Next Article in Journal
Proteomic Profiling of Pre- and Post-Surgery Saliva of Glioblastoma Patients: A Pilot Investigation
Next Article in Special Issue
AdipoRon’s Impact on Alzheimer’s Disease—A Systematic Review and Meta-Analysis
Previous Article in Journal
Evaluation of In Vitro Biological Activity of Flavanone/Chromanone Derivatives: Molecular Analysis of Anticancer Mechanisms in Colorectal Cancer
Previous Article in Special Issue
IL-1 Blockade Mitigates Autism and Cerebral Palsy Traits in Offspring In-Utero Exposed to Group B Streptococcus Chorioamnionitis
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
IJMS
Volume 25
Issue 23
10.3390/ijms252312981
ijms-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Brief Report

RNA Editing Signatures Powered by Artificial Intelligence: A New Frontier in Differentiating Schizophrenia, Bipolar, and Schizoaffective Disorders

by
Francisco J. Checa-Robles
1,2,
Nicolas Salvetat
1,2,
Christopher Cayzac
1,2,
Mary Menhem
1,2,
Mathieu Favier
1,2,
Diana Vetter
1,2,
Ilhème Ouna
2,
João V. Nani
3,4,
Mirian A. F. Hayashi
3,4,
Elisa Brietzke
5 and
Dinah Weissmann
1,2,*
1
ALCEDIAG, Parc Euromédecine, 34184 Montpellier Cedex 4, France
2
Sys2Diag, UMR 9005 CNRS/ALCEN, Parc Euromédecine, 34184 Montpellier Cedex 4, France
3
Department of Pharmacology, Escola Paulista de Medicina (EPM), Universidade Federal de São Paulo (UNIFESP), São Paulo CEP 04044-20, Brazil
4
National Institute for Translational Medicine (INCT-TM, CNPq/FAPESP/CAPES), Ribeirão Preto CEP 14040-900, Brazil
5
Department of Psychiatry, School of Medicine, Queen’s University, Kingston, ON K7L 7X3, Canada
*
Author to whom correspondence should be addressed.
Int. J. Mol. Sci. 2024, 25(23), 12981; https://doi.org/10.3390/ijms252312981
Submission received: 22 October 2024 / Revised: 22 November 2024 / Accepted: 29 November 2024 / Published: 3 December 2024
(This article belongs to the Special Issue Translating Molecular Psychiatry: From Biomarkers to Personalized Therapies)
Browse Figures
Versions Notes

Abstract

:
Mental health disorders are devastating illnesses, often misdiagnosed due to overlapping clinical symptoms. Among these conditions, bipolar disorder, schizophrenia, and schizoaffective disorder are particularly difficult to distinguish, as they share alternating positive and negative mood symptoms. Accurate and timely diagnosis of these diseases is crucial to ensure effective treatment and to tailor therapeutic management to each individual patient. In this context, it is essential to move beyond standard clinical assessment and employ innovative approaches to identify new biomarkers that can be reliably quantified. We previously identified a panel of RNA editing biomarkers capable of differentiating healthy controls from depressed patients and, among depressed patients, those with major depressive disorder and those with bipolar disorder. In this study, we integrated Adenosine-to-Inosine RNA editing blood biomarkers with clinical data through machine learning algorithms to establish specific signatures for bipolar disorder and schizophrenia spectrum disorders. This groundbreaking study paves the way for the application of RNA editing in other psychiatric disorders, such as schizophrenia and schizoaffective disorder. It represents a first proof-of-concept and provides compelling evidence for the establishment of an RNA editing signature for the diagnosis of these psychiatric conditions.

1. Introduction

Mental disorders are highly prevalent and often devastating diseases that negatively impact the lives of millions of people worldwide [1]. Bipolar disorder (BD) is a severe, recurrent, and often disabling mood disorder characterized by manic, hypomanic, or mixed episodes and alternating episodes of depression [2]. Within mood disorders, BD is one of the most frequent and disabling, affecting approximately 1–2% of the world’s population [3,4,5]. According to clinical data, 75% of symptomatic time for BD patients consists of depressive episodes, while manic or hypomanic episodes are less frequent. Of note, early diagnosis and treatment are associated with a more favorable prognosis [4,6]. BD is commonly underdiagnosed and often mistaken for other mental health conditions. As a result, diagnostic wandering for BD has been estimated approximately from 7 to 10 years [4,7,8,9], delaying appropriate treatment, increasing comorbidity development and suicide risk, and leading to massive costs for patients and society [10,11,12,13]. Schizophrenia (SZ) is characterized by positive symptoms such as delusions and hallucinations and negative symptoms including avolition, social withdrawal, and blunted affect [14,15]. Schizoaffective disorder (SA) is a persistent and severe illness characterized by the simultaneous manifestation of symptoms associated with SZ and affective disorders, including depression and/or (hypo)mania [2,16]. Due to symptom overlaps between BD and SZ spectrum disorders, reliable and accurate differential diagnosis represents a notable challenge for the appropriate treatment of these mental disorders [16,17,18]. Currently, the diagnosis of these pathologies is a complex and time-consuming process that relies on clinical assessment of self-reported symptoms, medical history, and clinician expertise, which is subjective and error-prone [19]. Accurate and reliable biological prediction for diagnosing fluctuating mood disorders remains an unmet clinical need, and molecular biomarkers could serve as a crucial foundation for diagnosis. Precision medicine aims to tailor medical interventions to individual characteristics identified by various “omics” approaches, including epigenomics [20,21,22]. Adenosine-to-Inosine (A-to-I) RNA editing is a post-transcriptional modification that consists of the deamination of an Adenosine (A) nucleotide into an Inosine (I) nucleotide at precise locations along the RNA molecule [23,24,25,26,27]. Depending on the location where RNA editing occurs along the RNA molecule, this mechanism can impact splicing, affect RNA stability, modulate gene expression, or directly modify protein structure and function [28,29,30,31]. This highly dynamic process enhances gene diversity and is linked to regulatory systems in psychiatry. A growing number of studies indicate that dysregulation of RNA editing is associated with neuropsychiatric disorders [28,32,33,34,35,36,37]. We have successfully identified blood RNA editing-related gene modifications that enable the differentiation with high accuracy between healthy controls and depressed patients [17], as well as between patients with BD and unipolar depression [17,38]. Recently, we further demonstrated the diagnostic value of a panel of eight RNA editing genes for BD in another cohort of Brazilian patients, using novel combinations that distinguished euthymic BD patients from depressed BD patients and healthy controls [39].
In this study, we first analyze RNA editing modifications in a panel of eight genes (CAMK1D, GAB2, IFNAR1, KCNJ15, LYN, MDM2, PDE8A, and PRKCB) across four groups of individuals (BD, SZ, SA, and healthy controls) using targeted next-generation sequencing (NGS). RNA editing biomarkers were then combined to identify the optimal signatures for distinguishing between each group. Finally, RNA editing profiles were integrated with clinical characteristics using a machine learning approach, and the diagnostic performance of the resulting algorithms was evaluated. These findings highlight the potential of artificial intelligence (AI)-based predictions using RNA-editing biomarkers to improve diagnostic precision in mental health.

2. Results

2.1. Characteristics of the Populations

This study conducted a comparative analysis involving 85 control (CTRL) subjects, 39 BD, 31 SZ, and 14 SA patients. Regarding the treatments, the following five main classes of psychiatric drugs were considered (see Table 1): antipsychotics, antidepressants, anxiolytics, antiepileptics, and hypnotics/sedatives. Additional details are available in Supplementary Material.

2.2. Target Editing Index (TEI)

RNA editing modifications were analyzed using a panel of eight genes (CAMK1D, GAB2, IFNAR1, KCNJ15, LYN, MDM2, PDE8A, PRKCB; see details in Supplementary Material) that we previously identified and whose editing combinations were suggested as biomarkers to distinguish BD from unipolar depression [38] and subcategories of BD patients [39]. A feature selection process was implemented, incorporating the removal of highly correlated features [40], p-value filtering, and the Boruta algorithm [41]. This rigorous process resulted in a final list of 32 significant RNA editing biomarkers, including 3 biomarkers from CAMK1D, 7 biomarkers from GAB2, 6 biomarkers from IFNAR1, 3 biomarkers from KCNJ15, 3 biomarkers from LYN, 6 biomarkers from MDM2, 1 biomarker from PDE8A, and 3 biomarkers from PRKCB (Supplementary Figure S1). An analysis was performed to examine these 32 RNA editing biomarkers across the 4 groups, summarized as TEI (Figure 1). In comparison to the CTRL group, the SZ and SA groups showed a significant difference (p-value FDR < 0.10) in TEI on KCNJ15 and PRKCB or KCNJ15 and LYN, respectively. As compared to SZ and/or SA groups, the BD group revealed a significant difference (p-value FDR < 0.10) in TEI on CAMK1D, GAB2, IFNAR1, KCNJ15, LYN, MDM2, and PDE8A. However, no significant differences in TEI values were noticed between the SZ and SA groups.

2.3. RNA Editing Biomarkers Combination Analysis

The 32 significant RNA editing biomarkers were combined to assess their collective diagnostic power using two distinct approaches. First, a biomarkers-only model approach using linear combinations of RNA editing biomarkers to maximize the area under the curve (AUC) from the receiving operating characteristic (ROC) curve [42] for pairwise comparisons across all diagnostic categories. Second, an integrated model approach combined the 32 RNA editing biomarkers with covariates such as sex and psychiatric treatments, leveraging a multiclass random forest (RF) [40,43] machine learning algorithm to enhance predictive accuracy.

2.3.1. Biomarkers-Only Model

Significant RNA editing biomarkers were combined to determine for each comparison the optimal linear signature, facilitating group separation. “Z” represents this combination, comprising a list of 32 significant RNA editing biomarkers. As shown in Figure 2, the combination of (Z1), (Z2), (Z3), (Z4), (Z5), and (Z6) resulted in AUC values of 0.910, 0.955, 0.990, 0.962, 0.999, and 1, with sensitivities of 82.1, 91.1, 97.8, 96.8, 100, and 100%, and specificities of 87.1, 87.1, 94.9, 87.1, 98.8, and 100%, respectively. This indicates a clear separation between CTRL and BD, CTRL and SZ + SA, BD and SZ + SA, CTRL and SZ, CTRL and SA, and SZ and SA.

2.3.2. Integrated RF Model

The 32 significant biomarkers with associated covariates were combined using a multiclass RF algorithm. RF models are particularly effective on data from clinical trials because of their ability to handle complex and non-linear relationships between variables, enabling them to uncover intricate patterns and interactions within the data. The algorithm was trained on 70% of the population. Then, the test was performed on the 30% of the population who never saw the algorithm. The results obtained were plotted in a 3D scatterplot (Supplementary Figure S2A). The RF model clearly discriminates CTRL vs. BD vs. SZ vs. SA with high performance (Supplementary Figure S2A). Feature importance analysis was conducted using multiple indicators to evaluate the contribution of each feature from various perspectives. This multi-way approach ensures a more balanced assessment, mitigating the risk of relying solely on a single indicator and accounting for potential biases in feature dominance (Supplementary Table S1 and Supplementary Figures S3–S5). These analyses revealed that antipsychotics were the most influential variable in the RF model, but RNA editing biomarkers also played an important and significant role, demonstrating that they are fully integrated and utilized in the model’s predictive framework. Finally, sex was the least influential variable in the model.
Given the relatively small number of patients and the lack of an independent cohort to replicate RF results, we performed a Monte Carlo (MC) simulation to estimate prediction error (Supplementary Figure S2B). The MC method generated 500 synthetic data points from the original dataset and proved to be an effective and reliable method for data augmentation, with the synthetic data reproducing the original dataset’s structure and statistical properties (Supplementary Figures S2B and S6).

2.4. Results Conclusion

First, we found that RNA editing of eight previously identified genes differed significantly between healthy controls and patients with SZ and SA, as well as between BD patients and SZ/SA individuals. Conversely, RNA editing was found to be comparable between SZ and SA groups for genes considered in this study.
Then, we demonstrated that the combination of selected significant biomarkers allowed us to discriminate CTRL vs. BD vs. SZ vs. SA with high accuracy.
Finally, we showed that using machine learning algorithms to integrate these RNA editing biomarker combinations with patients’ clinical features enabled us to create an RF model with high diagnostic validity.

3. Discussion

Personalized medicine has become increasingly relevant to many medical fields, including psychiatry, promising earlier intervention, more efficient drug therapies, and individualized care management. BD, SZ, and SA share similar depressive symptoms and episodes of mood fluctuations [16,18], and differential diagnosis may be difficult. SA patients exhibit a clinical pattern resembling SZ, but with a longitudinal progression more akin to that of BD patients [44]. Thus, due to this diagnostic difficulty, the resulting delay is associated with inappropriate treatment and a poorer outcome [16,18,45,46]. The development of objective and reliable tools for the differential diagnosis of BD and SZ spectrum disorders is therefore essential, and several recent studies have highlighted promising research into biomarkers [47,48,49,50]. In this context, RNA editing and AI-based predictive testing represent an exciting advancement in precision medicine [17,38,39]. Machine learning, a subfield of artificial intelligence, offers the potential to significantly accelerate the diagnostic process [51]. The capacity to discriminate between these populations of patients will decrease the risk of exacerbation of the depressive state or transition to other phases (e.g., mania [52]), reduce hospitalization rates and associated costs, and diminish the comorbidities, substance abuse, and suicide risk. Furthermore, it could help maintain patients’ professional and personal lives by preventing interruptions due to ineffective care. Due to the dynamic and evolving nature of epitranscriptomic modifications, such as RNA editing, which is believed to reflect environmental cues at the molecular level, their predictive value is also expected to play a role in monitoring treatment response and emerging adverse events [53]. RNA editing analysis via NGS technology and machine learning represent an innovative approach. Previous studies have demonstrated the capacity of these techniques to efficiently identify RNA editing profiles and blood signatures with high accuracy, reproducibility, and reliability [17,38,39].
Recently, applying machine learning and combining RNA editing with clinical information from a cohort of depressed patients and healthy controls, we successfully identified specific blood signatures that demonstrated high accuracy in differentiating BD from unipolar depression (major depressive disorder, MDD) [17]. Furthermore, we validated the efficacy and reliability of this diagnostic tool using an external cohort, where an optimized algorithm achieved comparable performance in discriminating BD from MDD [38].
Additionally, we found RNA editing biomarker combinations capable of differentiating euthymic from depressed or mixed-state BD patients [39]. In this study, we extend our findings by demonstrating that specific combinations of RNA editing biomarkers from the same panel of genes can differentiate patients with BD from those with SZ and SA.
We selected a list of 32 significant RNA editing biomarkers from 8 genes used for all comparisons (Z1 to Z6, Figure 1 and Supplementary Figure S1), which led to characterizing CTRL from patients with BD, SZ, and SA (Figure 2). Using a multiclass RF algorithm, the predictive capacity of these biomarkers to differentiate CTRL, BD, SA, and SZ was tested (Supplementary Figure S2). The genes analyzed in this study are associated with specific clinical conditions, particularly psychiatric disorders and immune responses. Notably, both central and peripheral immune alterations have been observed in BD and SZ [47]. Through a series of Ca2+-dependent signaling cascades, CAMK1D plays a key role in neuronal development, neuronal transmission, synaptic plasticity, cognition, and neuroinflammation processes that are critical to brain function and mental health. CAMK1D promotes basal dendritic growth of hippocampal neurons, activates CREB1-dependent gene transcription factor, regulates granulocyte functions, and contributes to cytokine-induced proliferation and neutrophil activation [54,55]. Single-nucleotide polymorphisms (SNPs) in CAMK1D have been linked to depressive episodes and suicide attempts in patients with depression [56]. Using in vitro assays, CAMK1D was shown to phosphorylate substrates including synapsin 1 and 2, and CREB 1, which were previously associated with SZ [57]. While MDM2 is primarily recognized for its role in cancer biology [58], emerging evidence suggests it may also play a significant role in mental disorders, especially through its regulation of stress response [59], ubiquitination activity, and cell survival in the brain. MDM2 catalyzes the ubiquitination of β-arrestin, a target for antidepressants, and is involved in AMPAR surface expression during synaptic plasticity within the central nervous system [60]. Additionally, post mortem analysis revealed significant reductions in MDM2 expression among subjects with SZ as compared to CTRL [61]. Emerging research suggests that IFNAR1 dysregulation and the resulting impact on type I IFN signaling can influence the pathophysiology of several mental disorders via neuroinflammation [62]. RNA editing shows a strong association with interferon response, and polymorphisms in the promoter region of IFNAR1 have been linked to increased depression risk [63,64,65]. Interferon-based treatments can lead to adverse effects, including neurological and neuropsychiatric disorders, as well as autoimmune disease [62]. Emerging evidence suggests that LYN dysregulation may contribute to the pathophysiology of certain mental disorders, primarily through its involvement in immune modulation and neural signaling pathways [66]. LYN has been found to enhance glutamatergic synaptic transmission and activate the mitogen-activated protein kinase (MAPK) pathway, which increases Brain-Derived Neurotrophic Factor (BDNF) expression [67]. LYN also modulates N-methyl-D-aspartate receptor (NMDAR) function, a pathway critical to SZ, as NMDAR hypofunction is implicated as a core component of the disorder [68]. PRKCB, a member of the protein kinase C (PKC) family, is particularly important in the brain through its influence on neurotransmitter release, synaptic plasticity, and neural signaling. Further findings highlight that PRKCB expression is downregulated in the peripheral blood mononuclear cells of depressed patients, with specific SNPs in PRKCB showing association with major depression [69]. Located at 16p12.2-p12.1, PRKCB encodes a protein essential to synaptic plasticity [70], and impaired synaptic plasticity is increasingly recognized as a pathogenic factor in SZ [71]. In addition, abnormalities in PKC have been implicated in BD, with decreased PKC activity observed in the prefrontal and temporal cortex of BD patients [72]. Indeed, endoxifen, a PKC inhibitor, significantly improved type I BD symptoms [73]. PDE8A is a cAMP-specific phosphodiesterase that plays a vital role in cellular signaling by hydrolyzing cAMP, a second messenger regulating various physiological processes such as neuronal function and inflammation [74,75,76]. In patients with MDD, expression of PDE8A in the temporal cortex is reduced by half, and PDEs are known to play key roles in signal transduction, inflammatory cell activation, memory, and cognition [77]. Memory and cognitive impairments are common in SZ, further linking PDE8A to these psychiatric conditions [78]. GAB2 is widely expressed in the central nervous system and is a pivotal component in immune responses, engaging in multiple signaling pathways [79,80]. GAB2 overexpression makes neurons vulnerable, increasing tau phosphorylation and leading to the Alzheimer’s disease phenotype [81]. Genetic analyses also identify the DISC1 locus as significant. This gene was originally discovered in a Scottish pedigree in which chromosomal translocation (1;11) (q42.1; q14.3) was linked to multiple psychiatric illnesses, including SZ, BD, and MDD. GRB2, through its docking on GAB2, interacts with DISC1 in the brain, competing with NDEL1 (nuclear distribution element like-1) and playing a critical role in pathways involving ErbB receptors related to SZ [82]. Lastly, KCNJ15, a member of the inwardly rectifying potassium channel family, plays a role in maintaining potassium ion homeostasis. Potassium channels are involved in the regulation of neuronal excitability and neurotransmission, processes often disrupted in neurological diseases including epilepsy [83] and Parkinson’s disease [84]. KCNJ15 has been shown to influence immune activity [85], with specific variants linked to an increased risk of Alzheimer’s disease in the Chinese population, possibly acting through immune modulation [85]. Thus, all the identified genes exhibit a relationship, either direct or indirect, with neuronal and/or inflammatory mechanisms, particularly BD and SZ spectrum disorders.
Medications are included as covariates within the RF model. By incorporating in the model RNA editing biomarkers, psychiatric treatment classes (based on ATC classification), and sex as features, the model captures complex relationships among these factors and their combined influence on diagnostic categories. RF models are especially valuable for medication adjustments as they effectively manage the complex non-linear interactions often encountered in clinical trials. Additionally, the model’s ability to assess feature importance adds transparency, allowing clinicians to understand which variables most significantly influence the predictions. Psychiatric treatment ATC classes frequently appear as root variables in the decision trees, although RNA-editing biomarkers also hold prominent positions. Moreover, the results indicate that the model’s predictions are not sensitive to biological sex. Analysis of the model’s explainability suggests that the threshold values of RNA-editing biomarkers within the decision trees closely align with the influence of medications, indicating a meaningful interplay between treatment effects and biomarker responses in the model’s classification process. Furthermore, all these results demonstrate the importance of RNA editing biomarkers in the RF model and their adjustment with medications. The performances and spatial classification of the four classes obtained from simulated data using the MC method enhance the reliability of this model as a first step to large cohort validation. The MC simulation approach proves to be an effective method for data augmentation, addressing the challenges of limited data availability.
Some limitations should be considered in this study. First, patients in the BD group were in different mood states (hypomanic/manic, depressive, mixed, and euthymic bipolar). Second, sex proportion was significantly different among the groups. Third, the small sample size (particularly for the SA group) may induce a potential bias in this study, and without replication, it is important to take the results of this study with caution.
This initial proof-of-concept analysis presents compelling evidence for the establishment of an RNA editing signature for diagnosis and potentially for prognosis and/or treatment prediction. Further validation using a larger cohort and replication of these results will be required.

4. Materials and Methods

This study involved the following complementary methodological approaches:
Recruitment of healthy control individuals (CTRL, n = 85) and patients with bipolar disorder (BD, n = 39), schizophrenia (SZ, n = 31), and schizoaffective disorder (SA, n = 14), using 2 independent cohorts.
RNA extraction from blood samples of recruited individuals and targeted sequencing of a previously defined panel of RNA editing genes.
Bioinformatics analysis for ensuring the quality of sequencing data, filter data based on length and quality score, perform the alignment with reference human genome, and identify editing events (site, isoform, and/or pattern, referred to as biomarkers; see Supplementary Material).
Biostatistics for calculating a Target Editing Index (TEI; see below and Supplementary Material for more details) are obtained in the following way: finding significant differences in editing between groups of individuals (CTRL vs. BD vs. SZ vs. SA) and combining identified biomarkers using machine learning approaches (Random Forest algorithm), which consider patient clinical characteristics.
Additional information and references supporting the methods used in this study, particularly for bioinformatics and biostatistical analysis, are provided in Supplementary Material.

4.1. Subjects and Clinical Assessment

The patient samples included in this study originated from two distinct cohorts. The Helsinki Declaration’s guiding principles were closely followed throughout this study. All participants, aged between 18 and 65 years, signed a written informed consent before entering this study. The cohort of BD patients was approved by the Research Ethics Committee of UNIFESP [CEP No. 1427/16]. Samples from patients with SZ and SA were obtained from PrecisionMed Inc. These samples were collected with informed consent under Western Institutional Review Boards (WIRB®) protocol #20041439 and protocol no 1022, reviewed and approved by their Institutional Review Board. The Structured Clinical Interview for DSM-IV Axis I disorders (SCID-1) was conducted with eligible participants. The Hamilton Depression Rating Scale-17 items (HDRS) and the Young Mania Rating Scale (YMRS) were used to assess the severity of depressive and manic symptoms, respectively. Patients in this study were diagnosed with SZ according to M.I.N.I. (Mini International Neuropsychiatric Interview) and SCI-PANSS (Structured Clinical Interview for the Positive and Negative Syndrome Scale). A form document completed by the subjects’ psychiatrist confirmed their diagnosis according to DSM-IV criteria. Volunteers in the healthy control group included individuals without a history of psychotropic medication use, no lifetime or current mental disorders, and no family history of a major psychiatric disorder in first-degree relatives. More details about demographic and clinical characteristics are shown in Table 1.

4.2. RNA Extraction and Qualification

Samples were retrieved in PAXgene™ blood RNA tubes and extracted using the MagNA Pure 96 instrument (Roche, Bâle, Suisse). Total RNA concentrations and quality were determined with Qubit Fluorometer (Life technologies, Carlsbad, CA, USA) and LabChip GX (Perkin-Elmer, Waltham, MA, USA) instruments, respectively.

4.3. Targeted Next Generation Sequencing

A panel of blood RNA editing genes (PRKCB, PDE8A, CAMK1D, GAB2, IFNAR1, KCNJ15, LYN, and MDM2) was designed [39], and validated primers were used to amplify each gene on a Peqstar 96 × thermocycler (VWR, Radnor, PA, USA). PCR products underwent purification with magnetic beads, quantification, and indexing. The resulting library was pooled, purified, denatured, spiked with PhiX Control V3 (Illumina, San Diego, CA, USA), and loaded onto a sequencing cartridge. Finally, the library was sequenced using Illumina NextSeq 500 (Illumina, San Diego, CA, USA).

4.4. Bioinformatics Analysis

FastQC software (version 0.11.9) was used to verify the quality of the sequencing data. A minimal sequencing depth of 10,000 reads for each sample was considered for further analysis. A pre-treatment step was performed, assessing both length and quality scores. Bowtie2 was used to align the processed reads to the reference human genome sequence, GRCh38 [86]. For SVN calling, SAMtools mpileup was applied [87,88]. By counting the different nucleotides in each genomic region, proprietary scripts were used to identify the edited positions in the alignment. RNA editing values for all potential biomarkers (including sites, isoforms, and motifs) were evaluated against the background (for more information see Supplemental Methods in Supplementary Material).

4.5. Biostatistical Analysis

All statistics and figures were computed with the “R/Bioconductor” software (R version 4.4.1, Bioconductor Manager version 3.19) [89,90]. A differential analysis was carried out using the most appropriate test between the Mann–Whitney rank-sum test, Student’s t-test, or Welch’s t-test according to normality and sample variance distribution in each cohort. The false discovery rate (FDR) was managed through the use of the Benjamini and Hochberg (BH) procedure [91], and an adjusted p-value below 0.1 was considered statistically significant. Due to numerous significant biomarkers (p-value ≤ 0.05), a feature selection process was implemented by selecting the biomarkers less correlated (R = 0.6) and showing more importance related to the class, and using the functions find correlation from the R package “caret” version 6.0-94, the Boruta algorithm (a robust feature selection method based on the random forest classifier) from the R package “Boruta” version 8.0.0, and the function mutinformation (a feature selection method based on an entropy estimator) from the R package “infotheo” version 1.2.0.1 [40,41,92]. This rigorous feature selection process resulted in a final list of 32 significant RNA editing biomarkers. A “Target Editing Index” (TEI), resuming gene-specific editing values, was calculated by a linear combination of significant RNA editing variants maximizing the AUC of ROC (Figure 1) [42]. The TEI was transformed using the Boxcox transform, aiming to guarantee the normal distribution of data [93]. After assessing the target validity with the TEI analysis, we explored a signature of 32 combined RNA editing biomarkers to classify all classes. The 32 biomarkers were normalized by the corresponding target global editing and then combined using the mROC multivariate method for each comparison (Figure 2). The equation for the respective combination is provided and can be used as a new virtual marker Z, as follows: Z = a × biomarker 1 + b × biomarker 2 + c × biomarker 3, where a, b, and c are calculated coefficients and biomarkers 1, 2, and 3 are the levels of biomarker.
Finally, the same 32 biomarkers were combined using multiclass Random Forest (RF) [43], considering patient treatment and sex (see Table 1 and Supplementary Material). RF requires the use of a training set to construct the model (70% of the population; n = 120 samples) and a test set (30% of the population; n = 49 samples). This sharing has been randomized and respects the initial proportion of the various statutes in each set. The RF method combines Breiman’s “bagging” idea and the random selection of features in order to construct a collection of decision trees with controlled variance. To generalize the model, we performed 1000 learning RF trees (n = 120). The final classifier was generated by summing up the votes (probabilities) for each applied sample from each tree and was normalized by the number of trees. The end resulting probabilities reflect the majority vote of the 1000 RF trees. We used a grid learning approach for each individual tree, where we stated certain maximum parameter sets (ntree  =  1000, nodesize  =  1 and mtry  =  (1, 20)). RF results are shown on the test dataset (n = 49), which has never seen the algorithm. The implementation was performed using the R RandomForest package (version 4.6-14) and the R caret package (version 6.0-84). Variable importance analysis was assessed using the randomForestExplainer packages in R [94]. Monte Carlo (MC) simulation was performed to generate synthetic data [95]. For more information about RF model interpretability and MC simulation, see Supplemental Methods in Supplementary Material.

5. Conclusions

RNA editing coupled with AI paves the way for improved classification of patients with mood fluctuations into specific disease subcategories. This approach aligns with the principles of precision psychiatry, aiming to deepen our understanding of the pathophysiological mechanisms of depression and improve therapeutic care. In addition, RNA editing and machine learning represent valuable tools in attempting to think beyond diagnostic symptoms, better understand heterogeneity in psychiatric disorders, and consider the underlying features of psychopathology from a neuroscience-based approach.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/ijms252312981/s1. References [96,97,98,99] are cited in the supplementary materials.

Author Contributions

Conceptualization, N.S., M.A.F.H., E.B. and D.W.; methodology, N.S., C.C., I.O., J.V.N., M.A.F.H., E.B. and D.W.; software, F.J.C.-R. and N.S.; validation, F.J.C.-R., N.S., C.C. and I.O.; formal analysis, F.J.C.-R. and N.S.; resources, C.C., D.V., I.O. and J.V.N.; writing—original draft preparation, M.M.; writing—review and editing, F.J.C.-R., N.S., M.F., M.A.F.H., E.B. and D.W.; visualization, F.J.C.-R. and N.S.; supervision, M.A.F.H., E.B. and D.W.; project administration, D.W.; funding acquisition, M.A.F.H., E.B. and D.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by João V. Nani, recipient of a Fellowship from FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo—No. 2022/03297-3 and 2019/09207-3). Mirian AF Hayashi is supported by FAPESP [No. 2022/00527-8; 2020/01107-7; 2019/13112-8], CAPES, and CNPq. MAF Hayashi is also a recipient of a fellowship from CNPq [39337/2016-0].

Institutional Review Board Statement

The Helsinki Declaration’s guiding principles were closely followed throughout this study. The cohort of BD patients was approved by the Research Ethics Committee of UNIFESP [CEP No. 1427/16]. Samples from patients with SZ and SA were obtained from PrecisionMed Inc. These samples were collected with informed consent under Western Institutional Review Boards (WIRB®) protocol #20041439 and protocol no 1022, reviewed and approved by their Institutional Review Board.

Informed Consent Statement

All participants, aged between 18 and 65 years, signed a written informed consent before entering this study.

Data Availability Statement

Readers will be able to access the data, associated protocols, code, and materials in the paper by asking to Dinah Weissmann ([email protected]).

Conflicts of Interest

The authors declare the following financial interests/personal relationships that may be considered as potential competing interests: Francisco J. Checa-Robles reports a relationship with ALCEDIAG that includes: employment. Nicolas Salvetat reports a relationship with ALCEDIAG that includes: employment. Christopher Cayzac reports a relationship with ALCEDIAG that includes: employment. Mary Menhem reports a relationship with ALCEDIAG that includes: employment. Mathieu Favier reports a relationship with ALCEDIAG that includes: employment. Diana Vetter reports a relationship with ALCEDIAG that includes: employment. Dinah Weissmann reports a relationship with ALCEDIAG that includes: employment. The rest of the authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References

  1. Scangos, K.W.; State, M.W.; Miller, A.H.; Baker, J.T.; Williams, L.M. New and emerging approaches to treat psychiatric disorders. Nat. Med. 2023, 29, 317–333. [Google Scholar] [CrossRef]
  2. Pacchiarotti, I.; Tiihonen, J.; Kotzalidis, G.D.; Verdolini, N.; Murru, A.; Goikolea, J.M.; Valenti, M.; Aedo, A.; Vieta, E. Long-acting injectable antipsychotics (LAIs) for maintenance treatment of bipolar and schizoaffective disorders: A systematic review. Eur. Neuropsychopharmacol. 2019, 29, 457–470. [Google Scholar] [CrossRef]
  3. McDonald, K.C.; Bulloch, A.G.; Duffy, A.; Bresee, L.; Williams, J.V.; Lavorato, D.H.; Patten, S.B. Prevalence of Bipolar I and II Disorder in Canada. Can. J. Psychiatry 2015, 60, 151–156. [Google Scholar] [CrossRef]
  4. Nierenberg, A.A.; Agustini, B.; Kohler-Forsberg, O.; Cusin, C.; Katz, D.; Sylvia, L.G.; Peters, A.; Berk, M. Diagnosis and Treatment of Bipolar Disorder: A Review. JAMA 2023, 330, 1370–1380. [Google Scholar] [CrossRef]
  5. Merikangas, K.R.; Jin, R.; He, J.P.; Kessler, R.C.; Lee, S.; Sampson, N.A.; Viana, M.C.; Andrade, L.H.; Hu, C.; Karam, E.G.; et al. Prevalence and correlates of bipolar spectrum disorder in the world mental health survey initiative. Arch. Gen. Psychiatry 2011, 68, 241–251. [Google Scholar] [CrossRef]
  6. Judd, L.L.; Akiskal, H.S.; Schettler, P.J.; Endicott, J.; Maser, J.; Solomon, D.A.; Leon, A.C.; Rice, J.A.; Keller, M.B. The long-term natural history of the weekly symptomatic status of bipolar I disorder. Arch. Gen. Psychiatry 2002, 59, 530–537. [Google Scholar] [CrossRef]
  7. Dagani, J.; Signorini, G.; Nielssen, O.; Bani, M.; Pastore, A.; Girolamo, G.; Large, M. Meta-analysis of the Interval between the Onset and Management of Bipolar Disorder. Can. J. Psychiatry 2017, 62, 247–258. [Google Scholar] [CrossRef]
  8. Fritz, K.; Russell, A.M.T.; Allwang, C.; Kuiper, S.; Lampe, L.; Malhi, G.S. Is a delay in the diagnosis of bipolar disorder inevitable? Bipolar Disord. 2017, 19, 396–400. [Google Scholar] [CrossRef]
  9. Scott, J.; Graham, A.; Yung, A.; Morgan, C.; Bellivier, F.; Etain, B. A systematic review and meta-analysis of delayed help-seeking, delayed diagnosis and duration of untreated illness in bipolar disorders. Acta Psychiatr. Scand. 2022, 146, 389–405. [Google Scholar] [CrossRef]
  10. Beyer, J.L. The Use of Antidepressants in Bipolar Depression. Handb. Exp. Pharmacol. 2019, 250, 415–442. [Google Scholar] [CrossRef]
  11. Bauer, M.; Andreassen, O.A.; Geddes, J.R.; Vedel Kessing, L.; Lewitzka, U.; Schulze, T.G.; Vieta, E. Areas of uncertainties and unmet needs in bipolar disorders: Clinical and research perspectives. Lancet Psychiatry 2018, 5, 930–939. [Google Scholar] [CrossRef]
  12. Stang, P.E.; Frank, C.; Kalsekar, A.; Yood, M.U.; Wells, K.; Burch, S. The clinical history and costs associated with delayed diagnosis of bipolar disorder. Medscape Gen. Med. 2006, 8, 18. [Google Scholar]
  13. Scott, J.; Leboyer, M. Consequences of delayed diagnosis of bipolar disorders. Encephale 2011, 37 (Suppl. 3), S173–S175. [Google Scholar] [CrossRef]
  14. Nolan, C.J.; Roepke, T.A.; Perreault, M.L. Beyond the Binary: Gender Inclusivity in Schizophrenia Research. Biol. Psychiatry 2023, 94, 543–549. [Google Scholar] [CrossRef]
  15. Tandon, R.; Nasrallah, H.; Akbarian, S.; Carpenter, W.T., Jr.; DeLisi, L.E.; Gaebel, W.; Green, M.F.; Gur, R.E.; Heckers, S.; Kane, J.M.; et al. The schizophrenia syndrome, circa 2024: What we know and how that informs its nature. Schizophr. Res. 2024, 264, 1–28. [Google Scholar] [CrossRef]
  16. Malhi, G.S.; Green, M.; Fagiolini, A.; Peselow, E.D.; Kumari, V. Schizoaffective disorder: Diagnostic issues and future recommendations. Bipolar Disord. 2008, 10, 215–230. [Google Scholar] [CrossRef]
  17. Salvetat, N.; Checa-Robles, F.J.; Patel, V.; Cayzac, C.; Dubuc, B.; Chimienti, F.; Abraham, J.D.; Dupre, P.; Vetter, D.; Mereuze, S.; et al. A game changer for bipolar disorder diagnosis using RNA editing-based biomarkers. Transl. Psychiatry 2022, 12, 182. [Google Scholar] [CrossRef]
  18. Citrome, L.; Goldberg, J.F. The many faces of bipolar disorder. How to tell them apart. Postgrad. Med. 2005, 117, 15–23. [Google Scholar] [CrossRef]
  19. Carvalho, A.F.; Firth, J.; Vieta, E. Bipolar Disorder. N. Engl. J. Med. 2020, 383, 58–66. [Google Scholar] [CrossRef] [PubMed]
  20. Stolfi, F.; Abreu, H.; Sinella, R.; Nembrini, S.; Centonze, S.; Landra, V.; Brasso, C.; Cappellano, G.; Rocca, P.; Chiocchetti, A. Omics approaches open new horizons in major depressive disorder: From biomarkers to precision medicine. Front. Psychiatry 2024, 15, 1422939. [Google Scholar] [CrossRef] [PubMed]
  21. Shih, P.B. Metabolomics Biomarkers for Precision Psychiatry. Adv. Exp. Med. Biol. 2019, 1161, 101–113. [Google Scholar] [CrossRef] [PubMed]
  22. Mengelkoch, S.; Gassen, J.; Lev-Ari, S.; Alley, J.C.; Schussler-Fiorenza Rose, S.M.; Snyder, M.P.; Slavich, G.M. Multi-omics in stress and health research: Study designs that will drive the field forward. Stress 2024, 27, 2321610. [Google Scholar] [CrossRef] [PubMed]
  23. Walkley, C.R.; Li, J.B. Rewriting the transcriptome: Adenosine-to-inosine RNA editing by ADARs. Genome Biol. 2017, 18, 205. [Google Scholar] [CrossRef] [PubMed]
  24. Heraud-Farlow, J.E.; Walkley, C.R. What do editors do? Understanding the physiological functions of A-to-I RNA editing by adenosine deaminase acting on RNAs. Open Biol. 2020, 10, 200085. [Google Scholar] [CrossRef] [PubMed]
  25. Mannion, N.; Arieti, F.; Gallo, A.; Keegan, L.P.; O’Connell, M.A. New Insights into the Biological Role of Mammalian ADARs; the RNA Editing Proteins. Biomolecules 2015, 5, 2338–2362. [Google Scholar] [CrossRef]
  26. Bazak, L.; Haviv, A.; Barak, M.; Jacob-Hirsch, J.; Deng, P.; Zhang, R.; Isaacs, F.J.; Rechavi, G.; Li, J.B.; Eisenberg, E.; et al. A-to-I RNA editing occurs at over a hundred million genomic sites, located in a majority of human genes. Genome Res. 2014, 24, 365–376. [Google Scholar] [CrossRef]
  27. Nishikura, K. A-to-I editing of coding and non-coding RNAs by ADARs. Nat. Rev. Mol. Cell Biol. 2016, 17, 83–96. [Google Scholar] [CrossRef]
  28. Eisenberg, E.; Levanon, E.Y. A-to-I RNA editing—Immune protector and transcriptome diversifier. Nat. Rev. Genet. 2018, 19, 473–490. [Google Scholar] [CrossRef]
  29. Solomon, O.; Oren, S.; Safran, M.; Deshet-Unger, N.; Akiva, P.; Jacob-Hirsch, J.; Cesarkas, K.; Kabesa, R.; Amariglio, N.; Unger, R.; et al. Global regulation of alternative splicing by adenosine deaminase acting on RNA (ADAR). RNA 2013, 19, 591–604. [Google Scholar] [CrossRef]
  30. Kawahara, Y.; Megraw, M.; Kreider, E.; Iizasa, H.; Valente, L.; Hatzigeorgiou, A.G.; Nishikura, K. Frequency and fate of microRNA editing in human brain. Nucleic Acids Res. 2008, 36, 5270–5280. [Google Scholar] [CrossRef]
  31. Wulff, B.E.; Nishikura, K. Modulation of microRNA expression and function by ADARs. Curr. Top. Microbiol. Immunol. 2012, 353, 91–109. [Google Scholar] [CrossRef] [PubMed]
  32. Slotkin, W.; Nishikura, K. Adenosine-to-inosine RNA editing and human disease. Genome Med. 2013, 5, 105. [Google Scholar] [CrossRef] [PubMed]
  33. Maas, S.; Kawahara, Y.; Tamburro, K.M.; Nishikura, K. A-to-I RNA editing and human disease. RNA Biol. 2006, 3, 1–9. [Google Scholar] [CrossRef]
  34. Yang, Y.; Okada, S.; Sakurai, M. Adenosine-to-inosine RNA editing in neurological development and disease. RNA Biol. 2021, 18, 999–1013. [Google Scholar] [CrossRef]
  35. Silberberg, G.; Lundin, D.; Navon, R.; Ohman, M. Deregulation of the A-to-I RNA editing mechanism in psychiatric disorders. Hum. Mol. Genet. 2012, 21, 311–321. [Google Scholar] [CrossRef] [PubMed]
  36. Di Narzo, A.F.; Kozlenkov, A.; Roussos, P.; Hao, K.; Hurd, Y.; Lewis, D.A.; Sibille, E.; Siever, L.J.; Koonin, E.; Dracheva, S. A unique gene expression signature associated with serotonin 2C receptor RNA editing in the prefrontal cortex and altered in suicide. Hum. Mol. Genet. 2014, 23, 4801–4813. [Google Scholar] [CrossRef] [PubMed]
  37. Weissmann, D.; van der Laan, S.; Underwood, M.D.; Salvetat, N.; Cavarec, L.; Vincent, L.; Molina, F.; Mann, J.J.; Arango, V.; Pujol, J.F. Region-specific alterations of A-to-I RNA editing of serotonin 2c receptor in the cortex of suicides with major depression. Transl. Psychiatry 2016, 6, e878. [Google Scholar] [CrossRef]
  38. Salvetat, N.; Checa-Robles, F.J.; Delacretaz, A.; Cayzac, C.; Dubuc, B.; Vetter, D.; Dainat, J.; Lang, J.P.; Gamma, F.; Weissmann, D. AI algorithm combined with RNA editing-based blood biomarkers to discriminate bipolar from major depressive disorders in an external validation multicentric cohort. J. Affect. Disord. 2024, 356, 385–393. [Google Scholar] [CrossRef]
  39. Hayashi, M.A.F.; Salvetat, N.; Cayzac, C.; Checa-Robles, F.J.; Dubuc, B.; Mereuze, S.; Nani, J.V.; Molina, F.; Brietzke, E.; Weissmann, D. Euthymic and depressed bipolar patients are characterized by different RNA editing patterns in blood. Psychiatry Res. 2023, 328, 115422. [Google Scholar] [CrossRef]
  40. Caret Package—RDocumentation. Available online: https://rdocumentation.org/packages/caret/versions/6.0-94 (accessed on 5 July 2023).
  41. Kursa, M.B.; Rudnicki, W.R. Feature Selection with the Boruta Package. J. Stat. Softw. 2010, 36, 1–13. [Google Scholar] [CrossRef]
  42. Su, J.Q.; Liu, J.S. Linear Combinations of Multiple Diagnostic Markers. J. Am. Stat. Assoc. 1993, 88, 1350–1355. [Google Scholar] [CrossRef]
  43. Liaw, A.; Wiener, M.C. Classification and Regression by randomForest. R. News 2002, 2, 18–22. Available online: https://journal.r-project.org/articles/RN-2002-022/RN-2002-022.pdf (accessed on 4 March 2024).
  44. Benabarre, A.; Vieta, E.; Colom, F.; Martinez-Aran, A.; Reinares, M.; Gasto, C. Bipolar disorder, schizoaffective disorder and schizophrenia: Epidemiologic, clinical and prognostic differences. Eur. Psychiatry 2001, 16, 167–172. [Google Scholar] [CrossRef] [PubMed]
  45. Berk, M.; Dodd, S.; Callaly, P.; Berk, L.; Fitzgerald, P.; de Castella, A.R.; Filia, S.; Filia, K.; Tahtalian, S.; Biffin, F.; et al. History of illness prior to a diagnosis of bipolar disorder or schizoaffective disorder. J. Affect. Disord. 2007, 103, 181–186. [Google Scholar] [CrossRef] [PubMed]
  46. Kahn, R.S.; Sommer, I.E.; Murray, R.M.; Meyer-Lindenberg, A.; Weinberger, D.R.; Cannon, T.D.; O’Donovan, M.; Correll, C.U.; Kane, J.M.; van Os, J.; et al. Schizophrenia. Nat. Rev. Dis. Primers 2015, 1, 15067. [Google Scholar] [CrossRef]
  47. Skorobogatov, K.; De Picker, L.; Wu, C.L.; Foiselle, M.; Richard, J.R.; Boukouaci, W.; Bouassida, J.; Laukens, K.; Meysman, P.; le Corvoisier, P.; et al. Immune-based Machine learning Prediction of Diagnosis and Illness State in Schizophrenia and Bipolar Disorder. Brain Behav. Immun. 2024, 122, 422–432. [Google Scholar] [CrossRef]
  48. Varathan, A.; Senthooran, S.; Jeyananthan, P. Role of different omics data in the diagnosis of schizophrenia disorder: A machine learning study. Schizophr. Res. 2024, 271, 38–46. [Google Scholar] [CrossRef]
  49. Fryar-Williams, S.; Tucker, G.; Strobel, J.; Huang, Y.; Clements, P. Molecular Mechanism Biomarkers Predict Diagnosis in Schizophrenia and Schizoaffective Psychosis, with Implications for Treatment. Int. J. Mol. Sci. 2023, 24, 15845. [Google Scholar] [CrossRef]
  50. Chen, J.; Amdanee, N.; Zuo, X.; Wang, Y.; Gong, M.; Yang, Y.; Li, H.; Zhang, X.; Zhang, C. Biomarkers of bipolar disorder based on metabolomics: A systematic review. J. Affect. Disord. 2024, 350, 492–503. [Google Scholar] [CrossRef]
  51. Bzdok, D.; Meyer-Lindenberg, A. Machine Learning for Precision Psychiatry: Opportunities and Challenges. Biol. Psychiatry Cogn. Neurosci. Neuroimaging 2018, 3, 223–230. [Google Scholar] [CrossRef]
  52. Goldberg, J.F.; Truman, C.J. Antidepressant-induced mania: An overview of current controversies. Bipolar Disord. 2003, 5, 407–420. [Google Scholar] [CrossRef] [PubMed]
  53. van der Laan, S.; Salvetat, N.; Weissmann, D.; Molina, F. Emerging RNA editing biomarkers will foster drug development. Drug Discov. Today 2017, 22, 1056–1063. [Google Scholar] [CrossRef] [PubMed]
  54. Wayman, G.A.; Lee, Y.S.; Tokumitsu, H.; Silva, A.J.; Soderling, T.R. Calmodulin-kinases: Modulators of neuronal development and plasticity. Neuron 2008, 59, 914–931. [Google Scholar] [CrossRef] [PubMed]
  55. Takemoto-Kimura, S.; Suzuki, K.; Horigane, S.I.; Kamijo, S.; Inoue, M.; Sakamoto, M.; Fujii, H.; Bito, H. Calmodulin kinases: Essential regulators in health and disease. J. Neurochem. 2017, 141, 808–818. [Google Scholar] [CrossRef]
  56. Lin, E.; Kuo, P.H.; Liu, Y.L.; Yu, Y.W.; Yang, A.C.; Tsai, S.J. A Deep Learning Approach for Predicting Antidepressant Response in Major Depression Using Clinical and Genetic Biomarkers. Front. Psychiatry 2018, 9, 290. [Google Scholar] [CrossRef]
  57. Gadelha, A.; Coleman, J.; Breen, G.; Mazzoti, D.R.; Yonamine, C.M.; Pellegrino, R.; Ota, V.K.; Belangero, S.I.; Glessner, J.; Sleiman, P.; et al. Genome-wide investigation of schizophrenia associated plasma Ndel1 enzyme activity. Schizophr. Res. 2016, 172, 60–67. [Google Scholar] [CrossRef]
  58. Karni-Schmidt, O.; Lokshin, M.; Prives, C. The Roles of MDM2 and MDMX in Cancer. Annu. Rev. Pathol. 2016, 11, 617–644. [Google Scholar] [CrossRef]
  59. Scrable, H.; Burns-Cusato, M.; Medrano, S. Anxiety and the aging brain: Stressed out over p53? Biochim. Biophys. Acta 2009, 1790, 1587–1591. [Google Scholar] [CrossRef]
  60. Colledge, M.; Snyder, E.M.; Crozier, R.A.; Soderling, J.A.; Jin, Y.; Langeberg, L.K.; Lu, H.; Bear, M.F.; Scott, J.D. Ubiquitination regulates PSD-95 degradation and AMPA receptor surface expression. Neuron 2003, 40, 595–607. [Google Scholar] [CrossRef]
  61. Andrews, J.L.; Goodfellow, F.J.; Matosin, N.; Snelling, M.K.; Newell, K.A.; Huang, X.F.; Fernandez-Enright, F. Alterations of ubiquitin related proteins in the pathology and development of schizophrenia: Evidence from human and animal studies. J. Psychiatr. Res. 2017, 90, 31–39. [Google Scholar] [CrossRef]
  62. Tan, P.H.; Ji, J.; Hsing, C.H.; Tan, R.; Ji, R.R. Emerging Roles of Type-I Interferons in Neuroinflammation, Neurological Diseases, and Long-Haul COVID. Int. J. Mol. Sci. 2022, 23, 14394. [Google Scholar] [CrossRef] [PubMed]
  63. Liu, H.; Golji, J.; Brodeur, L.K.; Chung, F.S.; Chen, J.T.; deBeaumont, R.S.; Bullock, C.P.; Jones, M.D.; Kerr, G.; Li, L.; et al. Tumor-derived IFN triggers chronic pathway agonism and sensitivity to ADAR loss. Nat. Med. 2019, 25, 95–102. [Google Scholar] [CrossRef] [PubMed]
  64. Yoshida, K.; Alagbe, O.; Wang, X.; Woolwine, B.; Thornbury, M.; Raison, C.L.; Miller, A.H. Promoter polymorphisms of the interferon-alpha receptor gene and development of Interferon-induced depressive symptoms in patients with chronic hepatitis C: Preliminary findings. Neuropsychobiology 2005, 52, 55–61. [Google Scholar] [CrossRef] [PubMed]
  65. Rifai, M.A.; Sabouni, M.A. Utilizing genomic polymorphisms to personalize hepatitis C therapies. Curr. Opin. Organ Transplant. 2012, 17, 198–203. [Google Scholar] [CrossRef]
  66. Ingley, E. Functions of the Lyn tyrosine kinase in health and disease. Cell Commun. Signal 2012, 10, 21. [Google Scholar] [CrossRef]
  67. O’Neill, M.J.; Bleakman, D.; Zimmerman, D.M.; Nisenbaum, E.S. AMPA receptor potentiators for the treatment of CNS disorders. Curr. Drug Targets CNS Neurol. Disord. 2004, 3, 181–194. [Google Scholar] [CrossRef]
  68. Zhu, L.; Wu, X.; Xu, B.; Zhao, Z.; Yang, J.; Long, J.; Su, L. The machine learning algorithm for the diagnosis of schizophrenia on the basis of gene expression in peripheral blood. Neurosci. Lett. 2021, 745, 135596. [Google Scholar] [CrossRef]
  69. Costas, J.; Gratacos, M.; Escaramis, G.; Martin-Santos, R.; de Diego, Y.; Baca-Garcia, E.; Canellas, F.; Estivill, X.; Guillamat, R.; Guitart, M.; et al. Association study of 44 candidate genes with depressive and anxiety symptoms in post-partum women. J. Psychiatr. Res. 2010, 44, 717–724. [Google Scholar] [CrossRef] [PubMed]
  70. Colgan, L.A.; Hu, M.; Misler, J.A.; Parra-Bueno, P.; Moran, C.M.; Leitges, M.; Yasuda, R. PKCalpha integrates spatiotemporally distinct Ca(2+) and autocrine BDNF signaling to facilitate synaptic plasticity. Nat. Neurosci. 2018, 21, 1027–1037. [Google Scholar] [CrossRef]
  71. Forsyth, J.K.; Lewis, D.A. Mapping the Consequences of Impaired Synaptic Plasticity in Schizophrenia through Development: An Integrative Model for Diverse Clinical Features. Trends Cogn. Sci. 2017, 21, 760–778. [Google Scholar] [CrossRef]
  72. Pandey, G.N.; Rizavi, H.S.; Ren, X. Protein and mRNA expression of protein kinase C (PKC) in the postmortem brain of bipolar and schizophrenic subjects. J. Psychiatr. Res. 2020, 130, 362–371. [Google Scholar] [CrossRef] [PubMed]
  73. Ahmad, A.; Sheikh, S.; Khan, M.A.; Chaturvedi, A.; Patel, P.; Patel, R.; Buch, B.C.; Anand, R.S.; Shah, T.C.; Vora, V.N.; et al. Endoxifen: A new, protein kinase C inhibitor to treat acute and mixed mania associated with bipolar I disorder. Bipolar Disord. 2021, 23, 595–603. [Google Scholar] [CrossRef]
  74. Giorgi, M.; Pellegrini, M.; Massimi, M. Role of Phosphodiesterases in Biology and Pathology 2.0. Int. J. Mol. Sci. 2024, 25, 5339. [Google Scholar] [CrossRef]
  75. Bollen, E.; Prickaerts, J. Phosphodiesterases in neurodegenerative disorders. IUBMB Life 2012, 64, 965–970. [Google Scholar] [CrossRef] [PubMed]
  76. Epstein, P.M.; Basole, C.; Brocke, S. The Role of PDE8 in T Cell Recruitment and Function in Inflammation. Front. Cell Dev. Biol. 2021, 9, 636778. [Google Scholar] [CrossRef]
  77. Chimienti, F.; Cavarec, L.; Vincent, L.; Salvetat, N.; Arango, V.; Underwood, M.D.; Mann, J.J.; Pujol, J.F.; Weissmann, D. Brain region-specific alterations of RNA editing in PDE8A mRNA in suicide decedents. Transl. Psychiatry 2019, 9, 91. [Google Scholar] [CrossRef]
  78. Guo, J.Y.; Ragland, J.D.; Carter, C.S. Memory and cognition in schizophrenia. Mol. Psychiatry 2019, 24, 633–642. [Google Scholar] [CrossRef]
  79. Zhan, A.; Xu, X.; Chen, L.; Wang, X.; Yanfeng, X.; Dan, W.; Zhan, Y.; Shi, Q. Decreased expression of Gab2 in patients with temporal lobe epilepsy and pilocarpine-induced rat model. Synapse 2014, 68, 168–177. [Google Scholar] [CrossRef] [PubMed]
  80. Malagrino, F.; Puglisi, E.; Pagano, L.; Travaglini-Allocatelli, C.; Toto, A. GRB2: A dynamic adaptor protein orchestrating cellular signaling in health and disease. Biochem. Biophys. Rep. 2024, 39, 101803. [Google Scholar] [CrossRef]
  81. Chen, X.X.; Guo, R.R.; Cao, X.P.; Tan, L.; Tan, L.; Alzheimer’s Disease Neuroimaging Initiative. The impact of GAB2 genetic variations on cerebrospinal fluid markers in Alzheimer’s disease. Ann. Transl. Med. 2018, 6, 171. [Google Scholar] [CrossRef]
  82. Balu, D.T.; Coyle, J.T. Neuroplasticity signaling pathways linked to the pathophysiology of schizophrenia. Neurosci. Biobehav. Rev. 2011, 35, 848–870. [Google Scholar] [CrossRef]
  83. Gao, K.; Lin, Z.; Wen, S.; Jiang, Y. Potassium channels and epilepsy. Acta Neurol. Scand. 2022, 146, 699–707. [Google Scholar] [CrossRef]
  84. Chen, X.; Feng, Y.; Quinn, R.J.; Pountney, D.L.; Richardson, D.R.; Mellick, G.D.; Ma, L. Potassium Channels in Parkinson’s Disease: Potential Roles in Its Pathogenesis and Innovative Molecular Targets for Treatment. Pharmacol. Rev. 2023, 75, 758–788. [Google Scholar] [CrossRef]
  85. Zhou, S.Z.; Shen, L.; Fu, Z.B.; Li, H.; Pan, Y.L.; Yu, R.Z. Exploring the common diagnostic gene KCNJ15 and shared pathway of ankylosing spondylitis and ulcerative colitis through integrated bioinformatics. Front. Physiol. 2023, 14, 1146538. [Google Scholar] [CrossRef]
  86. Langmead, B.; Salzberg, S.L. Fast gapped-read alignment with Bowtie 2. Nat. Methods 2012, 9, 357–359. [Google Scholar] [CrossRef]
  87. Li, H. A statistical framework for SNP calling, mutation discovery, association mapping and population genetical parameter estimation from sequencing data. Bioinformatics 2011, 27, 2987–2993. [Google Scholar] [CrossRef]
  88. Li, H.; Handsaker, B.; Wysoker, A.; Fennell, T.; Ruan, J.; Homer, N.; Marth, G.; Abecasis, G.; Durbin, R.; Genome Project Data Processing, S. The Sequence Alignment/Map format and SAMtools. Bioinformatics 2009, 25, 2078–2079. [Google Scholar] [CrossRef]
  89. Gentleman, R.C.; Carey, V.J.; Bates, D.M.; Bolstad, B.; Dettling, M.; Dudoit, S.; Ellis, B.; Gautier, L.; Ge, Y.; Gentry, J.; et al. Bioconductor: Open software development for computational biology and bioinformatics. Genome Biol. 2004, 5, R80. [Google Scholar] [CrossRef]
  90. R Development Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. 2010. Available online: https://www.r-project.org/ (accessed on 10 November 2024).
  91. Benjamini, Y.; Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J. R. Stat. Soc. Ser. B (Methodol.) 1995, 57, 289–300. [Google Scholar] [CrossRef]
  92. Meyer, P.E. Information-Theoretic Variable Selection and Network Inference from Microarray Data. Ph.D. Thesis, Universite Libre de Bruxelles, Brussels, Belgium, 2008. [Google Scholar]
  93. Box, G.E.P.; Cox, D.R. An Analysis of Transformations. J. R. Stat. Soc. Ser. B (Methodol.) 1964, 26, 211–243. [Google Scholar] [CrossRef]
  94. Paluszynska, A.; Biecek, P.; Jiang, Y. randomForestExplainer: Explaining and Visualizing Random Forests in Terms of Variable Importance. R Package Version 0.10.1. 2022. Available online: https://CRAN.R-project.org/package=randomForestExplainer (accessed on 10 November 2024).
  95. McCracken, D.D. The Monte Carlo Method. Sci. Am. 1955, 192, 90–96. [Google Scholar] [CrossRef]
  96. Kramar, A.; Faraggi, D.; Fortuné, A.; Reiser, B. mROC: A computer program for combining tumour markers in predicting disease states. Comput. Methods Programs Biomed. 2001, 66, 199–207. [Google Scholar] [CrossRef] [PubMed]
  97. Breiman, L. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef]
  98. Venables, W.N.; Ripley, B.D. Modern Applied Statistics with S, 4th ed.; Springer: New York, NY, USA, 2022; Available online: https://www.stats.ox.ac.uk/pub/MASS4/ (accessed on 10 November 2024)ISBN 0-387-95457-0.
  99. Kuhn, M. Building Predictive Models in R Using the caret Package. J. Stat. Softw. 2008, 28, 1–26. [Google Scholar] [CrossRef]
Ijms 25 12981 g001
Figure 1. TEI analysis differentiates between the four groups (CTRL, BD, SZ, and SA). (A) TEI values comparing Ctrl, SZ, SA, and BD (° p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001). (B) Differential TEI analysis per-group comparison, showing FDR-adjusted p-values (bold values are <0.1). CTRL: Control; BD: Bipolar disorder; SA: Schizoaffective disorder; SZ: Schizophrenia.
Figure 1. TEI analysis differentiates between the four groups (CTRL, BD, SZ, and SA). (A) TEI values comparing Ctrl, SZ, SA, and BD (° p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001). (B) Differential TEI analysis per-group comparison, showing FDR-adjusted p-values (bold values are <0.1). CTRL: Control; BD: Bipolar disorder; SA: Schizoaffective disorder; SZ: Schizophrenia.
Ijms 25 12981 g001
Ijms 25 12981 g002
Figure 2. RNA editing signatures to discriminate CTRL vs. BD vs. SA vs. SZ. (A) Scatterplot 3D of the best combination to discriminate CTRL vs. BD vs. SZ + SA subgroups. (B) Scatterplot 3D of the best combination to discriminate CTRL vs. SA vs. SZ subgroups. (C) Table showing per-group comparison performances and decision threshold (cut-off). Z represents a combination of 32 significant RNA editing biomarkers. CTRL: Control; BD: Bipolar disorder; SA: Schizoaffective disorder; SZ: Schizophrenia; AUC: AUC ROC; Sp: Specificity; Se: Sensitivity; PPV: Positive predictive value; NPV: Negative predictive value.
Figure 2. RNA editing signatures to discriminate CTRL vs. BD vs. SA vs. SZ. (A) Scatterplot 3D of the best combination to discriminate CTRL vs. BD vs. SZ + SA subgroups. (B) Scatterplot 3D of the best combination to discriminate CTRL vs. SA vs. SZ subgroups. (C) Table showing per-group comparison performances and decision threshold (cut-off). Z represents a combination of 32 significant RNA editing biomarkers. CTRL: Control; BD: Bipolar disorder; SA: Schizoaffective disorder; SZ: Schizophrenia; AUC: AUC ROC; Sp: Specificity; Se: Sensitivity; PPV: Positive predictive value; NPV: Negative predictive value.
Ijms 25 12981 g002
Table 1. Demographic and clinical characteristics of the study population. Data are presented as mean ± SEM. p-values for age were obtained using Student’s t-test, while p-values for sex and psychotropic treatments were calculated using the chi-square test (bold values are ≤0.05). Controls: Healthy volunteers; SZ: Schizophrenia patients; SA: Schizoaffective patients; BD: Bipolar disorder patients.
Table 1. Demographic and clinical characteristics of the study population. Data are presented as mean ± SEM. p-values for age were obtained using Student’s t-test, while p-values for sex and psychotropic treatments were calculated using the chi-square test (bold values are ≤0.05). Controls: Healthy volunteers; SZ: Schizophrenia patients; SA: Schizoaffective patients; BD: Bipolar disorder patients.
Total SampleControlsSZSABD
Number, n16985311439
Age
Age (min–max)18–6418–6322–6235–6318–64
Age (mean ± SD)42.2 ± 10.540.3 ± 11.242.8 ± 9.845.6 ± 6.744.7 ± 9.7
  p value (vs. Ctrl) 0.240.020.03
  p value (vs. SZ) 0.270.42
  p value (vs. SA) 0.71
Sex
Male (n (%))91 (53.8)47 (55.3)23 (74.2)10 (71.4)11 (28.2)
Female (n (%))78 (46.1)38 (44.7)8 (25.8)4 (28.6)28 (71.8)
  p value (vs. Ctrl) <0.00010.001<0.0001
  p value (vs. SZ) 0.52<0.0001
  p value (vs. SA) <0.0001
Psychotropic Treatments
Anxiolytics (n (%))7(4.1)05 (16.1)2 (14.3)0 (0.0)
Hypnotics/Sedatives (n (%))1 (0.59)01 (3.2)0 (0.0)0 (0.0)
Antidepressants (n (%))31 (18.3)018 (58.0)8 (57.1)5 (12.8)
Antipsychotics (n (%))71(42.0)029 (93.6)14 (100.0)28 (71.8)
Antiepileptics (n (%))43 (25.4)012 (38.7)5 (35.7)26 (66.7)
  p value (vs. SZ) 0.39<0.0001
  p value (vs. SA) <0.0001
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Checa-Robles, F.J.; Salvetat, N.; Cayzac, C.; Menhem, M.; Favier, M.; Vetter, D.; Ouna, I.; Nani, J.V.; Hayashi, M.A.F.; Brietzke, E.; et al. RNA Editing Signatures Powered by Artificial Intelligence: A New Frontier in Differentiating Schizophrenia, Bipolar, and Schizoaffective Disorders. Int. J. Mol. Sci. 2024, 25, 12981. https://doi.org/10.3390/ijms252312981

AMA Style

Checa-Robles FJ, Salvetat N, Cayzac C, Menhem M, Favier M, Vetter D, Ouna I, Nani JV, Hayashi MAF, Brietzke E, et al. RNA Editing Signatures Powered by Artificial Intelligence: A New Frontier in Differentiating Schizophrenia, Bipolar, and Schizoaffective Disorders. International Journal of Molecular Sciences. 2024; 25(23):12981. https://doi.org/10.3390/ijms252312981

Chicago/Turabian Style

Checa-Robles, Francisco J., Nicolas Salvetat, Christopher Cayzac, Mary Menhem, Mathieu Favier, Diana Vetter, Ilhème Ouna, João V. Nani, Mirian A. F. Hayashi, Elisa Brietzke, and et al. 2024. "RNA Editing Signatures Powered by Artificial Intelligence: A New Frontier in Differentiating Schizophrenia, Bipolar, and Schizoaffective Disorders" International Journal of Molecular Sciences 25, no. 23: 12981. https://doi.org/10.3390/ijms252312981

APA Style

Checa-Robles, F. J., Salvetat, N., Cayzac, C., Menhem, M., Favier, M., Vetter, D., Ouna, I., Nani, J. V., Hayashi, M. A. F., Brietzke, E., & Weissmann, D. (2024). RNA Editing Signatures Powered by Artificial Intelligence: A New Frontier in Differentiating Schizophrenia, Bipolar, and Schizoaffective Disorders. International Journal of Molecular Sciences, 25(23), 12981. https://doi.org/10.3390/ijms252312981

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop