The keyword “children” appears most frequently, indicating that researchers aim to investigate biomarkers from early childhood. Among the 100 most-cited articles, nearly 40% of the studies focused on the exploring the biomarkers in childhood. Studying autism biomarkers in early childhood may be due to several reasons. Autism typically manifests and is diagnosed during childhood, which is also when symptoms are most pronounced. Identifying biomarkers during this period can aid in early diagnosis (Chawarska et al., 2007; Ozonoff et al., 2011; Lord et al., 2018). Additionally, the complex etiology of autism, involving the interaction of genetic and environmental factors, makes it easier to obtain accurate data by studying childhood biomarkers, minimizing interference from other factors. This helps us better understand the origins, development, and trends of autism (Hallmayer et al., 2011; Gaugler et al., 2014; Sandin et al., 2014). Moreover, certain biomarkers may be more stable in childhood, as the nervous and physiological systems are not fully developed, making changes in biomarkers easier to detect. Researching childhood biomarkers can also reveal key pathways and targets in the pathophysiology of autism, enabling early intervention and treatment, and thereby reducing future burdens for individuals with autism.

The detection of biomarkers in childhood has become an effective way to predict autism in adulthood. Researches show that genetic, neuroimaging, and biochemical markers from childhood provide significant predictive accuracy for adult outcomes. For example, genetic variations such as FMR1 mutations or 16q11.2 copy number variations increase the likelihood of severe neurodevelopmental challenges in adulthood (Stessman et al., 2014; Sandin et al., 2017). Neuroimaging studies reveal weakened functional connectivity in the default mode network of autistic children, persisting into adulthood and predicting social difficulties. In summary, exploring childhood biomarkers is crucial for predicting adult autism severity and enabling early interventions.

Significant progress has been made in the genetic research of autism biomarkers, particularly in the fields of genes (keywords cluster #0), chromosomes, and cytogenetics (keywords cluster #8). Many studies on autism’s genetic causes have employed genome-wide screening to identify shared genetic markers within families. By analyzing genetic patterns in parents and siblings, researchers aim to uncover hereditary factors contributing to autism. Studies confirm a correlation between the incidence of autism and genetics, with family genetic loci and genes acting as hereditary factors in children with autism (Philippe et al., 1999; Risch et al., 1999; Muhle et al., 2004; Morrow et al., 2008; Anney et al., 2010). Many genes are associated with autism spectrum disorders, and copy number variations and rare gene mutations (e.g., SHANK3, NRXN1, CHD8, SCN2A) are closely linked to autism (Glessner et al., 2009). Regarding chromosomes, research shows that abnormalities such as duplications or deletions in the chromosomes of individuals with autism significantly increase the prevalence of the disorder, especially the duplication of chromosome 15q11-13 (Cook et al., 1998; Philippe et al., 1999; Buxbaum et al., 2002; Shen et al., 2010). Furthermore, variations in susceptibility regions like 7q22-q31, 2q24.3-q31, and 11p12-p13 affect neurodevelopment through multiple mechanisms, increasing the risk of autism (Consortium I.M.G.S.o.A, 1998, 2001; Muhle et al., 2004; Vorstman et al., 2006).

Related studies also reveal a correlation between the occurrence of autism and the interaction of genetic and environmental factors, as well as interactions between genes. For instance, the review “The genetics of autism” indicates that interactions among multiple genes lead to idiopathic autism, while epigenetic factors and exposure to environmental modulators may contribute to the variable expression of autism-related traits (Muhle et al., 2004; Rossignol and Frye, 2012b; Rossignol et al., 2014).

Among the 100 top-cited articles, 14 articles focus on clarifying the relationship between ASD and oxidative stress. Oxidative stress is a physiological state caused by an imbalance between the production and clearance of Reactive Oxygen Species (ROS) and Reactive Nitrogen Species (RNS) in the body (Chauhan and Chauhan, 2006; Frustaci et al., 2012). This imbalance is thought to be closely related to the pathological mechanisms of various neurodevelopmental disorders.

Glutathione is a crucial antioxidant in the human body, playing a significant role in maintaining cellular redox balance. Abnormal glutathione metabolism is considered an important component of the pathological mechanism of autism. Studies have found that the total amount of glutathione, including reduced glutathione (GSH) and oxidized glutathione (GSSG) in patients with autism is often reduced. Specifically, the level of GSH is significantly reduced, while the level of GSSG is relatively elevated. This results in a decreased GSH/GSSG ratio, reflecting an imbalance in the cellular redox state. Changes in this ratio are considered an important marker of oxidative stress (James et al., 2004; Frustaci et al., 2012; Rose et al., 2012). Research on glutathione-related enzymes, such as glutathione synthase (GSS), glutathione reductase (GR), and glutathione peroxidase (GPx), reveals their key roles in glutathione synthesis, regeneration, and functional maintenance (Frustaci et al., 2012; Rose et al., 2012; Gu et al., 2015). In autism patients, enzyme activities, such as reduced GPx and GR activities, are often abnormal, resulting in impaired GSH regeneration and exacerbated oxidative stress.

In terms of identifying and validating oxidative stress markers, researchers have identified markers that show significant changes in individuals with autism (Chauhan et al., 2012). These specific biomolecules, including Malondialdehyde (MDA), 4-hydroxynonenal (4-HNE), and protein carbonyl, indicate the degree of lipid peroxidation and protein oxidation, respectively. Changes in the concentrations of these markers in blood, urine, and cerebrospinal fluid are thought to reflect the oxidative stress status of individuals with autism, providing important indicators for clinical diagnosis and disease assessment (Frustaci et al., 2012). Advanced technologies such as High-Performance Liquid Chromatography (HPLC), Gas Chromatography–Mass Spectrometry (GC–MS), and Enzyme-Linked Immunosorbent Assay (ELISA) have made the quantitative detection of these markers more accurate and sensitive, providing reliable data for studying oxidative stress in autism (Yao et al., 2006; Chauhan et al., 2012; El-Ansary and Al-Ayadhi, 2012).

Evaluating antioxidant enzyme activity is another critical aspect. Antioxidant enzymes, including Superoxide Dismutase (SOD), Catalase (CAT), and Glutathione Peroxidase (GPx), play a crucial role in scavenging reactive oxygen species and maintaining redox balance in the body (Chauhan et al., 2012; Frustaci et al., 2012). Studies have found that the activity of these enzymes often changes in individuals with autism, manifesting as either decreased or abnormally increased activity. Assessing changes in antioxidant enzyme activity provide further insights into the oxidative stress status of individuals with autism.

Mitochondrial dysfunction leads to energy metabolism disorders and excessive production of reactive oxygen species, triggering oxidative stress (Rossignol and Frye, 2012a; Rossignol and Frye, 2014). In autistic patients, research on mitochondrial dysfunction focuses on mitochondrial DNA, bioenergetics, and mitochondrial autophagy.

Mitochondria have a genome independent of nuclear DNA and are more susceptible to mutations. Many studies have found multiple mutations and variations in mitochondrial DNA in individuals with autism, affecting mitochondrial function and leading to energy metabolism disorders. Studies have validated that point mutations, deletions, and copy number variations in mtDNA are all associated with the severity of autism symptoms (Giulivi et al., 2010; Rossignol and Frye, 2012a).

Additionally, mitochondria in autistic individuals often exhibit abnormalities in ATP synthesis, redox response, and electron transport chain (ETC) function. Studies on mitochondrial function in cells and tissues of autistic individuals have found significant differences in oxidative phosphorylation efficiency, membrane potential, and ROS production. These dysfunctions may lead to insufficient energy supply to nerve cells, affecting neurodevelopment and function, thereby increasing the risk of autism (Giulivi et al., 2010; Rossignol and Frye, 2012a).

The relationship between mitochondria and autophagy has also garnered research attention. Autophagy is a crucial process for cells to clear damaged mitochondria and maintain intracellular stability. Numerous studies have found that autophagy function in autistic individuals may be impaired, resulting in the accumulation of damaged mitochondria, which produce more ROS and exacerbate oxidative stress (Tang et al., 2014).

The keyword “brain” has become the second most prominent keyword and the second-largest cluster. The timeline of keyword clustering shows that this is currently the hottest topic, indicating that brain-related research will undoubtedly continue to be a major focus in the future. Current research on the brain mechanisms of autism primarily employs techniques such as functional magnetic resonance imaging (fMRI), structural magnetic resonance imaging (sMRI), electroencephalography (EEG), magnetoencephalography (MEG) and event-related potentials (ERP) to uncover neuroimaging biomarkers for autism (Calhoun et al., 2008; Schumann et al., 2010; Abraham et al., 2017; Gilmore et al., 2018). These methods will continue to evolve, enhancing our understanding of both the structure and function of the autistic brain.

Brain research may be developed from several aspects, including deepening our understanding of brain structure, with an emphasis on connectivity and functional integration within brain networks. Most importantly, it will aim to establish a multimodal model to comprehensively examine the brain mechanisms in individuals with autism.

Current neuroimaging research has provided a substantial amount of preliminary data on the brain structure and function of autism patients, which still needs to be expanded. Using structural magnetic resonance imaging (sMRI), voxel-based morphometric analysis has found that individuals with autism exhibit increased white matter volume in the temporal and frontal lobes, as well as enlarged gray matter volumes in regions such as the postcentral gyrus and superior temporal gyrus. Cortical folding analysis suggests an initial increase in cortical folds in some brain areas, which decreases with age (Jiao et al., 2010; Ecker et al., 2015). Functional MRI (fMRI) reveals that autistic individuals show heightened activity in brain regions related to visual perception and pattern recognition, but lower activity in areas related to social and emotional processing (Di Martino et al., 2009; Schipul et al., 2011). Additionally, resting-state fMRI analysis of striatal connectivity reveals changes, such as decreased connectivity between the striatum and prefrontal cortex, potentially disrupting normal neural communication (Yerys et al., 2015). While there have been significant findings regarding the brain regions involved in autism, the brain remains a complex and mysterious organ. Investigating the potential underlying brain mechanisms in individuals with autism is an ongoing challenge and an exciting area of research. Understanding these mechanisms could provide deeper insights into the condition and open new avenues for more effective treatments and interventions. This remains a key interest for researchers in the field.

However, a major unresolved issue is that current research primarily relies on single techniques, each providing only a partial view of the brain. The challenge lies in integrating multimodal data and combining different technologies to more comprehensively and accurately reveal the neurobiological mechanisms of autism. Differences in the spatial and temporal resolution, as well as measurement indicators, between modalities highlight the need for new data analysis methods and models to achieve this integration. Developing such approaches will be a crucial breakthrough in autism brain research.

Autism does not exist in isolation; many autism patients also suffer from other psychiatric disorders such as attention deficit hyperactivity disorder (ADHD), anxiety disorders, depression, and epilepsy (Gillberg and Billstedt, 2000; Simonoff et al., 2008; Joshi et al., 2013; Lai et al., 2019). This comorbidity phenomenon not only increases the complexity of diagnosis and treatment but also poses greater challenges for patients and their families. Therefore, studying the comorbidity of autism with other psychiatric disorders is important for understanding the pathological mechanisms of autism and developing effective treatment strategies.

The most cited article “Identification of risk loci with shared effects on five major psychiatric disorders: A genome-wide analysis,” explores specific genetic variants’ effects on autism, attention deficit disorder, bipolar disorder, major depression, and schizophrenia. It identifies chromosome 3p21 and10q24, along with single nucleotide polymorphisms (SNPs) within two L-type voltage-gated calcium channel subunits, CACNA1C and CACNB2, as potential biomarkers for these disorders (Consortium C.-D.G.o.t.P.G, 2013). This underscores the research focus on common biomarkers across different psychiatric conditions. Therefore investigating shared biomarkers of autism in conjunction with other psychiatric disorders represents a significant research trend and hotspot moving forward.

The presence of multiple comorbid conditions makes accurate diagnosis challenging. For instance, when autism co-occurs with Attention Deficit Hyperactivity Disorder (ADHD), the typical ADHD symptoms of inattention, hyperactivity, and impulsivity can overlap and become confounded with autism’s own traits, such as distractibility and stereotyped behaviors. This overlap makes it difficult for clinicians to distinguish between the two based solely on external behavioral manifestations, often necessitating more comprehensive and in-depth assessment methods for accurate diagnosis. The high prevalence of comorbidity between autism and other psychiatric disorders suggests that these disorders may share common pathological bases. For example, both autism and ADHD involve brain function abnormalities, particularly in executive function, attention, and behavioral control (Castellanos and Proal, 2012). By studying these comorbidity phenomena, current research has begun to explore common neurobiological mechanisms between autism and other psychiatric disorders (Crespi et al., 2010; Consortium C.-D.G.o.t.P.G, 2013; Smaga et al., 2015).

Additionally, the challenges posed by comorbidities are even more pronounced in treatment. The comorbidity research can improve the diagnosis and treatment of autism. For patients with both autism and other psychiatric disorders, single-treatment methods are often ineffective. Understanding the comorbidity mechanisms of these disorders allows doctors to develop multi-level and multi-target treatment strategies, enhancing treatment efficacy (Matson and Goldin, 2013; Vasa and Mazurek, 2015).

Future research on autism biomarkers will focus on the comorbidity of autism with other psychiatric disorders, revealing common pathological mechanisms and promoting the development of new diagnostic and therapeutic methods, improving early detection and intervention capabilities for these complex disorders. Through comorbidity research, we hope to better understand and address autism and its associated psychiatric disorders (Loth et al., 2016; Frye et al., 2019).

In recent years, an increasing number of studies have found metabolic abnormalities in autism patients (Adams et al., 2011a; Adams et al., 2011b; Kang et al., 2018). These abnormalities involve various metabolic pathways, including energy metabolism, amino acid metabolism, lipid metabolism, and oxidative stress (Rossignol and Frye, 2014; Frye et al., 2019).

Although genetic factors play a significant role in the development of autism, the impact of environmental factors cannot be overlooked. Metabolic processes act as a crucial interface between genes and the environment. Investigating metabolic biomarkers in children with autism provides deeper insights into the condition’s pathogenesis. For example, metabolic biomarkers can reveal abnormalities in processes such as cellular energy metabolism and neurotransmitter synthesis. These disruptions may serve as potential underlying causes of impaired brain development in children with autism.

As mentioned above, some studies have found mitochondrial dysfunction in autism patients, leading to energy metabolism disorders (Rossignol and Frye, 2012b; Morris and Berk, 2015); other studies have shown elevated oxidative stress levels in autism patients, which may be related to neuronal damage (James et al., 2004; Ng et al., 2008; Frustaci et al., 2012). By studying these metabolic markers, scientists can better understand the biological basis of autism. Metabolic marker research provides an objective biological testing method, improving diagnostic accuracy and early detection capabilities. For instance, scientists can analyze metabolite levels in the blood, urine, or other body fluids of autism patients to find specific metabolic markers, assisting in the diagnosis of autism and providing targeted treatment (Gabriele et al., 2014; Rossignol and Frye, 2014).

In analyzing the top 100 highly-cited studies, we found that autism biomarker research, especially concerning genetic and neuroimaging biomarkers in autism patients, heavily relies on the advancement and application of cutting-edge technologies. Current autism biomarker detection predominantly involves established techniques like fMRI, sMRI, EEG, and MEG (Calhoun et al., 2008; Schumann et al., 2010; Abraham et al., 2017; Gilmore et al., 2018). With innovations in artificial intelligence and machine learning technologies, we believe the application of new technologies is a promising direction, particularly in the research of the diagnosis of autism biomarkers.

Artificial intelligence (AI) and machine learning demonstrate considerable application value in the diagnosis of autism biomarkers. Based on the precision and automation of AI and machine learning, these technologies can be applied to large-scale whole genome sequencing data to uncover complex genetic variation patterns associated with autism (Nisar and Haris, 2023; Maqsood et al., 2024). They can also be used to process vast amounts of MRI and fMRI data to capture and analyze complex relationships in brain structural data. These approaches are far more challenging with traditional manual methods, but AI and machine learning not only save time and effort but also provide higher accuracy (do Rêgo and Araújo-Filho, 2024; Maqsood et al., 2024; Wankhede et al., 2024).

Regarding genetic biomarkers diagnosis, autism-related genes such as SHANK3, NRXN1, and CNTNAP2 exhibit significant complexity and variability, with mutations appearing as single nucleotide polymorphisms (SNPs), small insertions or deletions (Indels), and copy number variations (CNVs). Deep learning algorithms like Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) can automatically extract more refined features from large datasets, making them more efficient and accurate than manual analysis (Nisar and Haris, 2023; Maqsood et al., 2024). From a brain structural perspective, AI and machine learning can analyze vast datasets through clustering methods to identify abnormalities in specific brain regions and autism-specific structural features (Khodatars et al., 2021; Bahathiq et al., 2022). For example, clustering analyses of gray matter volume data in autism patients and control groups can reveal the relationship between reduced frontal lobe gray matter volume and executive function deficits, as well as abnormal temporal lobe gray matter volume and social cognitive impairments.

Additionally, AI and machine learning provide the opportunity for dynamic tracking of biomarkers at different time points and stages, allowing for the creation of evolving pathways for biomarkers. By integrating data from other biological markers such as metabolic and genetic markers, the accuracy of disease diagnosis and prediction can be improved, providing a basis for timely adjustments to treatment plans and enhancing treatment outcomes for autism patients.Thus, we believe that the use of emerging technologies will undoubtedly become a focal point in future research.

We analyzed key areas of biomarker research in childhood autism, emphasizing its importance for early diagnosis, detection, and predictive studies. However, the continuous changes in autism biomarkers throughout development remain unclear. Current research often focuses on childhood and adulthood biomarkers separately, leaving gaps in understanding their evolution and longitudinal dynamics. Autism, as a lifelong neurodevelopmental disorder, involves symptoms and pathological characteristics that vary with age and developmental stage (Seltzer et al., 2004; Zwaigenbaum et al., 2013).

Longitudinal studies, which repeatedly measure the same group of individuals at multiple time points, can offer a more comprehensive and in-depth understanding. Through long-term tracking, researchers can observe changes in biomarkers during disease progression, understand their role in the disease course, and identify early predictors of autism (Singer, 2003; Ecker et al., 2015). However, there is currently a shortage of longitudinal studies, especially those that span the lifespan, investigating autism biomarkers.

The dynamic progression of autism biomarkers is indeed essential to understanding the disorder comprehensively. For instance, studies have identified a strong correlation between mutations in the FMR1 gene and autism (Hagerman et al., 2018; Richter and Zhao, 2021). However, longitudinal research on how such genetic anomalies influence autism phenotypes across developmental stages remains limited. While we know that mutations in FMR1 lead to the loss or dysfunction of its encoded protein, the precise mechanisms by which this anomaly affects brain development and behavioral symptoms from infancy to adulthood remain unclear (Verkerk et al., 1991).

Similarly, the dynamic interplay between multiple autism-related genes (e.g., SHANK3, PTEN) and their joint regulation of brain development and behavior over time has yet to be thoroughly investigated (Monteiro and Feng, 2017). Additionally, while glutamate level abnormalities have been observed in individuals with autism, studies have not examined how these levels change with age. For example, elevated glutamate levels detected in early childhood are not yet understood—whether they represent a transient developmental fluctuation or a persistent neurodevelopmental factor influencing adult social and cognitive symptoms remains unknown (Robertson and Baron-Cohen, 2017; Horder et al., 2018).

The brain’s plasticity underscores the significance of dynamic changes and developmental processes. While exploring brain mechanisms at a single time point, such as during childhood, is valuable, understanding the trajectory of biomarker evolution under this plasticity is crucial. Future research should prioritize studying the dynamic progression of biomarkers alongside current cross-sectional approaches. Although such longitudinal studies are time-and resource-intensive, they hold immense potential for deepening our understanding of autism mechanisms and advancing effective treatments.

Current research trends indicate a trend toward studying the comorbidity of autism with other psychiatric disorders (Gillberg and Billstedt, 2000; Simonoff et al., 2008; Joshi et al., 2013; Lai et al., 2019). However, due to the unique nature of autism, we believe that specific research on autism biomarkers is important. The specific study of autism biomarkers is crucial for more accurate diagnosis of the disorder itself.

To understand the core pathological mechanisms of autism, while the study of comorbid biomarkers can shed light on their association with other disorders, it cannot replace the need to explore autism’s intrinsic pathology. Comorbid biomarkers may blur diagnostic boundaries, whereas autism-specific biomarkers—such as mutations in SHANK3, NRXN1, and CHD8—are directly tied to autism, but may also be found in other conditions like intellectual disabilities and ADHD (Zoghbi and Bear, 2012; De Rubeis and Buxbaum, 2015). Therefore, they cannot be solely used for initial diagnosis. From a personalized treatment standpoint, autism-specific biomarkers provide targeted therapeutic strategies, enhancing the potential for improving core symptoms and quality of life.

Therefore, future research should not only explore the comorbidity of autism with other psychiatric disorders but also focus on advancing the specific study of autism biomarkers. This will enable more precise diagnosis of autism’s core features, uncover its unique pathogenesis, and guide personalized treatment strategies. Future research should aim to develop more specific and sensitive biomarkers, which may require the integration of multiple biological data, such as genomic, metabolomic, and imaging data. The combined application of multiple biomarkers could enhance diagnostic specificity and sensitivity, despite increasing research and application complexity. Additionally, large-scale validation studies are needed to ensure the reliability of these biomarkers in different populations and clinical settings. The application of new technologies such as machine learning and artificial intelligence may help identify more precise diagnostic patterns from complex data.