Bioinformatics Analysis of the PSD Gene Family: Metastasis and Prognosis in HCC Authors: Hongyi Sun, Minghao Guo, Wei Liu Published: 2026 ABSTRACT Hepatocellular carcinoma (HCC) is a common malignant tumor with a poor prognosis for advanced metastasis while there is currently a lack of prognostic markers and effective therapeutic targets in this area. The PSD gene family, comprising four genes, has been implicated in tumor metastasis in breast cancer, but its role in HCC remains unclear. To elucidate the role-played by the PSD gene family in the prognosis of HCC and explore the relationship between the PSD gene family and tumor metastasis, we conducted a bioinformatics analysis using TCGA and gene expression omnibus (GEO) databases to investigate PSD gene family expression in HCC and its correlation with patient survival. Furthermore, we performed quantitative real-time polymerase chain reaction (qPCR) validation of the expression levels of the PSD gene family in clinical samples. Gene correlation and enrichment analyses were also performed to explore its relationship with tumor metastasis. In conclusion, the expression levels of the PSD gene family are significantly downregulated in HCC (P < 0.01), and the lower expressions are associated with poor survival. In addition, there is a link between the PSD gene family and RAS pathway activation, suggesting its potential role in metastasis. Introduction Hepatocellular carcinoma (HCC), a malignant epithelial tumor originating from hepatocytes, accounts for approximately 90% of the incidence of all primary liver cancers. HCC is also the fifth most prevalent cancer and the fourth leading cause of cancer-related deaths worldwide, making it a serious public health problem. 1 Existing studies have shown that liver cirrhosis, hepatitis B virus or hepatitis C virus infections, alcoholic and non-alcoholic fatty liver disease and aflatoxin B1 exposure are all major risk factors for HCC.2,3 In addition, genes such as P53, RUNX, and CTNNB1 have been shown to be closely related to HCC.4,5 Although surgery and liver transplantation have become effective tools in the treatment of early HCC, patients with HCC still face poor 5-year survival and high recurrence rates.2,6 Alpha-fetoprotein, a biomarker widely used to predict HCC prognosis, still has some limitations in clinical practice due to its low sensitivity and specificity. 7 In recent years, targeted therapy has become a new approach to cancer treatment. However, due to the significant influence of non-genetic processes on targeted therapy resistance, exploring new prognostic biomarkers and applying them to the clinical treatment of HCC is of great importance in selecting appropriate HCC treatment regimens, increasing the survival rate and improving the prognosis of HCC patients. Cancer cell invasion and metastasis are important processes in the development of HCC and one of the major causes of death in HCC patients. 8 Cancer metastasis genes are activated in the early stage of cancer or during cancer metastasis, resulting in the emergence of cancer cell metastasis phenotypes. 9 The PSD gene family is a group of genes involved in cancer cell metastasis, including PSD, PSD4, PSD2, and PSD3, respectively, encoding four isoforms of the ADP-ribosylation factor 6 (Arf6) exchange factor (EFA6) A, B, C, and D which are associated with the formation of cellular tight junctions. 10 In breast cancer, deletion of EFA6B, encoded by the PSD4 gene, has been demonstrated to complete epithelial-mesenchymal transition (EMT) and promote the occurrence of metastatic phenotypes of cancer cells through activation of CDC42 and CDC42-MRCK-pMLC and CDC42N-WASP-Arp2/3 signaling pathways. 11 In MDCK cells and the human mammary tumoral MCF7 cell line, EFA6 encoded by the PSD gene also promotes the occurrence of metastatic phenotypes through Hippo signaling inhibition via direct binding to α-actinin 1.12,13 Basement membrane crossing is an important step in the invasion and metastasis of cancer cells. Since tight junction is one of the ways of basolateral membrane junction of adjacent hepatocytes, 14 aberrant changes in the PSD gene family in HCC may induce invasion and metastasis of cancer cells through basement membrane crossing, leading to a poor prognosis of HCC. High-throughput sequencing has enabled comprehensive profiling of mRNA in HCC, revealing differentially expressed genes (DEGs) that play crucial roles in disease progression and prognosis. Transcriptional analysis studies on HCC mRNA have identified numerous DEGs, such as AXIN1, TP53, and CTNNB1,4,5 which have been proved to be associated with poor prognosis and tumor progression, revealing the effectiveness of differential mRNA analysis in HCC. Although previous bioinformatics studies on HCC have revealed the impact of genes on HCC metastasis like Sortilin1 15 through DEGs analysis, current research has not fully elucidated the genetic factors driving this process. In our study, we used data from The Cancer Genome Atlas (TCGA) and other clinical and molecular databases like GSE64041 to explore the potential role of the PSD gene family in HCC. Through differential analysis of HCC mRNA, we found that compared with the normal population, the transcription level of PSD gene family decreased in HCC patients and decreased with the increase of the malignant degree of HCC cells. This result was also confirmed in clinical samples by quantitative real-time polymerase chain reaction (qPCR) analysis. Analysis of prognostic data suggested that the low expression of PSD gene family was associated with low survival and poor prognosis in HCC patients. Since changes in gene transcriptional levels generally affect patient prognosis by altering protein synthesis and cellular physiological functions, we concluded Protein-Protein Interaction (PPI) and KEGG analyses on PSD gene family to explore the potential causes of HCC due to PSD genes expression changes and our results confirmed the association between the PSD gene family and the metastasis of HCC cells. In conclusion, members of the PSD gene family can serve as molecular biomarkers for assessing HCC prognosis and potential targeted interventions. Methods Clinical data Comprehensive clinicopathological annotations for TCGA-LIHC (n = 374) were downloaded from the GDC Data Portal (https://portal.gdc.cancer.gov).16,17 Tumor extent was defined by the AJCC 7th edition Tumor-Node-Metastasis staging system (TNM) system: T1–T4 for primary size/invasion, N0/1 for nodal status, and M0/1 for distant metastasis. Histological grade (G1–G4) and overall stage (stage I–IV) were extracted from accompanying pathology reports; substages (e.g. T1a–T1c) were collapsed into the parent category (T1). Categorical variables were compared with Pearson’s χ2 test, and continuous variables with two-tailed Student’s t-test. 95% Confidence intervals were computed in R v4.3.0 with the DESeq2 wrapper functions. 18 Pan-cancer analysis An in-depth analysis of the expression patterns of the PSD gene family across 33 diverse tumor cohorts was conducted using data hosted on the TIMER platform. 19 This analysis encompassed both tumor and normal samples, with log2(TPM) serving as the normalization method to ensure data accuracy and comparability. Differential analysis Transcriptome profiles of 374 HCC and 50 matched non-tumor tissues (TCGA-LIHC) were retrieved from the cBioportal database (http://www.cbioportal.org/) and variance-stabilized with DESeq2 package. 20 Differential expression of the PSD gene family was computed against the normal controls; genes with|log2 fold change| > 1 and FDR < 0.05 (Benjamini–Hochberg) were considered significant. 21 Dataset GSE64041 from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) including RNA-seq data from both HCC tumor samples and normal liver tissue samples was utilized to validate the robustness of our findings and the same analytical pipeline was replicated. 17 All the data was processed using R software. Survival analysis PSD gene-family expression was treated as a categorical predictor. After excluding the top and bottom 5% of values to dampen outlier influence, samples were split at the cohort-specific median (median-cutoff) and again at the lower quartile (Q1-cutoff). Survival information from TCGA clinical data was obtained to generate the analysis. Overall survival curves were generated with the Kaplan–Meier estimator and compared by two-sided log-rank tests with R software. 22 Univariable and multivariable Cox proportional-hazards models were then fitted to estimate the association between survival and PSD expression, TNM categories, or overall stage. The proportional-hazards assumption was checked using Schoenfeld residuals; no significant time-dependent effects were detected. Hazard ratios (HR) with 95% confidence intervals were derived from Wald statistics, and the contribution of each covariate was evaluated with likelihood-ratio tests. All analyses were performed in R (survival package v3.5). 23 Functional enrichment analysis PSD-interacting proteins were retrieved with STRING (https://string-db.org/) v11 (confidence ⩾ 0.4, maximum 300 interactors) and imported into R. 24 Ontologic annotation was performed with clusterProfiler 4.8. Cellular Component (CC), Biological Process (BP), and Molecular Function (MF) terms were screened with the “elim” algorithm to remove parent–child redundancy, and P-values were Bonferroni-corrected.20,25,26 Enrichment maps were drawn when fold-enrichment ⩾ 1.5 and adjusted P ⩽ .05. KEGG pathway enrichment analysis was performed to further explore the roles of these genes in biological pathways.20,27 KEGG pathway analysis followed the same package (gene-count ⩾ 5, q ⩽ 0.05). The Gene Set Enrichment Analysis (GSEA) was employed to extend the investigation to the whole transcriptome. 20 In this step, we ranked genes based on their differential expression levels between the two types of samples and tested whether preset gene sets were significantly enriched at the top or bottom of this ranked list. 28 The enrichment score (ES) was a key indicator reflecting the degree of overrepresentation of the gene set at the extreme positions of the ranked list. The nominal P-value (NOM p-val) was determined based on permutation tests and was used to assess the statistical significance of the enrichment score. The False Discovery Rate (FDR q-val) evaluated the reliability of the enrichment results. Gene co-expression analysis Correlation analyses of the PSD gene family were performed based on transcriptomic expression data to characterize internal co-expression patterns among PSD family members. To evaluate whether metastatic status or lymph node involvement modulates PSD co-expression in HCC, samples were stratified according to clinical annotations, and correlation analyses were repeated within each subgroup. Correlation matrix plots and scatter plots were generated using an online platform for data analysis and visualization (https://www.bioinformatics.com.cn). Correlation strength was quantified by the Pearson correlation coefficient (R), with R > 0.3 considered indicative of a correlation and R > 0.5 defined as a strong correlation. Prediction of miRNAs TargetScan tool (http://www.targetscan.org) was used to predict MicroRNAs (miRNAs) targeting the PSD gene family.17,29,30 The results were filtered based on the scores of seed match, context++ score, and predicted relative knockdown efficiency (KD), with context++ score as the first priority, predicted relative KD as the secondary criterion, and seed match types prioritized in the order of 8mer>7mer-m8>7mer-1A. 31 The likelihood and strength of interactions between miRNAs and target mRNAs were further evaluated using the core prediction parameter (Target Score) of the miRDB tool (http://mirdb.org). The Binding site context++ score (negative values indicate inhibitory effects) is categorized as: −0.3 to −0.2 for strong binding and less than −0.3 for extremely strong binding. The predicted relative KD (negative values indicate inhibitory effects) is classified as: −3 to −2 for strong regulation (potential therapeutic targets) and less than −3 for extremely strong regulation. In addition, the miRDB Target Score defines targets as: ⩾80 for high-confidence, and 60-79 for medium-confidence. All data used in this study are derived from publicly available databases, facilitating access and verification by readers. Clinical samples and quantitative real-time polymerase chain reaction (qPCR) Four clinical samples were divided into two groups, with each group containing one HCC tumor tissue sample and one adjacent normal liver tissue sample. The primers designed for the PSD gene family in qPCR were as follows: PSD1F (“GACCAAGCTCTCAGGGTGTT”), PSD1-R (“GTGGAGATCCGTGTTGAGCA”), PSD2F (“TGCCAAAGACCTGCTGAAGACC”), PSD2-R (“CCGAACTTGTCATCCACCAGCT”), PSD3F (“TGTCTGTATGGTTGCTGCAATG”), PSD3-R (“TTTGGCCAACCTTTTGGCTG”), PSD4F (“TCTTGGAGACGGATGGGGAA”), PSD4-R (“TGTCCTCAGATTCTCATGTTCTGC”). Pre-denaturation was carried out at a set temperature of 95°C for 30 seconds. In the cycling stage, the denaturation condition was 95°C for 10 seconds, and the annealing and extension steps were at 60°C for 30 seconds, repeated for a total of 40 cycles. Fluorescence signals were collected during the annealing and extension. The melting curve was generated using the instrument’s default melting curve acquisition program. Results The PSD gene family is under-expressed in HCC The TIMER data base (http://timer.cistrome.org/) is an analysis network that can be used for differential gene expression analysis between tumor and normal samples. 32 Thus, we compared the expression levels of four PSD gene family members in normal samples and tumor samples in pan-cancer to explore the correlation between PSD gene family expression and the occurrence of various cancers through the TIMER data base. The results showed that in more than 30 cancers, only the PSD4 gene had an overall lower expression in tumor samples than normal samples, while the other three members of PSD gene family did not show particular differences (Figure 1A). Next, we explored the RNA transcriptional levels of the PSD gene family in HCC. Using the RNA-seq data of HCC from TCGA database, we performed a difference analysis on the data of 50 normal samples and 374 samples of HCC patients. Transcriptional analysis revealed downregulation of PSD, PSD3, and PSD4 in tumor samples, with PSD3 and PSD4 showing significantly lower expression levels in HCC tissues compared to normal controls (P < 0.01) (Figure 1B). In order to further corroborate these results, we collected the TPM-normalized HCC RNA-seq microarray dataset GSE64041 from the GEO database. These data indicated that the RNA expression levels of all members in PSD gene family except PSD3 were significantly lower in HCC samples than in normal samples (Figure 1C). All these results confirm that the PSD gene family is overall under-expressed in HCC samples compared to normal samples. (A) The RNA-seq data for PSD4 mRNA levels in tumors and adjacent normal tissues across 33 cancers on the TIMER platform. Blue for normal, red for tumor, and purple for metastatic samples. Statistical significance: “*” indicates P-value < 0.05, “**” indicates P-value < 0.01, and “***” indicates P-value < 0.001. (B) Boxplots present the normalized RNA-seq data for PSD, PSD2, PSD3, and PSD4 transcriptional levels in HCC tumors sourced from the TCGA database. (C) The normalized RNA-seq data for PSD, PSD2, PSD3, and PSD4 transcriptional levels, obtained from the GSE64041 database. (D) The normalized RNA-seq data for PSD, PSD2, PSD3, and PSD4 transcriptional levels across histological grades (ranging from G1 to G4) in HCC tumors. To explore the association between RNA expression of the PSD gene family and clinicopathologic characteristics, we used clinical data from HCC samples in TCGA to analyze the trend of PSD gene family expression with clinical staging. Tumor stage (T), tumor stage (S), and histological stage are commonly used clinical criteria for tumor staging, and according to these criteria, we analyzed the expression of four members of the PSD gene family in different stages (Supplemental Table S1). Figure 1D showed that in histological stage, the expression of four PSD gene family members decreased with the increasing staging, especially in the staging from G3 to G4, indicating that the PSD gene family plays a protective role in HCC progression. In addition, given the putative contribution of PSD-encoded EFA6 proteins to tumor metastasis, we calculated the average gene expression in patients with N1 and M1 in TNM staging and compared it to patients with N0 and M0. Among them, the expression of four PSD genes was basically lower in patients with stage M1 than in patients with stage M0, while only the PSD3 gene was lower in patients with stage N1 (Supplemental Table S1). Downregulated expression of PSD gene family negatively impacts survival in HCC patients To investigate whether PSD gene family expression status is associated with patient prognosis, we grouped the clinical data obtained from TCGA. Among the 424 samples obtained, we defined the samples that were above the median expression as the “high expression group” and vice versa as the “low expression group.” In order to avoid the influence of extreme values on data analysis, we respectively deleted the samples with expression in the upper and lower 5% of the two groups and performed Kaplan-Meier survival analysis on the remaining 382 samples. The analysis showed that lower expressions of PSD2 and PSD4 genes were associated with worse prognosis in HCC (P < 0.05) (Figure 2A). In addition, K-M plots showed that the lower quartile expression group of the PSD4 gene significantly reduced the survival of HCC patients (P < 0.05) (Figure 2B). These outcomes suggest that the reduced expression of PSD gene family is associated with worse survival in HCC patients. (A) The Kaplan-Meier plots on the influence of the expression levels of the four PSD family genes on the survival of cancer patients. P-value < 0.05 is deemed statistically significant. (B) Kaplan-Meier survival plots for HCC patients, categorized based on quartiles of PSD gene family expression levels. P-value < 0.05 is deemed statistically significant. We next established a unifactorial Cox risk ratio model by making different tumor staging criteria as single factors, with the aim of evaluating the correlation between each factor and patient prognosis. Tumor S stage showed significantly high risk in this model (P < 0.001) (Supplemental Table S2), so we built a multifactorial Cox risk proportionality model based on this result. The analysis showed that relative to the control group (low PSD gene expression group and S4 stage), high PSD gene expression and lower S stage impeded the occurrence of patient death (Supplemental Table S3), and this result was applicable in all four PSD gene families. In addition, the results of the univariate Cox regression analysis of HCC patient survival with different PSD gene family expression levels as a single factor showed that the low expression groups of the PSD2, PSD3, and PSD4 genes all have a higher risk associated with the death of HCC patients compared to the high expression groups (Supplemental Table S4). All these results confirm that low expression of the PSD gene family has an adverse effect on patient survival. PSD3 and PSD4 are correlated Gene expression correlation analysis is a routine genomics tool for clarifying gene function and disease mechanisms. To clarify the correlation between the four members of the PSD gene family, we first performed a correlation analysis of the PSD gene family among the data of 33 cancers provided in the TCGA database. In breast invasive carcinoma, rectum adenocarcinoma and HCC, PSD3 and PSD4 genes showed a high positive correlation (Table 1 and Figure 3A). Table 1.Correlation coefficients for the PSD gene family across multiple cancers.RPSD-PSD2PSD-PSD3PSD-PSD4PSD2-PSD3PSD2-PSD3PSD3-PSD4ACC 0.35 0.260.100.21−0.060.13BRCA−0.050.010.20−0.040.15 0.49 COAD0.010.070.020.050.26 0.41 DLBC−0.020.130.210.17 0.31 0.28GBM0.21 0.90 0.090.240.260.19KIRC−0.02 0.53 −0.17−0.040.04−0.12LGG0.12 0.75 −0.110.160.190.06HCC0.11−0.030.130.050.08 0.30 LUAD0.100.110.240.010.00 0.40 MESO0.010.070.06 0.41 −0.03−0.12OV0.160.170.200.200.18 0.33 PCPG0.18 −0.31 −0.09−0.06−0.06 0.43 READ0.090.270.060.27 0.51 0.60 TGCT0.23 0.40 0.06 0.55 0.160.20UCS 0.69 −0.22−0.06−0.15−0.020.04Correlation status among the PSD gene family across 14 different types of tumor cohorts in TCGA database. Correlation coefficients (R) ⩾ 0.30 are considered statistically significant. Correlation coefficients for the PSD gene family across multiple cancers. Correlation status among the PSD gene family across 14 different types of tumor cohorts in TCGA database. Correlation coefficients (R) ⩾ 0.30 are considered statistically significant. (A) Correlation coefficient plot on the relationships among the PSD gene family in HCC and a scatter plot on the internal correlation coefficients within the PSD gene family across HCC, rectal adenocarcinoma, and breast cancer. The circle size reflects the relative strength of the correlation coefficients between genes, while the color intensity indicates the absolute strength of these correlations. R > 0.3 is considered indicative of a correlation. (B) Analysis of the correlation within the PSD gene family in M1 and N1 patient samples within HCC. R > 0.5 is considered indicative of a strong correlation. The expression patterns of PSD3 and PSD4 genes in breast cancer align with those observed in HCC. Given the established role of PSD4 gene in encoding EFA6 in breast cancer metastasis, 11 we hypothesize that the PSD3 gene may be also involved in this process in breast cancer and may promote HCC metastasis by the same mechanism. To further explore this speculation, we performed a correlation analysis within the PSD gene family in HCC patients with TNM staging of N1 and M1. As expected, the PSD gene family exhibited significantly higher correlation coefficients in HCC patients with N1 and M1. This trend was most pronounced in the correlation between PSD and PSD2, as well as between PSD3 and PSD4 (Figure 3B). Given the downregulation of PSD3 and PSD4 in N1 and M1 patients observed in the differential analysis, we propose that these two genes mediate tumor metastasis in HCC through a mechanism analogous to their role in breast cancer. Enrichment analysis of PSD gene family functional networks To further investigate the underlying mechanisms of PSD gene family in HCC, we applied the 300 genes that interact most strongly with the PSD gene family obtained from STRING database (https://string-db.org/) 33 and performed the gene ontology (GO) analysis. In the cellular component, the formation of the cell’s front or leading edge is significantly enriched (Figure 4A). The leading edge is the front area formed during cell migration and is the main power area that pulls the cell forward. The formation of lamellipodia and filopodia is one of the ways in which the leading edge of cells is formed. 34 Existing studies have shown that dysregulation of mitochondrial dynamics in HCC cells both in vitro and vivo promotes the lamellipodia and filopodia formation, and in turn, cellular metastasis. 35 In addition, activation of small GTPases and RAS protein signaling pathways showed significant enrichment (Figure 4A). RAS and ARF6 encoded by the PSD gene are both members of the small GTPase family. Among them, RAS has an important role in mediating inflammation, immunity, and cell proliferation. The RAS protein encoded by the K-RAS gene has also been shown to have an important impact on human cancer. 36 To further investigate the underlying biological functions of these genes, we subsequently conducted KEGG and GSEA analyses. Strong enrichment was also exhibited by pathways including chemokine signaling, estrogen signaling, and circadian entrainment (Figure 4B and C). In sum, all the results consistently indicate that the PSD gene family plays an important role in both the initial stage of HCC genesis and the stage of migration of progressive cancer cells. (A) A Bubble plot on BP-GO (Biological Process—Gene Ontology), CC-GO (Cellular Component—Gene Ontology), and MF-GO (Molecular Function—Gene Ontology) analyses. A Lollipop plot provides the GO analysis results for the PSD gene family. (B) A bubble plot shows the KEGG pathways with high enrichment ratios. A bar plot illustrates the enrichment significance of the top pathways. A lollipop plot provides an alternative view of the KEGG results. (C) A Bubble plot showcases the GSEA results, focusing on enrichment ratios within the PSD gene family. A Ridge plot and Line plot complement the bubble plot by providing additional perspectives on the GSEA analysis. (D) A PPI network diagram illustrates the interaction structure related to the PSD gene family. A Bar plot summarizes the BP-GO, CC-GO, and MF-GO categories that exhibit high enrichment ratios among the genes associated with the PSD gene family. To understand the relationship between the PSD gene family and cancer-related genes, we used the online PPI networks analysis on the STRING website to look for genes with strong associations with them. Apparently, in agreement with the results of the functional analysis of the PSD gene family, the KRAS and HRAS genes showed a strong correlation with the PSD gene family (Figure 4D). Consistently, GO analyses identified the small GTPase and RAS protein signaling pathways as substantially enriched (Figure 4D). This further presents the strong correlation between RAS and PSD gene family and their significant role in HCC development. In addition, based on our gene co-expression analysis results, we also found that the PSD gene family was closely related to the PIK3CA and AKT1 genes, especially the PSD3 and the PSD4 genes (Supplemental Table S5). The proteins PI3K and AKT, encoded by PIK3CA and AKT1 genes, respectively, function downstream of the RAS signaling pathway. Activated RAS protein can regulate cancer cell migration and invasion in various cancers through the downstream PI3K-AKT signaling way.37,38 Predictive analysis of potential miRNAs regulating the downregulation of the PSD gene family MicroRNAs (miRNAs) are a class of highly conserved non-protein-coding small RNAs that negatively regulate gene expression to maintain cellular homeostasis by destabilizing target mRNAs and inhibiting the translation of target mRNAs. Due to the pivotal role of miRNAs in modulating gene expression, they have increasingly emerged as a crucial target for the development and mechanism of action of anticancer therapies. 39 To further explore the significance of the PSD gene family in clinical treatment based on the bioinformatics analysis, we used the Targetscan and miRBD platforms to screen for miRNAs that regulate the PSD gene family. Based on the evaluation of the biding of miRNAs with the PSD gene family, the inhibitory effect on the PSD gene family and the accuracy of miRNAs, we have identified five miRNAs with the most promising profiles: hsa-miR-6763-5p, has-miR-3150a-3p, has-miR-199b-3p, has-miR-4735-3p, and has-miR-1321 (Table 2). Targeting and inhibiting these miRNAs may be a potential method to improve the prognosis of HCC patients by upregulating the expression of the PSD gene family. Table 2.Potential miRNA targets of the PSD gene family.miRNAGeneSeed matchContext++ scorePredicted relative KDmiRDB Target Scorehsa-miR-6763-5p PSD 8mer−0.5−2.75474hsa-miR-3150a-3p PSD 8mer−0.49−2.69274hsa-miR-199b-3p PSD2 7mer-m8−0.29−4.49282hsa-miR-4735-3p PSD3 7mer-m8−0.29−4.8993hsa-miR-1321 PSD4 7mer-m8−0.19−3.30362The screening results of potential miRNA regulatory targets for the PSD gene family. For the context++ score of binding sites, more negative values indicate stronger inhibitory effects. The Binding site context ++ score (negative values indicate inhibitory effects) is categorized as: −0.3 to −0.2 for strong binding and <−0.3 for extremely strong binding. The predicted relative KD (negative values indicate inhibitory effects) is classified as: −3 to −2 for strong regulation (potential therapeutic targets) and <−3 for extremely strong regulation. In addition, the miRDB Target Score defines targets as: ⩾80 for high-confidence, and 60-79 for medium-confidence. Potential miRNA targets of the PSD gene family. The screening results of potential miRNA regulatory targets for the PSD gene family. For the context++ score of binding sites, more negative values indicate stronger inhibitory effects. The Binding site context ++ score (negative values indicate inhibitory effects) is categorized as: −0.3 to −0.2 for strong binding and <−0.3 for extremely strong binding. The predicted relative KD (negative values indicate inhibitory effects) is classified as: −3 to −2 for strong regulation (potential therapeutic targets) and <−3 for extremely strong regulation. In addition, the miRDB Target Score defines targets as: ⩾80 for high-confidence, and 60-79 for medium-confidence. The PSD gene family is downregulated in HCC clinical tissue samples To further verify the accuracy of the bioinformatics analysis results, we selected two pairs of clinical samples derived from HCC tissue and adjacent normal liver tissue (each pair consisting of one liver cancer tissue and one adjacent normal tissue) for qPCR analysis. The findings from the independent samples t-test analysis of the two pairs of clinical specimens revealed that, with the exception of the PSD2 gene, the expression levels of the PSD3 and PSD4 genes were all reduced in HCC tissue compared to adjacent normal tissue (Figure 5). Specifically, the expression levels of the PSD3 and PSD4 genes were significantly lower in HCC tissue (P < 0.05). The differential expression of PSD gene family in clinical samples by qPCR analysis. “NC” represents adjacent normal liver tissue samples and “HCC” represents tumor samples. “*” indicates P-value < 0.05. “**” indicates P-value < 0.01. Overall, the results validated in clinical samples are consistent with our bioinformatics analysis, indicating that the PSD gene family is downregulated in HCC patients. Discussion HCC cell metastasis is one of the major causes of poor prognosis in patients with HCC. Most of the current research and drug development on HCC metastasis have focused on inhibiting tumor angiogenesis, including the classic anticancer drugs sorafenib and lenvatinib.40,41 However, basement membrane breakthrough of tumor cells often occurs at the earliest stage of their metastasis and has an indispensable impact on their subsequent distant migration and other activities. The PSD gene family found in breast cancer that is associated with tumor cell migration is not explored deeply in HCC. We explored public datasets through bioinformatics analysis in order to provide a direction for the research of the possible biological mechanisms of the PSD gene family in HCC. Our analysis of the data confirmed that the mRNA expression of the PSD gene family was generally lower in HCC tissue than in normal liver tissue and it was further validated by qPCR results from clinical samples. Furthermore, low expression of PSD genes was associated with poor patient prognosis and histological grade, indicating that PSD genes may play a positive role in HCC. In pan-cancer analyses, although the upregulation of PSD3 gene is associated with tumor proliferation, migration and invasion in thyroid cancer, 42 there is no definitive evidence linking this phenomenon to poor prognosis. Therefore, the role of the PSD gene family in the prognosis of other cancer requires further investigation. Through correlation analysis of four members of the PSD gene family and GO enrichment analysis of the network of interacting genes, several results point to a potential mechanism of HCC that causes tumor cell metastasis similar to breast cancer, including the strong correlation of PSD3 and PSD4 genes together in breast cancer and HCC in correlation analysis, the significantly higher correlation of PSD3 and PSD4 genes in N1 and M1 samples and functional analysis with multiple functional enrichments for cell membrane changes including the formation of the leading edge. It is noteworthy that the PSD4 gene-induced EMT process observed in breast cancer is closely related to the plasticity, migration, and invasiveness of cancer cells. EMT is regulated by transcription factors, and Snail1, as one of the transcriptional regulators of EMT, is significantly more abundant in tumor cells exhibiting metastatic HCC than in those without. 43 Thus, we hypothesize that the PSD4 gene may also be involved in the regulation of the EMT process during liver cancer metastasis, thereby promoting basement membrane breakthrough and distant metastasis. However, whether the PSD gene family is involved in EMT and its correlation with EMT transcription factors remains unclear. Further research on EMT regulation is essential for future investigations. Furthermore, functional analysis identified a RAS protein signaling pathway related closely to the physiological function of the PSD gene family. This is consistent with the existing observations of the Ras/Raf/MEK/ERK signaling pathway that plays an important role in the occurrence and development of HCC. 44 The PSMD2 gene, a potential immune checkpoint in HCC reported by Gu et al 45 recently is closely related to the MAPK signaling pathway and the ubiquitin-mediated proteolysis pathway. As a classic MAPK signaling pathway, Ras and Raf proteins in the Ras/Raf/MEK/ERK signaling pathway ae also regulated by ubiquitination. Therefore, the PSD gene family and the PSMD2 gene may jointly participate in the regulation of this signaling pathway. However, contrary to the expression pattern of the PSD gene family, the PSMD2 gene is significantly overexpressed in HCC. Whether there is crosstalk between these two genes in regulating cell proliferation pathways in HCC remains to be determined in further research. In addition, our KEGG analysis showed that circadian rhythms may also be correlated with the occurrence of HCC. However, only Padilla et al 46 have reported how circadian dysfunction drives human-hepatocyte carcinogenesis via non-alcoholic fatty liver disease and how reprogramming within HCC subsequently promotes tumor metastasis. The correlation between circadian rhythms and the occurrence of liver cancer still needs further confirmation in the future. Targeting cancer-related genes with miRNAs has shown promising results in cancers such as ovarian cancer. To better integrate the PSD gene family with clinical targeted cancer therapy, we used Targetscan and miRBD platforms to identify the miRNAs most associated with the PSD gene family. We hope that by specifically targeting the PSD gene family with miRNAs, we can provide insights for achieving precise and low-side-effect targeted cancer therapy. However, the miRNAs shown in Table 2 have not yet been studied in vivo or in vitro in HCC, and their specific functions remain to be investigated. Inspired by the development of hepatitis B and C virus mRNA vaccines, we hypothesize the possibility of packaging specific miRNAs into nanoparticles for targeted cancer therapy in the form of vaccines in the future. 47 In addition to miRNAs, another class of RNA molecules, long non-coding RNAs (lncRNAs), play an important role in HCC by regulating key processes such as cell growth, apoptosis, invasion, and metastasis through mechanisms such as epigenetic regulation, interaction with miRNAs, and regulation of protein activity. Maternally Expressed Gene 3 (MEG3), a highly expressed lncRNA that inhibits cell growth and induces programmed cell death, is upregulated in most cancers. However, its expression is reduced in HCC. Upregulation of MEG3 can inhibit HCC cell proliferation by negatively regulating miRNA-664, exhibiting a certain anti-tumor effect. Furthermore, several other lncRNAs, including GAS5, MVIH, and UC.134, have shown close correlation with HCC. As a type of RNA other than miRNAs, lncRNAs participate in the regulation of HCC through multiple pathways, including synergistic interactions with miRNAs, demonstrating their significant potential as future therapeutic targets for HCC. 48 Based on previous studies and all of our analyses, we believe that the PSD gene family plays an important role in inhibiting HCC basement membrane breaching and thus metastasis. However, its biological mechanism in RAS protein signaling pathway and other aspects still needs further experimental studies. Conclusion Our study has successfully identified the protective role of the PSD gene family in HCC. This protective role includes improving survival in HCC patients and reducing basement membrane invasion and distant metastasis of liver cancer cells. Furthermore, we have identified five miRNA molecules closely related to the PSD gene family, aiming to provide options for developing targeted therapies for HCC. Limitations of the study Our bioinformatics analysis indicated that the PSD gene family is associated with HCC prognosis and metastasis; however, this association remains to be established experimentally through approaches such as gene knockdown or overexpression in relevant HCC cell lines (e.g. MHCC97) or in vivo models. Moreover, we only identified five potential miRNA targets regulating the PSD gene family from the Targetscan and miRBD platforms without testing or evaluating the actual targeting effects of these miRNAs. Furthermore, although we analyzed the relationship between the RAS signaling pathway and the PSD gene family, we did not further explore the regulatory mechanisms of the PSD gene family on this signaling pathway. Given that the PSD gene family was associated with the metastasis process, EMT, as a pathway closely related to tumor metastasis, its relationship with the PSD gene family still needs further investigation. Supplemental Material sj-docx-1-bbi-10.1177_11779322261425454 – Supplemental material for Bioinformatics Analysis of the PSD Gene Family: Metastasis and Prognosis in HCC Supplemental material, sj-docx-1-bbi-10.1177_11779322261425454 for Bioinformatics Analysis of the PSD Gene Family: Metastasis and Prognosis in HCC by Hongyi Sun, Minghao Guo and Wei Liu in Bioinformatics and Biology Insights