia [34].Considerable evidence suggests that excessive synaptic pruning may have an impact on schizophrenia, especially in the subclinical stages before the onset of generalized psychosis.Thus, the overexpression of complement component 4A protein may increase the risk of schizophrenia by enhancing synaptic pruning.Current research is aimed at testing this hypothesis in patients and model systems. The discovery of complement component 4A is undoubtedly an important milestone in psychiatric genetics, which successfully links risk factors with neurobiological mechanisms.However, the vast majority of the hundreds of GWAS sites are not so directly linked to the mechanism.For most sites, it is difficult to identify the sequence variation associated with risk; even if the risk variation is identified, its functional impact is difficult to define.Even if the functional impact of these variants can be understood at the molecular level, their relative risk is small, which means that it is challenging to understand the consequences at the circuit and behavioral levels. A systems biology approach will be necessary because the risk of mental illness is likely to arise from many small effect variants acting together.These challenges have so far limited the direct impact of GWAS-based gene discovery on mental illness. In recent years, the emergence of large-scale sequencing technology is expected to lead to faster progress.In autism and schizophrenia, a number of individual variants have been identified, each of which appears to significantly increase the risk individually [35]. While there is still uncertainty about the reliability and universality of these findings, variants with larger effects are more likely to be associated with neurobiological consequences: they usually occur in the coding sequence of genes and therefore disrupt specific biological processes.From the perspective of therapeutic development, these large effect variants are suitable for restoring gene function through genetic strategies; currently, efforts to design and test these strategies have been carried out in fragile X syndrome and angel syndrome [36-39].

Building on the basics: prospects for new approaches to neural circuitry

Current evidence reveals that the pathophysiological mechanisms of psychiatric disorders, unlike most neurological disorders with neurodegenerative or anatomical changes, alter the functional properties of interconnected populations of neurons, ultimately disrupting this interrelationship [66].Therefore, to reveal the neural mechanisms of emotion in health and disease States, it is necessary to construct a conceptual framework that develops step by step from the properties of individual neurons, including location and electrochemical characteristics, and further explore how these cells integrate functions in circuits. In order to better understand the human brain, the BRAIN project and other projects have not only developed tools to identify the molecular and electrochemical characteristics of brain circuit components, but also to study how these components are integrated into complete neural circuits and monitor their activities synchronously.Most of these tools have been widely used in model organisms, such as mice, to study the functional properties of circuits. Despite the challenges of translating from mouse models to humans, these studies provide valuable knowledge to reveal how the brain encodes emotions through spatially separated neural circuits in healthy States and how this encoding changes in psychopathological States.In the long run, these methods, if translated from the laboratory to the clinic, may bring great potential for diagnosis and treatment based on the biology of emotional circuits.

Tool for measuring integrated loop component activity

For decades, the field has quantitatively analyzed changes in early gene expression through a histological perspective as a proxy for neuronal activation.It is encouraging that similar methods can now be used to detect the inhibition of neuronal activity [67].The combination of these methods allows us to broadly and unbiasedly examine the activity of cells associated with behavior throughout the brain. Viral tools have also been optimized in the past decade to facilitate the delivery of biosensors to precise neurons based on their activity patterns [68], connectivity patterns, intrinsic promoters or enhancers [69], or enzymes that these cells are designed to express.Currently, a variety of novel fluorescent sensors are capable of detecting intracellular calcium, intracellular voltage, and levels of neurotransmitters and neuropeptides with high temporal resolution [70-72].These sensors have been combined with an expanded toolbox, including microendoscopy, wide-field imaging, and fiber-optic photometry, to assess the activity of biologically defined cell types in free-acting preclinical model organisms. In addition, many of these methods have been extended in the past few years, allowing scientists to monitor neural activity in multiple brain regions simultaneously [73], bringing the field closer to quantifying the activity of complete circuits distributed across multiple brain regions.Silicon-based probes developed in the last decade are able to measure the electrical activity of neurons from multiple brain regions simultaneously.Indeed, several experimental designs have enabled simultaneous monitoring of thousands of neurons from dozens of brain regions in preclinical models.Research is now aimed at combining this technology with methods to manipulate the activity of precise circuit elements to achieve the long-term goal of deploying these devices in the human brain [74].

Towards Human Neuroscience

As mentioned above, most of the research work on circuit tools has been carried out in mouse models.Ultimately, these new technologies need to reveal the pathological mechanisms of mental illness and apply them to humans.This exploration has already begun.Simultaneous intracranial electrical recordings from multiple sites have identified circuits that may play a role in anxiety and emotional dysfunction [81,82]. One study found that beta (13-30 Hz) oscillations between the hippocampus and amygdala were synchronized in subjects with high anxiety traits (see Fig. 2) [82].Follow-up preclinical studies revealed synchronization of beta oscillations between these regions during anxiety-related behavior in mice and linked it to activity in local somatostatin-positive interneurons.

Integration across tiers and computational methods

One of the central unsolved problems in psychiatry is how to make accurate individual predictions, such as determining who will respond to a particular treatment or identifying who is at risk of relapse.Data-driven computational psychiatry uses machine learning algorithms to predict key clinical outcomes of mental illness, including treatment response or risk of relapse.By analyzing large amounts of patient data to identify key genetic, sociodemographic or biomarkers to predict disease progression and other clinically relevant outcomes, these models have become important tools to promote precision medicine, even if little is known about the underlying mechanisms of disease. In fact, data-driven strategies have had several successes in predicting clinically relevant outcomes in psychiatry, particularly in predicting remission or treatment response in depression [90-94] or schizophrenia [95].However, there are concerns that current results may be overly optimistic and may not always be generalizable to a broader patient population [96-98]. For example, Chekroud et al. [96] found that a machine learning model for predicting antipsychotic treatment response worked in one trial but failed to generalize to data from other trials.Part of the reason is that most clinical models are still validated within the training dataset [96, 99]; and to truly validate the accuracy of a data-driven model, it must be tested on patients that the model has never seen [97].Moreover, these models tend to perform poorly in the presence of large amounts of unexplained heterogeneity [96, 98], due to the failure of purely data-driven approaches to consider the underlying mechanisms.

New Advances in Computational Psychiatry

In recent years, mechanistic models have greatly advanced our understanding of a variety of psychiatric disorders, including schizophrenia [117,125,129-135], obsessive-compulsive disorder [126,136-138], depression [91,139,140], and addiction [141-145].One of the most promising advances is the combination of theory-driven models with data-driven methods [89, 93, 146, 147].Specifically, the use of parameter estimates obtained from theory-driven models as predictor variables for machine learning models improves performance in predicting disease-related outcomes more than the use of traditional behavioral and brain activity measures alone [89, 117, 118, 148]. A recent study has shown that the use of a computational model to fit the reverse saccade behavior can distinguish different early stages of Huntington's disease, which can not be done by itself [148].The reason for this performance improvement lies in the ability of parameter estimation in theory-driven models to capture hidden heterogeneity in patient populations while reducing overall noise, thereby improving the performance of machine learning models. Genome-wide association study (GWAS) studies of disease phenotypes have revealed underlying genetic risks that are mapped onto cell biology by molecular neuroscience tools.Circuit technology connects the functions and behaviors of specific cells and circuit hierarchies.The computational method refines the disease phenotype and defines the basic components of the behavior and their mapping to the circuit.Together, these tools have created a virtuous circle that has increased our understanding of mental illness, paving the way for modern neuroscience-inspired treatments. The combination of data-driven and theory-driven methods puts new demands on big data, which requires us to be able to integrate individual in-depth assessment ( "deep data") with large patient population assessment ( "breadth data").Ideally, these data should be collected longitudinally, be able to predict remission, relapse, treatment response, or other clinically relevant variables prospectively, and be confirmed by independent validation data sets.To this end, different methods for collecting depth and breadth data are being explored. One potential approach is the use of smartphones for cognitive and symptom assessment [149,150].Smartphone assessment has the advantage of being able to monitor patients continuously over a long period of time and to cover patient populations that are underrepresented in the study due to geographic or mobility constraints.In addition, smartphones provide a unique opportunity to connect real-life symptom data with mechanistic cognitive structures and modeling methods [151].

Look to the future

Given the multiple advances at the molecular, circuit, and computational levels, there is optimism about the future contribution of neuroscience (see Figure 3).However, we still have a lot of work to do.In particular, the translation of neuroscience knowledge from rodents and other organisms with the potential for genetic manipulation into treatments for humans still requires careful consideration and more model systems to support. In this regard, stem cell-based models, such as two-dimensional cultures or three-dimensional organoids, are of great interest in studying circuit processes and directly connecting to genetic factors in human tissues.The use of intermediate large animal models (such as non-human primates) can also help to achieve this transformation.Moreover, the specificity of circuit technology often masks the complexity of how these circuits actually work in the brain — not individually, one at a time, but through tight connections. An unresolved question is whether predictive models constructed by computational methods that combine theory-driven and data-driven approaches can identify actionable therapeutic targets or remain relevant when applied to real-world patient populations.Despite the challenges, the future is promising: intensive research into the neurobiological underpinnings of mental illness promises to lead to the development of targeted and transformative new therapies.This beautiful vision, relying on the research accumulation of the past decades and with the help of today's revolutionary innovative methods, is gradually moving from conception to reality.