Network analysis in the study of schizophrenia symptoms: a narrative review

Full Text

INTRODUCTION

At present, there is a lack of methods for the comprehensive study of complex phenomena, such as economic processes, the functioning of neural networks, the interaction between various genes, etc. [1]. Conclusions based on identifying the properties of their simpler components, although useful, cannot provide an understanding of the subject as a whole [1]. Mathematical models of data-driven complex systems offer a modern perspective and are rapidly evolving into a new discipline called network science [1]. Since the late 20th century, network theory has sought to understand the origins and characteristics of structures connecting components of various complex systems [1]. Owing to a strong empirical base and a diverse set of analytical and algorithmic tools, network theory has become indispensable in the analysis of complex phenomena [1]. It has also been applied in psychiatry, where there is currently a conceptual shift that reflects the nature of mental disorders [2].

The categorical approach, which is based on classifying symptoms into rigid categories, has already been recognized as inconsistent with the accumulated clinical and scientific experience [2]. A dimensional approach is being introduced, in which symptoms are represented as continua of mental disorders [2]. As part of this approach, complex computational techniques are actively used and new models of mental disorders are being developed [3, 4]. Thus, the Research Domain Criteria (RDoC) concept suggests that symptoms of mental disorders arise from a combination of individual differences in cognitive and emotional processes [5]. Another concept, the Hierarchical Taxonomy of Psychopathology (HiTOP), is based on factor analysis methods for studying patterns of symptom co-occurrence across a broad transdiagnostic field of mental disorders [6].

Network theory is based on an alternative view of mental disorders as a network of interacting symptoms [7]. Given individual differences in the severity and dynamics of disorders in psychiatric patients, as well as in their interactions [7], the authors of the network approach consider it unproven and unlikely that symptoms are a consequence of mental disorders [7, 8]. When a network of symptom interactions is identified, reflecting individual differences in disease trajectory, the concept of a mental disorder as the cause of symptoms becomes redundant [7]. Thus, the authors of the network approach argue that significant causal relationships between symptoms actually constitute the "material" from which mental disorders are made [8].

The network approach offers its own conceptualization of the pathogenesis of mental disorders [9]. It has been hypothesized that the symptoms of the disorder are not a consequence of a common etiological factor, but rather mutually trigger one another [7–9]. The network of symptom interactions is a dynamic system in which elements can change over time, influence other elements, and be influenced by them [10]. Factors that potentially contribute to the development of the disorder or modify its course, such as psychological trauma, chronic stress, inflammatory processes, and others, may be considered part of the "external field" [9, 10]. Elements of this so-called "external field" may activate certain symptoms in the network [9, 10]. These symptoms, may in turn, trigger the activation of other connected symptoms, and so on [9, 10]. Symptoms may form feedback loops that can sustain symptom activation even after the external factors have subsided; phenomenologically, this persistent activation is recognized as a mental disorder [9].

Within the network approach, the issue of comorbidity in psychiatry is viewed in a different light: it can no longer be explained simply as a relationship between two disorders or as the result of shared neurobiological dysfunction [8]. Instead, causal relationships between symptoms represent pathways connecting various disorders, such as through bridge symptoms that are part of these disorders [8, 11]. Within this approach, the boundaries between the diagnostic categories become blurred [11].

The network approach implies a shift in therapeutic goals: from this perspective, therapy should include modifying the symptoms network [9]. The optimal targets for therapeutic intervention are considered to be the symptoms, the causal relationships between them, and the elements of the "external field" [8, 9]. Symptoms with the greatest number of connections to other symptoms are prioritised as treatment targets, since they may potentially activate a larger number of impairments [12]. Such symptoms are called central [12]. Bridge symptoms linking different pathological domains are also considered important, as they are presumed to play a key role in network stability [9].

The principal tool for identifying and studying symptoms networks is network analysis [4]. At present, the most widely used model for constructing psychopathological networks from continuous data is the partial correlation network [13]. Within this network, the nodes represent observed variables and the edges represent the partial correlation coefficients between two variables after controlling for all other variables in the dataset [14]. Such networks are usually estimated using regularization methods from machine learning, which limit model flexibility and thereby help prevent fitting to noise in the training sample. Regularization involves estimating a statistical model with an additional penalty for model complexity, resulting in the elimination of potentially spurious edges [13, 14]. The nodes in such a network are typically arranged using the Fruchterman-Reingold algorithm, which places nodes with the highest correlation coefficients closer together and thereby enhances the interpretability of the graph [13].

Network node centrality is assessed using several metrics: strength and expected influence, which are of greatest interest in the analysis of psychopathological networks, as well as closeness and "betweenness" [15, 16]. Symptom networks usually form clusters; that is, groups of variables more strongly connected to one another than to the remaining elements of the network [13]. The relationships between clusters can be assessed by identifying bridge symptoms between them; these have the same metrics as central symptoms [17].

To study causal relationships between network components, directed acyclic graphs have been introduced into network analysis; in these graphs, all connections between nodes are directed and represented by arrows reflecting causal influence and there are no feedback loops [8]. For constructing networks based on binary data, Ising models are used, in which pairwise interactions are examined between variables that can exist in one of two distinct states [18]. To compare different patient groups in the context of network analysis, the network comparison test is used [19]. To address the issues of statistical significance and reproducibility in network analysis, methods for assessing sample size and network stability have been developed [13].

Network analysis is widely used in the study of mental disorders. Symptom networks have been constructed for depressive disorders [20, 21], anxiety disorders [22], post-traumatic stress disorder [23, 24], eating disorders [25], and autism spectrum disorders [26]. Network analysis is an equally common tool for studying psychotic disorders, particularly schizophrenia [15].

Over time, the contribution of schizophrenia to the global disease burden has been steadily increasing, particularly in years of working life lost [27]. Only a small proportion of patients demonstrate a good response to antipsychotic medications [28]. In patients with schizophrenia, as well as in patients with other schizophrenia spectrum disorders, pronounced long-term impairments in social functioning are observed [29]. High mortality among patients with schizophrenia remains a serious problem [30]. The economic burden of schizophrenia is also substantial, with average costs ranging from 0.3% to 0.6% of gross domestic product [31].

For schizophrenia patients, there is a need to develop and implement pharmacological and psychosocial treatments aimed at improving functioning and quality of life [30]. The network approach may offer a solution to this problem by proposing new targets for therapeutic interventions [9].

A number of review studies have been conducted on the use of network analysis in the study of schizophrenia. Thus, in one systematic review, positive symptoms were proposed as central, while negative symptoms served as a bridge between positive symptoms and depression [32]. Childhood psychological trauma played an important role in the network, and after successful therapy, the interconnectedness of symptoms within the network increased [32]. In another systematic review on network analysis of psychotic experiences, these experiences were not central in the network; however, childhood trauma exerted a pronounced influence on the network, being directly associated with positive symptoms or acting as a "bridge" between them and other disturbances [33].

The authors of another systematic review proposed cognitive metrics, as well as social and occupational functioning, as central nodes [34]. Affective symptoms, when included in network analysis, also showed a high level of centrality in the network, while negative symptoms played a less important role in the network compared to positive symptoms [34]. The results of network analysis in the study of schizophrenia were included in a systematic review exploring the assessment of cognitive functions in various mental disorders [35]. The authors of that review identified cognitive impairments as highly central across different psychopathological networks; processing speed and working memory proved to be particularly important variables [35].

At the same time, a number of important aspects of schizophrenia from the perspective of network theory remain unexplored. These include, for example, the role of disorganization phenomena in the network of schizophrenia symptoms, the influence of insight on the network, and the side effects of therapy when these characteristics are incorporated into network analysis. The relationship between symptom networks constructed on data from patients with schizophrenia and data from healthy individuals also remains unclear.

The study aimed to summarize the results of a network analysis of schizophrenia symptoms.

METHODS

Eligibility criteria

Inclusion criteria:

• original studies containing the results of network analysis based on data from psychiatric patients, presented in the form of partial correlation networks;
• study participants were patients with schizophrenia in accordance with the criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV and DSM-V or the International Classification of Diseases (ICD)-10 and ICD-11;
• publication date of the included study was between January 2019 and August 2025;
• networks presented in the studies must include symptoms of schizophrenia;
• articles were published in English.

Exclusion criteria:

• studies exploring network meta-analyses;
• studies exploring functional brain networks;
• studies exploring molecular interaction networks;
• studies exploring semantic networks;
• studies exploring social networks;
• systematic reviews;
• other articles that did not meet the inclusion criteria (exploring pharmacological, neural, microbiome networks, etc.) were excluded.

Information sources

The literature search was conducted in the PubMed (MEDLINE) database between October 2024 and September 2025. The search period covered publications from January 1, 2019, to August 31, 2025, in order to include the most relevant studies on this topic.

Search strategy

The search was conducted using the following keywords in the titles and abstracts of articles: "Network analysis" and "Schizophrenia". PubMed search term looked as follows: ("network"[Title/Abstract] AND "analysis"[Title/Abstract] AND "schizophrenia"[Title/Abstract]) AND ((medline[Filter]) AND (2019/1/1:2025/8/31[pdat])).

Selection process

Each publication was identified through a manual search. Some of the studies identified during the screening phase were excluded from further analysis because they did not meet the inclusion criteria (see the "Eligibility criteria" section). The selection process was performed in stages; the sequence is presented in Figure 1.

 

Figure 1. Literature search and selection process.

Source: Filippov, Petrova, 2026.

 

Data analysis

The publications identified and selected for analysis were studied in their entirety. The information from the selected sources was summarized and divided into three areas:

1. The role of separate impairments in the schizophrenia symptoms network.
2. The role of functioning, quality of life, and other patient characteristics in the schizophrenia symptoms network.
3. Transdiagnostic studies (including research examining links between schizophrenia and other disorders, comparing symptoms networks of schizophrenia, other mental disorders, and healthy individuals).

RESULTS

The role of separate impairments in the schizophrenia symptoms network

A global evaluation of the schizophrenia symptom network, visualized using the Fruchterman-Reingold algorithm, can provide researchers with valuable insights into the interaction structure among the study variables [13]. For instance, when mapping the network of parameters assessed by the Brief Negative Symptom Scale (BNSS) in patients with schizophrenia, the nodes cluster into 4–5 domains corresponding to avolition / asociality, anhedonia, alogia, and blunted affect, thereby reflecting the latent structure of negative symptoms [36].

Network analysis also facilitates the revision of established psychometric tools. For example, the Positive and Negative Syndrome Scale (PANSS), which traditionally categorizes symptoms into positive, negative, and general psychopathology subscales, is widely considered inadequate for reflecting the dimensional structure of schizophrenia [37]. Network analysis of patients experiencing a first episode of psychosis evaluated with the PANSS demonstrates that symptoms cluster into distinct domains corresponding to positive symptoms, negative symptoms, cognitive impairment/disorganization symptoms, and excitement/hostility [37].

Cognitive deficits have been identified as central nodes within the schizophrenia symptom network [38–40], with attention impairment [41] and reduced information processing speed being particularly prominent [42, 43]. Negative symptoms are also commonly identified as highly central across various studies [44, 45]. Depending on the study, the most central nodes have included motivation and pleasure deficit [38], such as avolition [46, 47], aspontaneity [48], anhedonia [49, 50], asociality [51, 52], poor rapport [41, 53], and motor retardation [45, 54], as well as expressive deficits manifested as alogia [47, 48] and blunted affect [48, 53, 55]. Furthermore, depressive and anxiety symptoms have exhibited high centrality indices in numerous studies [56–58]. While some authors have identified delusions [55, 59, 60], paranoia [61], and hallucinations [60] as central symptoms, other studies have found that positive symptoms exhibit the lowest centrality values [49, 51, 54]. Disorganization symptoms have also emerged as central in several investigations [38, 62, 63]. Moreover, some studies have proposed uncooperativeness, impaired abstract thinking [62], disrupted interpersonal relationships [46, 60], and the presence of unusual sensory experiences [61] and unusual thought [54] content as central features.

Because symptoms networks map the structure of symptom interactions, they allow for the identification of correlations between different phenotypic manifestations of schizophrenia [64, 65]. For instance, an association has been detected between diminished emotional expression and both disorganization symptoms [64] and information processing skills [65]. However, elucidating the relationships between distinct, non-adjacent domains of the disorder requires the identification of "bridge symptoms" within the network [17]. For example, impaired visual learning has been shown to bridge neurocognitive deficits and disorganization symptoms, avolition, and the metacognitive domain [38]. Affective symptoms are linked to positive symptoms via sleep disturbances [44], suspiciousness, and stereotyped thinking, whereas they are connected to negative symptoms through fatigue and observable sadness (depressed mood) [66]. Emotional distress serves as a key bridge symptom between positive and negative symptoms, as well as between dissociative phenomena and PANSS-measured pathology [67]. In other studies, conceptual disorganization [41, 48, 55], stereotyped thinking, depression [41, 55], hostility [48, 55], and impaired abstract thinking have also frequently emerged as prominent bridge symptoms between network clusters [48, 68].

The role of functioning, quality of life, and other patient characteristics in the schizophrenia symptoms network

To investigate the impact of schizophrenia symptoms on functioning and quality of life (QoL), these variables can be integrated directly into network analysis. When included, QoL [65, 69], global functioning [40, 44, 70], or specific functional subdomains, such as functional capacity [64, 71, 72], activities of daily living (ADL) [46, 64, 71], and social skills [61, 72], frequently occupy central positions within the network.

Quality of life is primarily linked to negative and affective symptoms [69]; directed acyclic graphs (DAGs) have established a causal influence of these symptoms on QoL [69]. Overall functioning correlates with impairments in social cognition [51, 64, 72] and metacognition [38], neurocognitive deficits [51, 64], negative symptoms [38, 51, 72], and disorganization phenomena [38]. Social skills are associated with suspiciousness, depression [39], negative symptoms [39, 72] (specifically motivation and pleasure deficits [64, 73]), as well as social cognitive deficits [72, 74] and neurocognitive impairments [39]. Vocational skills correlate with deficits in the motivation and pleasure domain [64]. Functional capacity is associated with social and neurocognitive performance, the severity of negative symptoms, and, most notably, diminished emotional expression [64, 72].

In one study, functional capacity acted as a bridging node connecting social and neurocognitive deficits to global functioning [71]. In another study, the relationship between functional capacity, neurocognitive deficits, and negative symptoms was mediated by disorganization phenomena. Additionally, some investigations have identified social functioning [50] — specifically social avoidance — as a bridge between different network domains [57].

Network analysis can also incorporate other patient characteristics to explore how schizophrenia symptoms interact with developmental risk factors, course of illness, and treatment-emergent adverse events [43, 56, 75].

Among developmental risk factors, childhood trauma is frequently examined in network analyses [76, 77]. In one such study, childhood abuse emerged as a central node in the network, alongside measures of resilience [76]. Specifically, childhood emotional abuse was associated with the severity of positive and negative symptoms [76], depression, and guilt [77]. Sexual abuse was linked to guilt, while physical abuse was associated with negative symptoms [77]. Childhood neglect was associated with disorganization symptoms, depression [76], and negative symptoms [76, 77]. Body image disturbances acted as a bridge symptom linking childhood trauma to positive symptoms and depression [76]. Emotional distress may serve as a bridge between childhood trauma and both positive and negative symptoms [77]. Furthermore, the symptoms described above are associated with aspects of trauma, abuse, and neglect through the mediating effects of social functioning, neuroticism, depression, and resilience [76]. Additionally, when studying risk factors for schizophrenia, smoking was found to be associated with the domains of cognitive impairment and positive symptoms [43].

Some researchers focus on the level of insight when performing network analyses of schizophrenia symptoms. Thus, in several studies, the level of insight was identified as one of the central nodes of the network [56, 78], and in one study, it acted as a bridge symptom linking different psychopathological domains [68]. A number of studies have demonstrated an association between clinical insight and quality of life [78, 79], adverse extrapyramidal symptoms [79], self-esteem [78, 80], guilt, morning depression, and suicidal ideation [80]. A directed acyclic graph demonstrated that the level of insight causes depression symptoms [80]. A study of subjective well-being in patients with schizophrenia demonstrated an association with obsessions and compulsions, depressive symptoms, and cognitive deficits [81].

Among the adverse extrapyramidal effects of schizophrenia treatment, parkinsonism was associated with domains of cognitive impairment, positive and negative symptoms [82], while expressive deficit played the greatest role in the symptom network [83]. One study also found an association between clozapine use and the development of abulia [52].

Investigations of immunological markers in schizophrenia have revealed associations of interleukin-6, tumor necrosis factor-α, and interferon-γ levels with depression and suspiciousness [84].

Studies examining the psychophysiological characteristics of schizophrenia have also shown a relationship between cognitive impairment and oculomotor parameters [85].

Network comparison tools [19] help to conduct a comparative assessment of patients with schizophrenia stratified by sex, phase of illness, and socioeconomic conditions [47, 59, 80].

Researchers are particularly interested in comparing networks reflecting different aspects of the course of schizophrenia. An increase in the number and strength of connections between symptoms has been established during exacerbations of the disease, whereas during the long-term course of the illness or after achieving remission, network becomes more sparse [44, 68, 71]. During exacerbations of schizophrenia, negative symptoms were more strongly connected with the other nodes of the network than in remission, when emotional and volitional deficit became more prominent [59]. In the latter case, the relative isolation of negative symptoms from other disturbances and from one another may indicate their primary nature [59]. It was also found that as schizophrenia progresses, the association between functioning parameters and negative symptoms becomes stronger [85]. Upon achieving remission, short-term memory is directly linked to social skills [39], and as the duration of the disease increases, metacognitive impairments become associated with symptoms of disorganization [38]. At the same time, some researchers argue that the structure of the schizophrenia network does not undergo significant changes over time [86]. The main symptoms of treatment-resistant schizophrenia were associated with social functioning through domains of positive and negative symptomatology; in contrast, treatment-responsive schizophrenia is more strongly associated with general symptoms of the PANSS [87]. In addition, the association of schizophrenia symptoms with childhood trauma was stronger in cases with treatment-resistant disease than during the first episode [88].

Network comparison is an important aspect of longitudinal studies as well as analysis of treatment effects. Symptom networks of schizophrenia reflecting the condition of the same group of patients at different points in time show little difference in structure, although the strength of connections between nodes often changes [36, 53, 71]. During clinical trials, the drug roluperidone¹ led to a decrease in the centrality of abulia in the network [89, 90]. The use of quetiapine managed to reduce symptoms of agitation, which played a central role in the network [48]. When comparing symptom networks reflecting the structure of schizophrenia among patients with different socioeconomic statuses and levels of involvement in medical care, no significant differences were found [80]. Alogia, blunted affect, and asociality were central symptoms in the network corresponding to the female patient group; alogia and abulia had higher centrality among male patients [47]. In patients who had committed violent acts, network nodes were less interconnected [70].

Transdiagnostic studies

Network analysis can be used to study the relationships between schizophrenia and other mental disorders. For example, in a study of schizophrenia comorbid with substance abuse, the use of illicit psychoactive substances acted as a central symptom and also provided a link between positive and negative symptoms and the level of patient functioning [44].

In a large cross-sectional transdiagnostic study, symptoms of depressive, anxiety, obsessive-compulsive, and psychotic disorders formed three clusters: general psychopathology, cognitive impairment, and use of illicit psychoactive substances [91]. Depressive and anxiety symptoms, as well as verbal memory and attention, were identified as central nodes in the network [91]. However, cannabis use, psychotic, and negative symptoms were found to be most closely associated with cognitive deficits [91].

When studying catatonia among patients with schizophrenia, bipolar affective disorder, depression, and somatic disorders, agitation, impulsivity, and perseveration played a central role in the symptom network [92].

In a study of the impact of childhood trauma on the development of schizophrenia, major depressive disorder, and bipolar affective disorder, researchers constructed a network including various types of trauma and psychopathological symptoms [93]. It was found that emotional abuse was associated with symptoms of depression, physical abuse with symptoms of agitation, sexual abuse with positive symptoms and disorganization, and childhood neglect with negative symptoms [93].

If symptoms from different diagnostic categories are not combined into one network but are studied separately, the resulting networks can be compared with each other [47, 52, 94]. For example, communication difficulties were found to be central in the symptom networks of both schizophrenia and autism [94]. Negative symptoms showed weaker interconnections in patients with schizophrenia compared to those with bipolar affective disorder and healthy individuals [47]. At the same time, the positive symptoms of bipolar disorder were found to be more strongly interconnected [52]. In addition, in patients with schizophrenia, depressive symptoms were less strongly associated with deficits in motivation and pleasure but were more strongly linked to positive symptomatology than in bipolar disorder [93]. In patients with schizophrenia (unlike major depressive disorder), the structure of depression was influenced by fatigue, sleep duration, appetite, concentration difficulties, and apathy [95]. The symptom network of major depressive disorder was characterized, in contrast to schizophrenia, by a stronger connection between symptoms of depression and disorganization [93]. No statistically significant differences (p = 0.396) were found between the network structures for schizophrenia and mood disorders, including negative and depressive symptoms, parkinsonism, and motor retardation [45]. Another study found that the symptoms networks of schizophrenia, schizoaffective disorder, and psychotic bipolar disorder were very similar, demonstrating a connection between depressive symptoms and negative symptoms, as well as a link between manic and productive symptoms [96].

Results were obtained by comparing networks based on symptoms of patients with schizophrenia and healthy individuals. For example, the network of disturbances in body perception, childhood trauma, and schizotypy had a similar structure in both schizophrenia and healthy individuals, but the presence of the disease was associated with stronger connections between nodes [97]. Similar results were obtained from a comparative analysis of symptom networks among patients with psychotic disorders, individuals with a hereditary risk, and a control group [61]. Similarities were also found when comparing symptom networks in patients with schizophrenia and their siblings [43]. When comparing networks of neurocognitive parameters in patients with schizophrenia and healthy individuals, information processing speed and working memory were the central nodes in the first and second group, respectively [42].

DISCUSSION

The network approach is a relatively new way of conceptualizing mental disorders, in which they are represented as networks of symptoms and their interactions [9]. In such networks, symptoms are influenced by "external field" factors and interact with each other, maintaining a symptoms network corresponding to a mental disorder [8]. Comorbidity in this approach arises through connections, i.e. "bridge" symptoms between networks representing different diagnostic categories [8, 11]. Within the framework of the network approach, therapy should include interventions targeting network elements: symptoms, the connections between them, and "external field" factors [8, 9].

Network analysis is both a method for studying symptom networks and a tool for applying the network approach. In this process, a partial correlation network is constructed, where nodes represent symptoms and edges represent associations between them [13]. Regularization methods from machine learning are used to construct such networks [13, 14]. In the vast majority of studies included in the review, the regularization method used was a combination of the Extended Bayesian Information Criterion and the Graphical Least Absolute Shrinkage and Selection Operator (EBICglasso). This combination helps to identify and exclude spurious edges from the network [13]. The tool is available as the EBICglasso function in the qgraph package for the R programming environment [98].

The visualization of symptom networks is carried out using the Fruchterman-Reingold algorithm, which places network nodes closer to each other when the correlation coefficient between them is higher [13]. In addition, in a number of studies included in the review [37, 45, 85], the authors applied Exploratory Graph Analysis (EGA), which allows for the identification of clusters formed in the symptom network based on an assessment of polychoric correlations using the "walktrap" algorithm [99]. This method is implemented in the EGA function in the EGA package in the R programming environment and is highly effective in determining the number of domains within the network [100].

An important aspect of network analysis is the search for central symptoms in the network [12]. The results of this analysis were also presented in many of the studies included in the review [38, 44, 56]. The network can be assessed based on the centrality of its nodes, calculated using the centrality function included in the qgraph package for the R statistical programming environment [98]. Identifying bridge symptoms that connect separate domains within the network is also of great importance [9]. For this purpose, the bridge function from the networktools package in the R statistical programming environment can be used [17]. In many of the studies included in the review, bridge symptoms were identified [48, 66, 67].

The use of directed acyclic graphs makes it possible to assess the orientation of causal relationships between individual elements of the network [8], as demonstrated in a number of the studies reviewed [69, 80, 93]. To construct such graphs, various functions from the bnlearn package in the R statistical programming environment are primarily used [101]. In one study [69], such a graph was constructed using Mendelian randomization, represented by the MRPC function in the corresponding package [102].

The review included longitudinal and transdiagnostic studies [48, 53, 91], as well as studies that conducted comparative analyses of different patient cohorts [47, 96, 97]. For these purposes, a network comparison test has been developed, implemented using the NetworkComparisonTest package for the R environment [19].

All results of network analysis require an assessment of accuracy and stability to enhance statistical reliability [14, 17, 19]. Before data collection begins, it is recommended to assess the sample size to ensure that it is appropriate for the expected network structure. After constructing the network, bootstrapping methods should be used to gain an understanding of the accuracy and stability of the network parameters and characteristics [14]. Tools for these operations are available in the bootnet package for the R environment [14].

None of the studies included in the review used network analysis based on binary data. To work with such data, the eLasso function from the IsingFit package in the R programming environment is used, which combines regularization methods with Ising models [18].

In the study of schizophrenia using network analysis, the most commonly identified central variables include neurocognitive deficits, affective symptoms, and negative symptoms [38, 44, 56]. Disorganization often acts as a bridge symptom between individual domains [41, 48, 55]. When patient functioning parameters are included in network analysis, they invariably become central in the network [40, 45, 70], and negative symptoms, as well as socio- and neurocognitive impairments, are most associated with them [51, 64, 72]. Childhood trauma has a complex impact on the development of schizophrenia, with different forms of abuse forming connections with specific domains of schizophrenic symptoms [76, 77]. The level of insight, which is associated with multiple symptoms of schizophrenia, also plays an important role within the symptom network [56, 78, 79]. Over the course of the illness, the overall network structure remains relatively stable, but connectivity between nodes decreases, and negative symptoms begin to exert a greater impact on functioning [36, 53, 71]. When psychotic symptoms are controlled, the symptom network also becomes more sparse [44, 68, 71]. Healthy individuals demonstrate a more fragmented network structure than patients with schizophrenia [61, 97]. Affective symptoms and cognitive impairment play the role of symptoms linking schizophrenia with other mental disorders [91].

The most promising therapeutic targets are the central symptoms, owing to their numerous connections with other network parameters [12]. According to the review results, central symptoms include negative symptoms, affective symptoms, and symptoms of cognitive deficit. The strong association between patients' functioning and both negative symptoms and cognitive impairments underscores their importance as key targets for medical interventions. Bridge symptoms, which connect symptom networks of different disorders or isolated domains within a sparse network, may also serve as potential therapeutic targets [17]. According to the review findings, such symptoms include cognitive and affective disturbances, which provide a link between schizophrenia and other psychiatric disorders, as well as disorganization phenomena, which act as bridges between separate clusters within the schizophrenia symptom network. Childhood psychological trauma influences the development of schizophrenia, and the level of insight into illness can modify its course, making these factors potential targets for psychosocial interventions.

This study is more descriptive than systematic in nature. The search was limited to publications indexed in MEDLINE and published between 2019 and 2025, which could have led to incomplete coverage of the topic. A common limitation of the studies included in this review is the heterogeneity of patient samples, variables analyzed, and clinical assessment methodologies.

CONCLUSION

Network analysis is a promising tool for investigating both schizophrenia and psychiatric disorders in general. The growing body of research utilizing this methodology demonstrates an increasing interest among investigators in identifying underlying pathways and interactions among clinical symptoms. As a core tool of the network approach in psychiatry, network analysis offers novel targets for therapeutic interventions and helps advance a new conceptualization of mental disorders. The findings of network analyses in patients with schizophrenia underscore the significance of negative, cognitive, and affective symptoms, as well as disorganization, within the structure of schizophrenia. The findings demonstrate a close relationship between patient functioning and most symptoms in the network. Using network analysis, specific patterns in the dynamics of the schizophrenia symptom network have been established, characterized by a weakening of associations between impairments after achieving remission and during the long-term course of the disease. Utilizing this tool, affective and cognitive disturbances emerge as the primary candidates mediating the link between schizophrenia and other psychiatric disorders. Based on the review findings, affective and negative symptoms, disorganization phenomena, cognitive deficits, as well as the effects of childhood trauma and insight into illness level were proposed as potential targets for therapeutic interventions.

Article history

Submitted: 7 Nov. 2025

Accepted: 8 Jul. 2026

Published Online: 31 Aug. 2026

Authors' contribution: Ivan Filippov: conceptualization, data curation, writing the original draft. Natalia Petrova: conceptualization, review and editing the original draft, supervision.

Funding: The research was carried out without additional funding.

Conflict of interest: The authors declare no conflicts of interest.

Generative AI use statement: Nothing to disclose.

¹ It is not approved in the Russian Federation.

About the authors

Ivan A. Filippov

Saint Petersburg State University

Author for correspondence.
Email: filippov20121@gmail.com
ORCID iD: 0000-0001-9307-919X
SPIN-code: 8154-1925
ResearcherId: LKJ-6132-2024

Postgraduate Student, Department of Psychiatry and Narcology, Medical Institute

Russian Federation, Saint Petersburg

Nataliia N. Petrova

Saint Petersburg State University

Email: filippov20121@gmail.com
ORCID iD: 0000-0003-4096-6208
SPIN-code: 3341-2372

MD, Dr. Sci (Med.), Professor, Head of Department of Psychiatry and Addiction

Russian Federation, Saint Petersburg

References

Supplementary files