PDZD11 gene expression data and clinical information for HCC patients were retrieved from TCGA through the GDC data portal (https://portal.gdc.cancer.gov/) (29), including 425 samples (50 paracancerous and 375 tumor tissues). PDZD11 expression and clinical data from the ICGC dataset (https://dcc.icgc.org/) were also used to verify survival analysis. Perl was used to sort and merge gene expression data, while R’s “limma” package (version 4.2.3) extracted PDZD11 expression data. The “limma” and “beeswarm” packages visualized this data with scatter plots. Perl also extracted survival data, removing incomplete entries, and matched complete survival information with PDZD11 data, resulting in data for 370 eligible patients.

HCC patients were divided into high or low PDZD11 expression groups based on the median expression level. The Kaplan-Meier method evaluated variations in OS within the TCGA HCC group. The ‘survival’ package handled statistical analysis, while the survminer mapped survival curves. The receiver operating characteristic (ROC) analysis assessed the prediction accuracy and specificity of PDZD11. Additionally, key prognostic indicators were identified using univariate and multivariate Cox regression analysis with the ‘survival’ package in R (version 4.2.3).

We developed a nomogram for forecasting patient outcomes based on a multivariate Cox regression analysis. To evaluate the nomogram’s predictive accuracy, we also constructed a calibration plot. Both the nomogram and the calibration plot were generated using the ‘survival’ and ‘rms’ packages in R (version 4.2.3) (30).

The study enrolled 78 patients newly diagnosed with HCC and 62 HC from the Central Hospital of Yongzhou. There were two categories of participants: HCC and HC. Diagnosis of HCC was confirmed based on China’s 2022 Guidelines for Primary Liver Cancer (31). Exclusion criteria included the presence of benign or malignant tumors in other locations, as well as other uncontrollable conditions such as severe infections, renal failure, and heart failure. The HC group underwent standard biochemical and immunological assessments to exclude the presence of hepatitis, cirrhosis, liver, gallbladder, and biliary tract tumors, as well as benign and malignant tumors in other regions. The research protocol received approval from the Ethics Committee of the Central Hospital of Yongzhou. This research followed the guidelines set forth in the Declaration of Helsinki and its amendments.

Blood samples were collected from each subject before surgery or treatment, subsequently centrifuged at 3,500 g for 10 min, and the supernatant was immediately frozen at -80°C, and the storage duration did not exceed six months.

Information on clinicopathological characteristics, including factors like gender, age, tumor dimensions and quantity, cancer stage, lymph node status, distant spread, TBA, TBIL, DBIL, ALT, AST, WBC, and serum AFP levels, were meticulously gathered.

PDZD11 protein concentrations were quantified utilizing a commercially available sandwich ELISA kit (EIAab, Wuhan, China), following the manufacturer’s protocol (https://www.eiaab.com.cn/). Each sample was assayed in duplicate. The intra- and inter-assay coefficients of variation were <5.2% and <9.6%, respectively.

Data analysis utilized R (Version 4.2.3), SPSS (Version 20.0), GraphPad Prism (Version 9.5), and CurveExpert (Version 1.4). Normally distributed data were presented as means ± SD, and correlations between HCC and age were analyzed by the independent samples t-test. Non-normally distributed data were shown as medians and interquartile ranges (M [P25, P75]), and correlations between HCC and PDZD11, AFP, TBA, TBIL, DBIL, ALT, AST, and WBC were evaluated with the Mann–Whitney U-test. Categorical variables were presented using frequencies and percentages, and correlations between HCC and gender, Child-Pugh class, tumor size, tumor number, and tumor stage HBV or HCV were calculated by the Chi-square test. Diagnostic accuracy was assessed via ROC curve analysis, yielding AUC, threshold values, sensitivity, and specificity. Principal component analysis was conducted using the PCA expression plot feature in GraphPad Prism. Univariate logistic regression models were performed to test the association between PDZD11 and several clinical metrics, as well as the presence of HCC. Statistical significance was set at a p-value below 0.05.

A Wilcoxon rank-sum test showed significantly higher PDZD11 levels in HCC samples compared to adjacent ones ( Figure 1A ). By applying the Wilcoxon signed-rank test to 50 paired samples of liver cancer and adjacent tissues, we verified that PDZD11 expression was markedly reduced in the adjacent tissues ( Figure 1B ). Kaplan-Meier analysis revealed that HCC patients with high PDZD11 levels had shorter OS compared to those with low levels (p < 0.01) ( Figure 1C ). In the same way, analysis of the ICGC dataset verified that elevated PDZD11 levels in HCC correlate with reduced OS (p < 0.01) ( Figure 1D ). The ROC curve analysis showed that PDZD11 had an AUC of 0.981, indicating high accuracy ( Figure 1E ). This suggests PDZD11 could be a valuable biomarker for diagnosing and predicting HCC prognosis.

We assessed predictors of OS through both univariate and multivariate Cox regression models ( Figure 2 ). According to the univariate analysis, both cancer stage and PDZD11 levels were significantly associated with OS in HCC patients (p < 0.001) ( Figure 2A ). According to multivariate analysis, both stage and PDZD11 expression (all p < 0.01) were identified as independent prognostic indicators ( Figure 2B ).

A clinical tool was developed using a nomogram based on stage and PDZD11 expression to predict 1-, 3-, and 5-year OS in HCC patients. Figure 3A illustrates that higher total points correlate with poorer survival outcomes. While the calibration plots for 1-year and 3-year predictions ( Figures 3B, C ) suggested potential underestimation or overestimation of mortality by the nomogram, the calibration plot for the 5-year predictions ( Figure 3D ) demonstrated satisfactory predictive accuracy, with the bias-corrected line closely matching the ideal curve, indicating a strong correlation between observed and predicted values. These results demonstrated a satisfactory performance of the constructed nomogram.

This study examined a cohort consisting of 78 patients diagnosed with HCC and 62 HC, as detailed in Table 1 . Among the 140 participants enrolled, 128 (91.4%) were male and 12 (8.6%) were female. The mean ages for the HCC patients and HC were 58.8 ± 1.0 years and 59.3 ± 0.7 years, respectively. The two groups showed no notable variations in age or gender distribution. Table 1 displays the initial traits of individuals with HCC and HC.

The ELISA analysis for PDZD11 was conducted on a cohort comprising 78 HCC patients and 62 HC. Serum PDZD11 levels of HCC patients were markedly higher than those in HC (p < 0.0001, Figure 4A ). In particular, the median serum levels of PDZD11 were 131.72 ng/ml (range: 77.42–264.86 ng/ml) in HCC patients and 85.45 ng/ml (range: 64.59–100.9 ng/ml) in HC. The results indicate that PDZD11 could act as a possible blood-based marker for HCC. Nonetheless, there were no notable associations between serum PDZD11 concentrations and factors such as age, sex, tumor dimensions, tumor count, or Child-Pugh classification (p > 0.05). Additional examination showed that patients with stage III-IV tumors had increased serum PDZD11 protein levels compared to those with stage I-II tumors (p = 0.043), and these levels were notably higher than in the HC group (p < 0.0001, Figure 4B ). On the other hand, there was no notable disparity between the HC group and patients with stage I-II tumors (p = 0.961, Figure 4B ). Notably, there was a higher correlation between serum PDZD11 levels and poor performance status (PS) in HCC patients (ECOG PS 3-4) compared to those with good PS (ECOG PS 0-2) (p = 0.0171, Figure 4C ). Furthermore, both the poor and good PS groups had markedly higher serum levels of PDZD11 compared to the HC group.

To assess PDZD11 and AFP’s diagnostic performance as individual biomarkers, a ROC curve was constructed. As illustrated in Table 2 and Figure 5A , both markers demonstrated significant diagnostic efficacy in differentiating HCC from HC, with PDZD11 achieving an AUC of 0.713 (95% CI: 0.626 - 0.799) and AFP achieving an AUC of 0.932 (95% CI: 0.884 - 0.979). Youden’s index was employed to ascertain the optimal threshold values for PDZD11 and AFP. For PDZD11, the optimal threshold for distinguishing HCC from HC was 105.50 ng/mL, corresponding to a sensitivity of 0.615 and a specificity of 0.807. For AFP, the optimal threshold for AFP was determined to be 6.085 ng/ml, yielding a sensitivity of 0.831 and a specificity of 0.984. Importantly, the capability to differentiate HCC from HC improved markedly, as the AUC for the PDZD11 and AFP combination hit 0.958, whereas AFP alone had an AUC of 0.932 (p < 0.0001, Figure 5A ). Moreover, the sensitivity of the diagnosis using the combined markers (PDZD11 + AFP) increased to 0.941, compared to 0.831 for AFP alone ( Table 2 ).

HCC patients were categorized into high- and low-expression groups according to the level of PDZD11 protein expression ( Supplementary Table S1 ). As illustrated in Supplementary Table S1 , the findings indicated a significant association between PDZD11 expression and the Child-Pugh class (p < 0.0001). Conversely, an analysis using PCA on the PS score, tumor dimensions, tumor count, AFP, TBA, TBIL, DBIL, ALT, AST, and WBC revealed no notable distinctions between the two groups ( Figure 5B ). Furthermore, univariate logistic regression analysis demonstrated a weak association between HCC and PDZD11, AFP, TBA, TBIL, DBIL, ALT, and AST levels ( Table 3 , Figure 5C ).

The PDZD11 levels of 12 patients, encompassing a total of 24 post-progression blood samples, were continuously monitored ( Figure 5D ). Additionally, we examined the clinical information of patients who had paired ‘before and after’ abdominal CT of tumor progression to evaluate the relationship between PDZD11 levels and tumor progression. For instance, one patient exhibited clinical progression as confirmed by abdominal CT two months after transarterial chemoembolization (TACE) combined with loplatin and pirarubicin interventional therapy. Although AFP levels remained within normal ranges both before and after progression, PDZD11 levels were significantly increased compared to pre-progression levels ( Figure 5E ). This suggests that PDZD11 may possess greater sensitivity than AFP in assessing the therapeutic response of HCC patients.