Introduction

Prostate cancer (PC) is one of the most common malignant tumors in men and the second leading cause of cancer-related mortality among males1. Research indicates that approximately 15% to 25% of PC patients suffer from clinical depression, with the incidence of depression in patients with castration-resistant prostate cancer (CRPC) being about 80% higher than that in the general male population2,3. Due to the complex mechanisms of drug resistance in CRPC, along with a median survival of only 14 months1, these patients also exhibit a significantly increased risk of major depressive disorder (MDD) and suicide4. Furthermore, depression can adversely affect the quality of life and treatment adherence of patients, interfere with the therapeutic efficacy of PC, and exacerbate the disease burden2. Conversely, antidepressant treatment has been shown to reduce the risk of PC recurrence5. Therefore, addressing depression in patients with PC holds significant clinical importance.

Existing studies suggest that MDD and CRPC are interconnected through complex molecular mechanisms, including immune dysregulation, Focal Adhesion Kinase (FAK) activation, neuroendocrine disruption, and inflammatory responses. Dysregulation of immune function is one of the core mechanisms linking MDD and CRPC. In mouse models, depressive states alter the immune profile, promoting an immunosuppressive tumor microenvironment characterized by dysfunctional T cells and macrophages and elevated inflammatory cytokines such as TNF-α and IL-6, thereby stimulating prostate cancer cell proliferation, survival, and invasion6,7. Changes in neuroendocrine and inflammatory pathways also represent significant mechanisms connecting MDD and CRPC. Depression-induced neurotransmitter imbalances and activation of the hypothalamic-pituitary-adrenal (HPA) axis lead to elevated glucocorticoid levels, which can directly regulate androgen receptors or estrogen-related receptor (ER) signaling in prostate cancer cells, promoting tumor growth and resistance to hormonal therapies8. Additionally, the systemic inflammation triggered by depressive states, accompanied by increased levels of cytokines (such as TNF-α, IL-6, and CRP), can activate transcription factors like nuclear factor kappa B (NF-κB), driving tumor angiogenesis, cell survival, and anti-apoptotic mechanisms7. Furthermore, a clinical study involving 98 PC patients found that those with severe depression exhibited higher levels of FAK, which were significantly correlated with aggressive tumor characteristics. This suggests that FAK may serve as a potential molecular target for the exacerbation of prostate cancer induced by depression, with its pathway activation partially elucidating the mechanisms of tumor progression in depressive states9.

The aforementioned mechanisms are interrelated and contribute to a vicious cycle. However, the mechanisms underlying the comorbidity of MDD and CRPC remain incompletely understood. For instance, beyond the mechanisms already discussed, does MDD promote the progression of CRPC through other critical pathways? Can these new mechanisms provide potential solutions for CRPC resistance? Additionally, apart from the hormonal changes associated with androgen deprivation therapy, does CRPC induce or exacerbate depressive symptoms through other mechanisms? Can these new mechanisms help reduce the risk of suicide in CRPC patients? Therefore, exploring new mechanisms is particularly important for addressing the comorbidity of MDD and CRPC.

Despite recognizing the detrimental effects of the MDD/CRPC comorbidity, effective intervention strategies based on molecular mechanisms remain very limited. A pressing issue is how to translate mechanistic research into potential therapeutic targets and drugs to break this vicious cycle. This study aims to systematically elucidate the molecular mechanisms underlying the comorbidity of CRPC and MDD and to explore potential therapeutic strategies. The research objectives include: integrating and analyzing transcriptomic data from CRPC and MDD to reveal their shared molecular basis and core regulatory networks, with the goal of identifying novel pathogenic pathways; applying Mendelian randomization (MR) analysis to validate the causal relationships between hub gene polymorphisms and both CRPC and MDD; and predicting potential therapeutic drugs based on the identified comorbid hub targets, utilizing molecular docking techniques for preliminary validation. This systematic research approach may uncover new pathways that have not been previously addressed in existing studies, providing a basis for the development of novel intervention strategies that could improve the symptoms and treatment outcomes of MDD/CRPC comorbidity (Fig. 1).

Fig. 1
figure 1

Research on the potential mechanisms and therapeutic drug for the co-occurrence of major depressive disorder in castration-resistant prostate cancer.

Full size image

Results

Co-occurrence gene analysis of CRPC/MDD

The GSE70768 and GSE98793 datasets were downloaded from the GEO database for screening DEGs in patients with CRPC or MDD. The datasets were normalized and standardized using the “limma” package, resulting in 991 DGEs for CRPC, with 318 upregulated genes and 673 downregulated genes (Fig. 2A-B). Additionally, 471 DGEs were obtained for MDD, with 171 upregulated genes and 300 downregulated genes (Fig. 2H-I). Using WGCNA, 42 CRPC-related gene modules were identified in the GSE70768 dataset (Fig. 2C-F), and 11 MDD-related gene modules were identified in the GSE98793 dataset (Fig. 2J-M). The “brown” module showed the highest correlation with CRPC (r = 0.58), containing 2559 genes (Fig. 2G), while the “pink” module showed the highest correlation with MDD (r = -0.22), containing 2603 genes (Fig. 2N). Finally, 11 co-occurrence genes for CRPC/MDD were obtained through overlapping DGEs and co-expressed genes (Fig. 2O).

Fig. 2
figure 2

Analysis of comorbid genes between CRPC and MDD. (A, B) Differential gene expression analysis of CRPC. (C–G) Weighted gene co-expression network analysis of CRPC. (H, I) Differential gene expression analysis of MDD. (J–N) Weighted gene co-expression network analysis of MDD. (O) Identification of comorbid genes between CRPC and MDD.

Full size image

PPI network construction and GO, KEGG analysis

The PPI network was constructed and visualized using the GeneMania database, as shown in Fig. 3A. GO terms mainly involved Ras protein signal transduction, small GTPase-mediated signal transduction, gonad development, hormone metabolism, NADP + 1-oxidoreductase activity, and nucleoside-triphosphatase regulator activity (Fig. 3B). KEGG pathways mainly included Folate biosynthesis, Apelin signaling pathway, Histidine metabolism, and Rap1 signaling pathway (Fig. 3C).

Fig. 3
figure 3

Construction of PPI network and GO, KEGG analysis. (A) PPI network diagram of comorbid genes in CRPC/MDD. (B) Bubble chart of GO functional enrichment analysis. (C) Bubble chart of KEGG pathway enrichment analysis (top 15).

Full size image

Identification and validation of hub genes

In the field of machine learning, we utilized the Random Forest (RF) algorithm to predict the top 10 important genes ranked by Mean Decrease Accuracy in both metastatic CRPC and MDD. Using a LASSO regression model, we identified 10 significant genes for CRPC and 6 significant genes for MDD. Additionally, based on a Support Vector Machine (SVM) model, we selected 7 optimal feature genes for CRPC and 11 optimal feature genes for MDD (Fig. 4A-B). The overlapping genes identified by the three machine learning methods were considered potential diagnostic biomarkers for the comorbidity of CRPC and MDD, resulting in the identification of three diagnostic markers: AUTS2, AOC1, and ANKRD37 (Fig. 4C).

SHAP analysis indicated that in CRPC, the importance ranking of hub genes was as follows: ANKRD37, AOC1, and AUTS2 (Fig. 4D); while in MDD, the ranking was ANKRD37, AUTS2, and AOC1. This finding enhances the transparency of the model and the interpretability of the results (Fig. 4E).

Subsequently, we assessed the diagnostic performance of the hub genes by plotting Receiver Operating Characteristic (ROC) curves. Internal dataset validation revealed that the hub genes exhibited significant diagnostic value in both diseases (AUC > 0.7). External validation was conducted using the GSE35988 (Platform GPL9075) and GSE19738 datasets for CRPC and MDD, respectively. The results demonstrated that the diagnostic markers had significant diagnostic value in CRPC (AUC > 0.7), outperforming their diagnostic performance in MDD (AUC > 0.5) (Fig. 4F-G). Decision Curve Analysis (DCA) curves indicated that in the MDD training cohort, the hub gene model achieved a peak net benefit of 0.4 within the threshold probability range of 16% to 96% (Fig. 4I); in the CRPC training cohort, a peak net benefit of 0.13 was observed within the threshold probability range of 4% to 57% (Fig. 4H), suggesting the potential value of hub genes in guiding clinical intervention strategies.

Finally, we employed the Kaplan-Meier method to construct survival curves to evaluate the prognostic value of hub genes in CRPC. The results indicated that AOC1 and ANKRD37 are significant prognostic biomarkers (P < 0.05), while AUTS2 did not demonstrate prognostic diagnostic value in CRPC (P = 0.33) (Fig. 4J).

Fig. 4
figure 4

Identification and validation of hub genes. (A) LASSO, SVM, and RF algorithm for CRPC. (B) LASSO, SVM, and RF algorithm for MDD. (C) Three hub genes were obtained after overlapping machine learning algorithms. (D) SHAP interpretation of model for CRPC. (E) SHAP interpretation of model for MDD. (F) ROC curve of hub genes in the internal and external dataset of CRPC. (G) ROC curve of hub genes in the internal and external dataset of MDD. (H) Decision curve analysis of CRPC. (I) Decision curve analysis of MDD. (J) Survival curves of hub genes in CRPC.

Full size image

Gene set enrichment analysis (GSEA)

We employed GSEA analysis to evaluate the pathways and molecular mechanisms associated with the hub genes that regulate CRPC and MDD. By combining the Normalized Enrichment Score (NES), False Discovery Rate (FDR), and p-value of pathways, we identified five significant signaling pathways (Fig. 5A-B). Our results indicate that the androgen response, apoptosis, inflammatory response, TNF-α signaling via the NF-κB pathway, as well as the IL-2/STAT-5 pathway, play crucial roles in the pathogenesis of CRPC and MDD.

Analysis of the correlation between hub genes and immune cells

We compared the differences in immune cell infiltration scores between the normal group and the disease group. The ssGSEA results showed significant differences in the levels of T cells in CRPC and MDD (Fig. 5C-D and F-G). Specifically, six immune cell infiltration scores (CD56 bright natural killer cell, Central memory CD4 T cell, Central memory CD8 T cell, MDSC, Regulatory T cell, Type 17 T helper cell) showed significant differences between the normal and disease groups in CRPC, while five immune cell infiltration scores (Activated dendritic cell, Effector memory CD8 T cell, Macrophage, Monocyte, Natural killer T cell) showed significant differences between the normal and disease groups in MDD.

We constructed an immune landscape of hub genes (Fig. 5E-H). However, we did not observe significant correlations between the expression levels of hub genes and immune cell infiltration scores. In MDD, we only found a correlation between the expression level of ANKRD37 and the infiltration score of T follicular helper cells, and a correlation between the expression level of AOC1 and the infiltration scores of CD56 bright natural killer cell, Effector memory CD8 T cell, Immature B cell, and Natural killer T cell. In CRPC, we did not observe any significant correlations.

Fig. 5
figure 5

GSEA and hub genes immune correlation analysis. (A, B) GSEA analysis of CRPC and MDD. (C–E) Immune infiltration analysis of CRPC. (F–H) Immune infiltration analysis of MDD.

Full size image

Mendelian randomization (MR)

We employed a two-sample MR analysis to explore the causal relationship between the Hub gene and the risk of CRPC and MDD. Due to insufficient SNPs for ANKRD37, we only conducted two-sample MR analysis for AOC1 and AUTS2. IVW analysis only found suggestive evidence of an association between AUTS2 and MDD risk (OR = 1.016, 95% CI = 0.019–0.033, p = 0.031). Sensitivity analysis indicated that the overall estimate was not driven by any single SNP (Fig. 6).

Fig. 6
figure 6

Mendelian randomization (MR) analysis. (A–C) MR of the causal relationship between AOC1 and PC. (D–F) MR of the causal relationship between AOC1 and MDD. (G–I) MR of the causal relationship between AUTS2 and PC. (J–L) MR of the causal relationship between AUTS2 and MDD.

Full size image

Drug prediction and molecular Docking

Using the DSigDB database in the Enrichr platform, we predicted potential therapeutic drugs for the comorbidity of CRPC/MDD. Figure 7A displays the top 20 potential therapeutic drugs based on their combined score and adjusted P value, with WP1066 having the highest combined score and considered the most promising drug for treating the comorbidity of CRPC/MDD. To elucidate the binding of WP1066 with the hub genes, we further validated it using molecular docking techniques (Fig. 7B-D). The molecular docking binding energies of AUTS2, AOC1, and ANKRD37 with WP1066 are shown in Table 1, with all three groups having binding energies lower than − 5 kcal·mol-1, indicating good binding between the ligands. The binding energy between WP1066 and AOC1 is -10.2 kcal·mol-1, suggesting a strong binding between WP1066 and AOC1.

Fig. 7
figure 7

Drug prediction and molecular docking. (A) Predicted drugs (top 20 rankings). (B–D) Molecular docking results of AUTS2, AOC1, ANKRD37, and WP1066.

Full size image
Table 1 The binding affinity between small molecule ligands and protein.
Full size table

Discussion

There is a significant clinical association between prostate cancer (PC) and depression. Epidemiological studies indicate that the prevalence of depression among prostate cancer patients ranges from 15% to 25%, with risk factors including androgen deprivation therapy (ADT), psychosocial factors, and advanced disease stages2,10. Furthermore, depression can increase the mortality risk in patients with castration-resistant prostate cancer (CRPC) by up to 50% 11. The comorbidity of major depressive disorder (MDD) and CRPC has become an urgent public health issue, yet effective intervention strategies based on molecular mechanisms remain limited. Translating mechanistic research into potential therapeutic targets and drugs to disrupt the vicious cycle of comorbidity is a critical challenge that needs to be addressed. This study is the first to integrate transcriptomic data and employ bioinformatics methods, Mendelian randomization analysis, and molecular docking techniques to explore the comorbid mechanisms of CRPC and MDD at the molecular level. Through machine learning, we successfully identified and preliminarily validated hub genes such as AUTS2, AOC1, and ANKRD37 for their diagnostic and prognostic value in the CRPC/MDD comorbidity. Building on this foundation, we innovatively predicted the JAK/STAT pathway inhibitor WP1066 as a potential therapeutic agent for this comorbidity and elucidated the synergistic roles of key pathways, including immune inflammation, androgen response, apoptosis, and JAK/STAT signaling in the context of comorbidity. These findings provide a new molecular mechanistic basis and candidate drugs for the diagnosis, prognostic assessment, and targeted treatment of CRPC/MDD comorbidity, holding potential clinical translational significance. It is important to note that this study is primarily exploratory, aimed at generating new hypotheses and potential targets, and the identified molecular mechanisms, causal relationships, and drug predictions require validation through future experimental studies.

In this study, we identified 11 genes associated with the comorbidity of CRPC and MDD. Using machine learning techniques, we ultimately identified three hub genes—AUTS2, AOC1, and ANKRD37—that demonstrated strong diagnostic value in both diseases, suggesting their potential as new targets for clinical diagnosis, treatment, and prevention. Notably, AOC1 and ANKRD37 showed significant relevance for prognostic assessment in CRPC. Furthermore, our two-sample Mendelian randomization (MR) analysis provided preliminary evidence linking AUTS2 to the risk of MDD. However, due to the statistical nature of MR analysis, the results may be influenced by unmeasured confounding factors and the limitations of the GWAS data used. Therefore, these preliminary findings require validation through functional experiments.

The AUTS2 gene is highly expressed in key areas of the brain, such as the cerebral cortex, hippocampus, and thalamus, and is localized in neurons12,13, participating in neurodevelopment12,14. Mutations in the AUTS2 gene can lead to neurodevelopmental disorders14,15,16,17,18, increase the risk of depression19,20,21, and affect the efficacy of antidepressant drugs22. In cancer, overexpression of AUTS2 can lead to abnormal differentiation and proliferation of B cells and T cells, promoting tumor initiation and progression23,24,25,26,27. Conversely, reduced expression of AUTS2 in ccRCC and malignant NK cells may be associated with the malignancy and poor prognosis of tumors28,29. The relationship between the AUTS2 gene and depression and cancer exists, but how it mediates CRPC and MDD is still unclear. It is worth noting that AUTS2 is closely related to immune-inflammatory responses30,31, and this study found that immune-inflammatory responses may mediate the comorbidity of CRPC/MDD. Further research on the relationship between AUTS2 and immune-inflammatory responses may be of significant importance in elucidating the mechanism of CRPC/MDD comorbidity.

Amine oxidase copper-containing 1 (AOC1), also known as diamine oxidase (DAO)32,33, is closely associated with the occurrence and development of cancers such as hepatocellular carcinoma34, gastric cancer35, central nervous system lymphoma36, and neuroglial tumors37, and is considered a candidate new antigen for use in lung squamous cell carcinoma vaccines38. Studies have shown that AOC1 is downregulated in prostate cancer, and its expression level is negatively correlated with the malignancy of tumors. Upregulation of AOC1 may inhibit the proliferation and migration of prostate cancer through ferroptosis, suggesting that AOC1 may become a new strategy for treating prostate cancer39. In addition, AOC1 (also known as DAO) has a potential association with depression. Abnormal serum levels of DAO may be related to cognitive impairments in dementia, Alzheimer’s disease, and prenatal behavioral disorders40,41,42. A study evaluated independent predictive factors for anhedonia in MDD patients and found that plasma DAO levels were lower in MDD patients43, indicating a potential connection between DAO and MDD. Abnormal expression of AOC1 may lead to cancer and neuropsychiatric disorders. Further research on AOC1, especially its specific mechanisms related to the comorbidity of CRPC/MDD, will provide new targets and strategies for the diagnosis, treatment, and prevention of related diseases.

ANKRD37 is associated with various cancers and neurological disorders. It is upregulated in colon cancer and renal clear cell carcinoma44,45, while downregulated in aggressive B-cell lymphoma, correlating with tumor metastasis and clinical outcomes46. Moreover, overexpression of ANKRD37 is causally linked to reduced hippocampal volume and may be associated with late-onset Alzheimer’s disease47. Despite the potential associations of ANKRD37 with cancer and neurological disorders, further research is needed to explore its relationship with CRPC and MDD. A comprehensive investigation into the role of ANKRD37 in CRPC and MDD, along with in vitro and in vivo experimental validation, will contribute to the development of novel therapeutic strategies for CRPC/MDD.

Through GSEA analysis, we found that biological processes such as inflammatory responses, androgen responses, apoptosis, TNF-αsignaling via NF-κB pathway, and the IL-2/STAT-5 signaling pathway play crucial roles in the pathogenesis of CRPC and MDD, providing new insights into the mechanisms of comorbidity between CRPC and MDD. It is important to emphasize that the biological pathways and mechanisms revealed by these enrichment analyses require further validation through subsequent experimental studies.

There may exist a complex inflammation-immune-neuroendocrine axis between CRPC and MDD, in which testosterone, cytokines, and neurotransmitters interact to jointly drive the comorbidity process. Spoletini et al. proposed a bidirectional mechanism of interaction between cancer and depression, emphasizing the role of inflammation and immune activity48. Firstly, the interaction between PC and MDD involves complex inflammatory mechanisms. Cancer patients with comorbid depression exhibit higher levels of pro-inflammatory cytokines, suggesting that depression may exacerbate the inflammatory response in cancer patients49. Testosterone can inhibit pro-inflammatory factors by binding to immune cell receptors, these anti-inflammatory effects of androgens have been observed not only in the prostate but also in the male bladder50,51. Although PC patients eventually develop CRPC resistant to traditional ADT, current clinical guidelines still recommend continuous ADT to maintain serum testosterone levels below 50 ng/dL 52. However, it is worth noting that castration levels of testosterone may exacerbate inflammation and depressive symptoms, which constitute an important clinical consideration in the treatment decisions for CRPC patients53.

Secondly, the interaction between PC and MDD also involves complex immunological mechanisms. Studies have shown that PC cells can produce chemicals that induce MDD, and MDD can in turn promote the recruitment of myeloid-derived suppressor cells (MDSCs) to the tumor microenvironment and secrete interleukin-6 (IL-6), thereby activating signal transducer and activator of transcription 3 (STAT3) within cancer cells6,54. Clarifying the immunological relationship between CRPC and MDD may inspire new therapeutic targets. This study explored the immunological association between CRPC and MDD, and although our ssGSEA analysis showed significant differences in T cell levels between CRPC and MDD. Previous studies have identified androgen receptor activation as a key immune regulator within the prostate gland. Specifically, dihydrotestosterone exhibits strong anti-inflammatory properties by reducing the activity of Th1/Th17 cells, which are the predominant T-cell subtypes in benign prostatic hyperplasia55,56. Our study revealed a significantly lower level of Th17 cells in CRPC compared to the control group. Conversely, although not statistically significant, there was a trend towards increased expression of Th2 lymphocytes in CRPC compared to the control group. Elevated Th2 levels are known to be associated with cancer, as these cells can create an immunosuppressive microenvironment within tumors57.

Inflammation of the bone marrow system promotes tumor progression by affecting T cell activity, and indicators reflecting activated bone marrow inflammation can be used to predict cancer prognosis58,59. Recent studies have shown that bone marrow inflammation is associated with the progression and drug resistance of CRPC60. CXCR2 is a key driver of tumor-induced bone marrow inflammation, with its ligands upregulated in tumors, recruiting bone marrow cells to the tumor site, exacerbating bone marrow inflammation, and promoting tumor progression and treatment resistance60. Targeting bone marrow chemotaxis, particularly by inhibiting CXCR2, could be a novel strategy for CRPC treatment, especially in patients resistant to androgen receptor blockade agents60,61. NF-κB has been shown to regulate the transcription of CXCR2 ligands CXCL1, CXCL2, and CXCL8 62,63,64. Through gene set enrichment analysis, this study found that TNF-α signaling via NF-κB pathway plays a critical role in the pathogenesis of CRPC and MDD. The NF-κB signaling pathway is activated in various cancers and is considered a key mechanism in inflammation and cancer development65. Studies suggest that constitutive activation of NF-κB may be related to prostate cancer bone metastasis66,67. Additionally, NF-κB can upregulate the expression of CXCR2 and CXCR4 ligands, suppress tumor immunity, and promote tumor progression68,69. NF-κB plays a crucial role in regulating cell apoptosis and drug resistance, and targeting the NF-κB pathway inhibition may be a potential strategy to overcome CRPC resistance and enhance anticancer efficacy65.

The pathological mechanism of depression is closely related to excessive apoptosis of neurons in specific brain regions70,71. Neuronal apoptosis exacerbates neuronal inflammation by inhibiting the PI3K/Akt pathway, promoting the progression of depression. Conversely, apoptosis inhibition shows potential therapeutic benefits by alleviating neuronal inflammation and improving depressive symptoms71,72,73,74. Antidepressant treatment has shown therapeutic potential by upregulating the expression of Bcl-2 family anti-apoptotic proteins71. However, this anti-apoptotic effect may pose challenges in the treatment of CRPC patients. Specifically, inducing cell apoptosis is a strategy in the treatment of CRPC drug resistance, which is expected to inhibit the growth of CRPC cells75,76. Nevertheless, the increased expression of anti-apoptotic proteins may promote resistance to apoptosis in PC cells, disrupting the balance between pro-apoptotic and anti-apoptotic pathways in PC cells, especially in CRPC77,78. Failure to address this conflict may result in suboptimal treatment outcomes and could even exacerbate certain aspects of the disease. Therefore, investigating balanced strategies for apoptosis regulation in the treatment of comorbid CRPC/MDD, such as targeting tissue-specific apoptosis pathways or developing novel drug delivery systems, may be crucial for establishing new therapeutic approaches.

Through gene enrichment analysis, we have found that the IL-2/STAT-5 pathway may play an important role in the pathogenesis of CRPC and MDD. Additionally, another reason we focus on the IL-2/STAT-5 pathway is that drug prediction has identified the JAK/STAT pathway inhibitor WP1066 as a potential therapeutic agent for the treatment of comorbid CRPC/MDD.

The IL-2/STAT-5 pathway is a branch of the JAK/STAT pathway. IL-2 regulates the activity of STAT-5 by activating the JAK protein kinase, playing a crucial role in immune regulation and inflammatory responses79,80. The JAK/STAT signaling pathway is closely associated with comorbidities of CRPC/MDD. The lineage plasticity of PC depends on the inflammatory signal transduction of the JAK/STAT pathway. Pharmacological inhibition of the JAK-STAT signaling pathway sensitizes PC organoids to androgen receptor signaling inhibitors81. Specifically, STAT3 and STAT5 are involved in the androgen resistance of CRPC, which is of significance for developing drugs to combat drug resistance in CRPC82. Additionally, JAK-STAT signal transduction plays a key role in depression, regulating synaptic function in neurons and providing new targets and strategies for the treatment of depression83.

Based on the drug prediction analysis using hub genes, this study identified the JAK/STAT pathway inhibitor WP1066 as a promising therapeutic candidate for managing CRPC in conjunction with MDD. Subsequent molecular docking studies predicted favorable binding interactions between WP1066 and the hub genes. However, it is important to note that molecular docking results are purely computational predictions, and the binding affinity indicated does not necessarily correlate with actual biological activity or therapeutic efficacy. The true efficacy of WP1066 must be validated through rigorous in vitro and in vivo experiments.

The unique aspect of WP1066 lies in its ability to simultaneously target key signaling pathways associated with tumor progression and neuropsychiatric symptoms (JAK/STAT), offering a potential “dual-target” therapeutic strategy for patients with CRPC accompanied by depressive symptoms. WP1066 demonstrates significant therapeutic potential across various cancers by inhibiting STAT3 phosphorylation and promoting apoptosis, effectively suppressing the viability and invasiveness of cancer cells, including bladder cancer, oral squamous cell carcinoma, and renal cell carcinoma84,85,86,87. In PC, WP1066 effectively blocks disease progression by inhibiting the expression of HIF-1α 88. Furthermore, WP1066 has been shown to reverse chemoresistance in breast and ovarian cancer cells89,90, indicating its potential therapeutic value in overcoming cancer resistance. Recent studies also suggest that WP1066 exerts antitumor effects by modulating the tumor immune microenvironment91. The antitumor activity of WP1066 in CRPC, along with its ability to overcome resistance and its immune-regulatory potential, warrants further investigation and clinical development. In addition to its antitumor effects, WP1066 has also shown positive impacts in the neuropsychiatric domain. Recent research indicates that WP1066 exhibits the potential to improve cognitive function by inhibiting STAT3 activity92. Given that STAT3 plays a crucial role in neuroinflammation and neurodegenerative diseases, its inhibition may help ameliorate the neurophysiological processes associated with depression93. In animal models of epilepsy, WP1066 significantly inhibited STAT3 phosphorylation, thereby alleviating anxiety behaviors and memory deficits induced by seizures94. Considering that patients with CRPC often experience depressive symptoms, which are frequently accompanied by cognitive dysfunction and anxiety, the neuroprotective and cognitive-enhancing effects of WP1066 are directly relevant to alleviating depressive symptoms in these patients. It is important to emphasize that current evidence regarding the potential therapeutic effects of WP1066 on MDD remains indirect and speculative, and its precise effects and mechanisms in psychiatric disorders require clarification through future direct experimental studies.

Given the significant antitumor activity of WP1066 in CRPC and its potential to improve cognitive function and alleviate anxiety, we strongly recommend considering clinical trials of WP1066 in CRPC patients presenting with depressive symptoms. As a small molecule compound, WP1066 exhibits good oral bioavailability and has undergone Phase II clinical trials in various cancers, including glioblastoma (ClinicalTrials.gov Identifier: NCT05879250), thereby establishing a foundation for its clinical translation in the context of CRPC. Future clinical trials should investigate the potential of WP1066 as a monotherapy, in combination with ADT, or as part of sequential treatment regimens, while closely monitoring tumor response, depression symptom scores, and patient quality of life. We anticipate that further in vivo and in vitro research will provide additional evidence to support the potential of WP1066 as a therapeutic strategy for the comorbidity of CRPC/MDD.

Conclusions

In summary, this study constructed a mechanistic model of the comorbidity of CRPC/MDD and identified three hub genes (AUTS2, AOC1, ANKRD37) as comorbidity biomarkers, showing promising diagnostic value in both diseases. Among them, AOC1 and ANKRD37 have prognostic significance in CRPC, while the risk association of AUTS2 with MDD was validated through two-sample MR. In addition to immune response, inflammatory response, and androgen response, this study emphasized the role of apoptosis in comorbidity, and drug prediction suggested that the JAK/STAT pathway inhibitor WP1066 may be a potential treatment for CRPC/MDD comorbidity. These findings provide valuable insights into the complex relationship between CRPC and MDD, and indicate new treatment strategies for future validation in subsequent studies.

Materials and methods

Data acquisition

Retrieve gene expression microarray datasets for CRPC and MDD from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Select the GSE70768 dataset from GEO as the CRPC dataset, which includes 13 CRPC samples from males and 73 matched benign prostate tissue samples. Choose the GSE98793 dataset as the MDD dataset, consisting of 16 male MDD samples and 16 male healthy control samples. Additionally, the GSE35988 (Platform GPL9075) dataset is used as the CRPC validation set, and the GSE19738 dataset is used as the MDD validation set.

Differential gene expression analysis and construction of co-expression gene modules

We performed differential gene analysis on the GSE70768 and GSE98793 datasets using the R software package limma. A conservative threshold (|log2FC|>0.1, p < 0.05) was applied to screen for differentially expressed genes (DEGs) in CRPC and MDD patients.

In addition to focusing on DEGs, we further applied Weighted Gene Co-expression Network Analysis (WGCNA) to identify whole-genome co-expression modules for CRPC and MDD. Using the entire gene expression profiles, we calculated the Median Absolute Deviation (MAD) for each gene, removed the bottom 50% of genes with the smallest MAD, used the R package WGCNA’s goodSamplesGenes algorithm to remove outlier genes and samples, and then constructed a scale-free co-expression network using WGCNA. Subsequently, we overlapped the DEGs with the co-expression module genes to obtain the comorbid genes for CRPC/MDD.

Construction of PPI network and GO, KEGG analysis

We built a Protein-Protein Interaction (PPI) network and visualized it using the GeneMANIA database (https://genemania.org/). GO and KEGG95,96,97analyses were performed using the R package clusterProfiler, and genes were sorted and visualized based on gene count values and p-values.

Identification and validation of hub genes

During the feature selection and model training processes for hub genes, we employed three types of machine learning methods: Random Forest (RF), LASSO regression, and Support Vector Machine (SVM). First, we utilized the randomForest package in R to construct decision trees and applied cross-validation to identify the optimal feature genes. Next, we implemented the glmnet package in R for LASSO regression to eliminate redundant factors. Subsequently, we conducted SVM analysis using the e1071 and caret packages in R to select genes that contributed most significantly to the group differences. By overlapping the hub genes identified through these three algorithms, we identified diagnostic markers for the comorbidity of CRPC and MDD. Additionally, we employed the SHapley Additive exPlanations (SHAP) method to interpret the predictions of the machine learning models, assigning importance values to the hub genes to elucidate the prediction processes98. We then plotted the receiver operating characteristic (ROC) curve to assess the diagnostic value of the hub genes and evaluated their clinical utility through decision curve analysis (DCA). Furthermore, we used the Kaplan-Meier method to construct survival curves to evaluate the prognostic value of the hub genes in CRPC.

GSEA

Gene Set Enrichment Analysis (GSEA) is commonly used to analyze and interpret pathway-level changes between normal and disease groups. We conducted single-gene GSEA analysis using GSEA 3.0 software (http://software.broadinstitute.org/gsea/index.jsp) and downloaded the h.all.v7.4.symbols.gmt subset from the Molecular Signatures Database (http://www.gsea-msigdb.org/gsea/downloads.jsp) to evaluate relevant pathways and molecular mechanisms. Based on gene expression profiles and phenotype grouping, we set the minimum gene set to 5, the maximum gene set to 5000, and performed 1000 resamplings.

Correlation analysis between hub genes and immune cells

We quantified the relative infiltration of 28 immune cell types using single-sample Gene Set Enrichment Analysis (ssGSEA) with the R software GSVA package and compared the differences in immune cell infiltration scores between the normal group and disease group of CRPC/MDD using the Wilcoxon test. To construct the immune landscape of hub genes, we determined the correlation between hub gene expression and immune cell content based on Spearman’s rank correlation coefficient.

Mendelian randomization

We employed a two-sample MR technique to explore the causal relationship between hub genes and the risk of CRPC and MDD. Using a genome-wide significance level of p < 5 × 10 − 8, minor allele frequency (MAF) > 1%, a clustering algorithm with a critical threshold of r2 = 0.001, and kb = 10000 to avoid linkage disequilibrium (LD). F-statistic was set larger than 10 to mitigate bias from weak instrumental variables. Genetic data for CRPC and MDD were obtained from publicly available genome-wide association studies (GWAS) databases (https://gwas.mrcieu.ac.uk/). The CRPC dataset was ebi-a-GCST90018685, and the MDD dataset was ebi-a-GCST003769. MR analysis was conducted using the “TwoSampleMR” package, with an inverse variance weighted (IVW) method to assess the association between hub genes and the risk of CRPC and MDD, and sensitivity analysis using the MR-Egger regression model.

Drug prediction and molecular Docking

After submitting the comorbid genes of CRPC and MDD to the Enrichr platform (https://maayanlab.cloud/Enrichr/), a drug feature database (DSigDB) in the platform was used to analyze and predict drugs for the comorbid genes. AutoDock Tools 1.5.6 was used for the hydrogenation of proteins and drug small molecules, saved in pdbqt format. Protein structures were sourced from the PDB or UniProt databases, while drug molecule structure data were obtained from the PubChem database. Molecular docking was performed using Auto Dock Vina 1.2.0 software, with a semi-flexible docking approach. The docking precision was set to 25, and the docking algorithm employed was the Lamarckian genetic algorithm.