Exploratory Study on the Prognostic Value and Biological Interpretability of PGF-Related H&E Pathomics Models in Hepatocellular Carcinoma

Exploratory Study on the Prognostic Value and Biological Interpretability of PGF-Related H&E Pathomics Models in Hepatocellular Carcinoma | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Exploratory Study on the Prognostic Value and Biological Interpretability of PGF-Related H&E Pathomics Models in Hepatocellular Carcinoma Long Chen, Xusheng Zhang, Kejun Liu, Weihu Ma, Ling Ding, Shicai Liang, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6945731/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract PURPOSE Placental growth factor (PGF) is implicated in hepatocellular carcinoma (HCC) progression, although its functional role remains unclear. This study aimed to develop a pathomics model for predicting PGF expression from H&E-stained HCC sections and investigate its prognostic relevance and molecular mechanisms. METHODS Retrospective analysis utilized H&E images and clinical data from TCGA and an external cohort. Prognostic significance of PGF was assessed via survival analysis. Image segmentation employed the OTSU algorithm, followed by PyRadiomics-based feature extraction. Key features were selected using mRMR and RFE algorithms, with a gradient boosting machine (GBM) model constructed for PGF prediction. Model performance was validated through ROC, calibration, and decision curve analyses. Prognostic stratification, Cox regression, and subgroup analyses were conducted for high/low pathomics score (PS) groups. Bioinformatics approaches identified differentially expressed genes (DEGs) and immune infiltration patterns. RESULTS PGF expression independently predicted poor HCC prognosis (HR=1.922, 95% CI:1.217-3.036, P =0.005). The pathomics model incorporating seven PGF-associated features demonstrated robust predictive performance (AUC=0.811; external validation AUC=0.740). High-PS patients exhibited significantly worse survival (HR=1.667, 95% CI:1.024–2.713, P =0.040). DEGs in high-PS subgroups showed enrichment in ribosome and coagulation pathways, accompanied by upregulated HBEGF (inflammation-related) and increased γδT cell infiltration. TP53 mutations were prevalent (>20% mutation rate) in high-PS cases. CONCLUSION PGF serves as an independent prognostic biomarker in HCC. The developed pathomics model enables non-invasive PGF expression prediction through H&E image analysis. Mechanistically, PGF-associated molecular alterations involve inflammatory signaling, immune microenvironment remodeling, and frequent TP53 mutations, providing insights into HCC pathogenesis. Hepatocellular carcinoma Placental growth factor Pathomics Machine learning Mechanism analysis Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 1. Introduction Hepatocellular carcinoma(HCC) is the sixth most common malignant tumor and the fourth leading cause of cancer-related deaths worldwide[ 1 ]. HCC poses an enormous health burden globally in terms of morbidity and mortality, particularly in China[ 2 ]. Although surgical resection and liver transplantation are effective treatments for patients with HCC, numerous clinical studies indicate that the 5-year recurrence rate of postoperative HCC is as high as 70%[ 3 , 4 ] .Therefore, there is an urgent need for more markers to assess the prognosis of patients with HCC. PGF is a member of the vascular endothelial growth factor family. PGF not only promotes angiogenesis but also has a pro-inflammatory effect[ 5 ]. It was found that the expression of PGF was significantly elevated in HCC tumor tissues compared to normal tissues[ 6 , 7 ]. Meanwhile, PGF is closely related to the pathophysiological process of liver diseases, primarily involved in regulating the activation of epithelial-mesenchymal transition, and plays an important role in the development of other tumors[ 8 , 9 ]. Additionally,, a competing risk model based on PGF has demonstrated strong predictive value for HCC prognosis, and PGF may serve as a new target for anti-tumor immunity in HCC[ 10 ]. Given the potential value of PGF in the development and prognosis of hepatocellular carcinoma, it is crucial to accurately detect its expression level. However, current conventional assays have some limitations, such as high price, high time cost, difficulty in specimen collection, and results affected by operator and antibody.H&E-stained sections are essential for clinical diagnosis and the most accessible image data, which have the advantages of being widely used, comprehensive information, stable, and inexpensive. However, there are limitations for pathologists to predict biomarker expression by observing H&E sections alone. Pathomics is a field of research based on artificial intelligence techniques, such as machine learning, which transforms pathology images into high-fidelity, high-throughput data that can be mined. This field encompasses quantitative features, including texture features, morphological features, margin gradient features, and biological properties, and aims to quantify aspects of pathological diagnosis, molecular expression, and disease prognosis[ 11 – 13 ]. Previous studies have shown that machine learning can accurately recognize information in H&E images that is often not captured by pathologists through the human eye[ 14 , 15 ] .For example,Shamai et al.[ 16 ] successfully predicted the expression content of various molecular biomarkers, such as estrogen receptor, in breast cancer H&E sections with high accuracy using a machine learning approach. The pathomics model constructed by Zhaoyong Yan et al.[ 17 ] successfully predicted TNF receptor superfamily member 4 (TNFRSF4) expression, and the model had good predictive performance for molecular expression and revealed a significant association between high pathohistological scores and poorer OS.This study demonstrates that molecular prediction in H&E sections can be achieved using machine learning, overcoming the limitations of traditional methods. Hence,we aim to use machine learning to predict the expression level of PGF in H&E sections of HCC patients, enabling us to evaluate their prognosis. This approach will serve as a complementary tool for prognostic evaluation in HCC patients. Additionally, we will integrate bioinformatics analysis to explore the potential molecular mechanisms underlying Pathomics. 2. Methodology 2.1. Patient cohort The flow of this study is illustrated in (Fig. 1 ). The public database cohort used in this study was sourced from the publicly available TCGA dataset, and all data were de-identified and approved by the ethics committee for public research purposes. Conversely, the hospital cohort was obtained from the General Hospital of Ningxia Medical University and received approval from its Research Ethics Committee(KYLL-2025-0520). 2.1.1. Public database queue First, 377 patients with HCC were selected from the TCGA database ( https://portal.gdc.cancer.gov/ ). The inclusion criteria were (a) patients with available H&E-stained section images from surgical resections. (b) Samples with primary solid tumor RNA-seq.(c) samples with a primary diagnosis and initial treatment of hepatocellular carcinoma. The exclusion criteria were(a) samples lacking survival status, survival time, or having a survival time of less than one month. (b) samples with missing clinical data,or substandard quality of pathological images. Ultimately, this study included a sample of 267 HCC cases in the TCGA cohort. 2.1.2. Hospital cohort In addition, we included 80 HCC patients who underwent surgical treatment at the General Hospital of Ningxia Medical University from January 2016 to December 2018. Samples with hepatocellular carcinoma at both initial diagnosis and initial treatment were included, and samples with missing pathologic images or substandard pathologic image quality were excluded, resulting in 79 patients being included in the external validation cohort. 2.2. survival analysis We downloaded the HCC-related RNAseq data from the TCGA website, and analyzed the intergroup differences between the HCC tumor group and the normal group by Toil pipeline[ 18 ], and the model outputs the probability of predicting the gene expression level of Pathomics score after the modeling was completed, and compared the Pathomics score in the PGF high and low by Wilcoxon test. groupings with or without differences between the subgroups. Subsequently, by plotting the Kaplan-Meier survival curves, we demonstrated the survival differences between the high and low PGF expression groups as well as the survival differences between the high PS and low PS groups. In addition, we investigated whether high PGF expression as well as high PS expression were independent influences on overall survival (OS) using univariate and multivariate Cox regression analysis. Finally, we performed exploratory subgroup analyses by univariate COX regression to analyze the effects of PGF (high expression group vs. low expression group) and PS (high expression group vs. low expression group) on patients' prognosis in different subgroups of each covariate; the likelihood ratio test was used to analyze the interactions between PGF as well as PS and other covariates. 2.3. Segmentation and processing of pathology images Firstly, image segmentation and processing were applied to the pathology images of the TCGA cohort. Given that the entire pathology image contains a significant amount of extraneous information, we used the OTSU algorithm ( https://opencv.org/ ) to divide the image into two parts: the irrelevant background and the relevant tissue region for the study[ 19 ]. After removing the irrelevant background, the 40× image was segmented into multiple subimages of 1024×1024 pixels (Fig. 2 A), and the 20× image was segmented into multiple subimages of 512×512 pixels, which were then up-sampled to 1024×1024 pixels. A final review was conducted by a pathologist to exclude subimages with poor image quality, such as contamination, blurriness, or blank areas exceeding 50%. Subsequently, 10 subimages from each segmented pathology image were randomly selected for further analysis (Fig. 2 B) to reduce computational cost while ensuring sufficient data[ 20 , 21 ]. The pathology images from the external validation queue were photographed and uploaded by specialized equipment and stored as 40×magnification images for subsequent model validation. 2.4. Image feature extraction We utilized PyRadiomics, a Python-based tool ( https://pyradiomics.readthedocs.io/en/latest/ ), to perform image normalization and extract relevant features, including first-order, second-order, and higher-order features, from the subimages. After extracting features from each of the 10 subimages per patient, the corresponding mean was calculated and used as the pathomics eigenvalue of each sample for subsequent data analysis. 2.5. Feature screening and model construction The TCGA cohort was randomly divided into training and validation sets in a 7:3 ratio. We employed the Maximum Relevance Minimum Redundancy (mRMR) algorithm and the Recursive Feature Elimination (RFE) algorithm to identify the optimal subset of features. Initially, the mRMR algorithm was used to filter out a set of relevant features, followed by the RFE algorithm to further refine the selection until the best subset of features was obtained. The RFE algorithm continued to identify relevant features vital for generating accurate models. Finally, we utilized the Gradient Boosting Machine (GBM) to assign importance scores to the selected pathomics features, thereby constructing a model capable of predicting the expression of PGF. 2.6. Evaluation and validation of the model For model evaluation, we assessed the predictive performance of the model by plotting ROC curves (Receiver Operating Characteristic curves) and calculating the respective AUC values (Area Under the ROC Curve, AUC) for the training, validation, and test sets. The calibration of the predictive models was assessed by plotting calibration curves and conducting Hosmer-Lemeshow goodness-of-fit tests. Decision curve analysis (DCA) was also performed to illustrate the clinical benefits of the predictive models. For model validation, we incorporated all pathology image data into the constructed pathomics model, calculated the Pathomics Score for each sample, and evaluated differences in PS between the training set, validation set, and the PGF high and low subgroups in the test set using the Wilcoxon test. These results were then visualized. Subsequently, the PS was integrated with clinical data and categorized into Low/High dichotomous variables PS after determining cutoff values using the survminer package. 2.7. Pathohistological mechanism analysis To explore the molecular mechanisms underlying Pathomics, we conducted Gene Set Enrichment Analysis (GSEA) using the KEGG and Hallmark gene sets to evaluate the enrichment of differentially expressed genes in high/low PS groups. Subsequently, we analyzed the differential expression of 200 inflammation-related genes between the high and low PS subgroups using the Wilcoxon rank-sum test, focusing on genes with a p-value of less than 0.05 for visualization[ 22 , 23 ]. We uploaded the gene expression matrices of the HCC samples from the TCGA cohort to the ImmuCellAI database to calculate immune cell infiltration for each sample. We then analyzed the differences in immune cell infiltration in high/low PS groups using the Wilcoxon rank-sum test. Finally, we downloaded mutation data for TCGA-LIHC patients from the TCGA database and intersected it with 267 samples from the TCGA cohort to analyze the mutation data. We visualized the top 15 genes with the highest mutation frequency, assessing the relationship between the mutation profiles of TCGA-LIHC patients and the PS grouping model predictions. 2.8. Statistical Methods Continuous variables were compared between groups using the Wilcoxon test. The calibration of the pathohistological prediction model was assessed by plotting ROC and calibration curves and performing the Hosmer-Lemeshow goodness-of-fit test. Kaplan-Meier survival curves illustrated variations in survival rates across different groups, and the log-rank test evaluated the significance of survival rate differences between groups. Model performance was assessed through likelihood ratio test analysis. All analyses were conducted using R software version 4.1.0, with hypothesis testing being two-sided and a p-value of less than 0.05 indicating a significant difference. 3. Results 3.1. Correlation analysis of PGF with HCC In both the TCGA cohort and the external validation cohort, we categorized patients into high- and low-expression groups based on the level of PGF expression and created baseline information tables for each cohort ( Appendix Table 1 、 Appendix Table 2 ). The results showed no statistically significant differences in the distribution of clinical variables between the groups and were comparable.Analysis of variance showed that the PGF expression level in HCC tumor tissues was significantly higher than that in normal tissues ( P < 0.0001) (Fig. 3 A). Kaplan-Meier curves showed that there was a significant difference in overall survival between patients with high and low PGF expression in the TCGA cohort ( P = 0.001),the median survival time of patients in the low PGF expression group being 84.4 months, compared to only 46.2 months for those in the high PGF expression group(Fig. 3 B). In univariate analysis (Fig. 3 C), high PGF expression emerged as a risk factor for overall survival (OS) in patients (HR = 2.123, 95% CI: 1.368–3.293, P < 0.001). When included in multivariate analysis (Fig. 3 D), high PGF expression remained a distinct risk factor for OS (HR = 1.922, 95% CI: 1.217–3.036, P = 0.005). In the interaction test analysis (Fig. 3 E), there was an interaction between the expression content of PGF and the prognosis of patients in the three subgroups of whether or not they were treated with medication, whether or not they had undergone ablative embolization, and the level of blood AFP ( P 400, elevated PGF did not correlate with patient prognosis. However, in subgroups not defined by these conditions, PGF expression levels could accurately reflect patient prognosis. While drug therapy and ablation embolization are commonly used clinical treatments that can affect patient survival time to varying degrees[ 24 ],they also influence PGF expression levels, thereby complicating the assessment of HCC patient prognosis based on PGF expression. In the subgroup with AFP > 400, PGF could not effectively predict the prognosis of HCC patients. This may be due to AFP's influence on PGF expression levels, though no detailed reports exist on the relationship between these factors in HCC patients domestically or internationally. This observation presents a novel perspective for exploring the impact of molecular levels on the survival prognosis of HCC patients. 3.2. Pathology image feature extraction and model construction A total of 267*10 sub-images of 1024×1024 pixels were obtained after segmenting and filtering the pathology images of the TCGA cohort. After image normalization, 1,488 features were extracted and finally the average value was taken to represent the histopathological features of this sample.[ 12 , 25 , 26 ]. The TCGA cohort was randomly divided into training and validation sets with a 7:3 ratio, comprising 188 cases in the training set and 79 cases in the validation set. A baseline profile table was created, indicating that the analysis of variance for each variable showed no significant differences between the groups ( P > 0.05) ( Appendix Table 3 ). Subsequently, in the training set, we used the mRMR method to screen the top 30 image features with strong predictive efficacy. These features were further refined using the RFE method, resulting in a total of 7 features significantly correlated with the prediction of the PGF molecule for subsequent analysis (Fig. 4 A).Finally, we calculated the importance scores of these 7 features in the model by the GBM algorithm (Fig. 4 B), and constructed the Pathomics model in the training set by the relevant features and their corresponding weights. 3.3. Evaluation of the model The results of our model efficacy evaluation showed that the HCC Pathomics model features had high predictive accuracy in both the training set (AUC = 0.881) and the validation set (AUC = 0.747) (Fig. 5 A,B). In addition, in the external test cohort (AUC = 0.74), the model had equally good predictive efficacy (Fig. 5 C).We then plotted calibration curves for the training, validation, and test datasets (Fig. 5 D, E, and F), and the calibration curves and the Hosmer-Lemeshow goodness-of-fit test showed that the pathohistological prediction model's predictive probability of whether PGF was highly expressed was in good agreement with the true value ( P > 0.05); the DCA curves (Fig. 5 G, H, and I) demonstrated that the model had clinical utility. 3.4.Prognostic value of the model We divided the training, validation, and test sets into high expression and low expression groups based on PGF expression levels. By comparing the distribution of PS values across these three datasets, we observed that the PS of the PGF high expression group were significantly higher than those of the PGF low expression group (Fig. 6 A, B, C). The TCGA cohort was divided into high and low PS groups based on their PS. A baseline information table for each clinical variable was created ( Appendix Table 4 ). The KM curves plotted (Fig. 6 D) showed that the median survival time for the high PS group was 56.17 months, while it was 82.87 months for the low PS group. High PS was significantly associated with lower OS in HCC patients ( P = 0.006), highlighting the positive predictive value of predictive models in assessing patient prognosis.Further univariate and multivariate Cox regression analyses revealed that high PS was an independent risk factor for overall survival (OS) (Fig. 7 A, B).Interaction test analysis showed (Fig. 7 C, D) that there was no significant interaction between PS and the majority of subgroups, indicating that PS scores are reliable predictors of patient prognosis across different subgroups for each covariate. 3.5. biological interpretability In the KEGG gene set ( Fig. 8 A), signaling pathways such as RIBOSOME and COAGUATION demonstrated significant differences between the high/low PS groups. In the Hallmark gene set ( Fig. 8 B), the COAGUATION signaling pathway likewise demonstrated obvious differences.. In the differential analysis of inflammation-related genes between the high and low PS subgroups, we found (Fig. 8 C) that the expression of genes such as HBEGF was significantly elevated in the high PS group. In the immune cell infiltration analysis (Fig. 8 D), γδT cells were significantly different in the high and low PS subgroups ( P < 0.01). Finally, Differential analysis of somatic gene mutations in the high/low PS groups showed that the mutation rate of the TP53 (29% vs. 25%) gene differed from that of the low PS group in the high PS group, and the mutation rate was higher than 20% in both the high and low PS groups(Fig. 8 E, F). 4. Discussion We developed a pathohistological model based on PGF expression levels for predicting the prognosis of HCC patients by PS, and validated its clinical effects as well as explored the biological interpretability behind it by an external test cohort. The main findings of this study were 1) PGF expression level had an independent prognostic value in HCC (HR = 1.922, 95% CI: 1.217–3.036, P = 0.005);2) a HE pathohistochemistry model consisting of seven pathological features was constructed to achieve non-invasive prediction of PGF (AUC = 0.811) and was externally test-cohort validation (AUC = 0.740); 3) the biological interpretability of the pathohistological model revealed that PS mapping PGF expression levels was associated with signaling pathways such as RIBOSOME and COAGUATION, inflammation-related genes such as HBEGF , immune cells such as γδT cells, as well as mutations in TP53 genes. PGF is a member of the pro-angiogenic VEGF family and is directly or indirectly involved in promoting immune escape and tumor metastasis, playing a significant role in tumor development[ 8 ]. Our study found a strong association between PGF and poor prognosis in HCC, indicating that the PGF molecule is an independent risk factor for the survival time of HCC patients. Furthermore, upregulation of PGF has been observed in cirrhotic patients[ 27 ]. And there is increasing evidence that overexpression of PGF is associated with early recurrence of HCC, and PGF may be an important prognostic indicator for HCC[ 28 , 29 ].Additionally, the research by Miao et al.[ 7 ] demonstrated a significant difference in overall survival (OS) between high and low levels of PGF expression in HCC patients. PGF is not only valuable in hepatocellular carcinoma but also in other tumors, such as colorectal cancer, renal cell carcinoma, and pancreatic cancer, as it can reflect the prognosis of patients to a certain extent[ 30 – 32 ] . The development of pathomics models represents a major leap forward in the integration of histopathology and molecular biology.Currently, with the rapid development of the field of Pathomics, its applications have expanded from the histological level, including tissue composition and cell types, to molecular pathology and disease prognosis. By utilizing various data mining algorithms based on machine learning methods, high-dimensional data can be transformed into predictive or prognostic information[ 33 – 35 ]. Recent seminal findings have shown the wide application of machine learning methods in Pathomics, demonstrating high value in metastasis detection in lymph nodes, Ki67 scoring in breast cancer, Gleason grading in prostate cancer, and tumor-infiltrating lymphocyte scoring in melanoma[ 36 ]. We have successfully constructed a H&E pathomics model from cases in the TCGA database using a machine learning approach to achieve non-invasive prediction of PGF expression levels. Most importantly, in an external validation cohort consisting of cases from our hospital, the model still demonstrated excellent prediction accuracy, even though the two cohorts were extremely different in terms of ethnicity, region, and so on. This demonstrates the stability and generalizability of H&E pathology images for molecular prediction, which can help us realize the possibility of molecular expression prediction from a pathology perspective. This non-invasive prediction may reach a newer height with the advancement of image analysis and processing tools. Meanwhile, models constructed from multicenter and large sample sizes will demonstrate higher predictive accuracy, thus providing a more accurate assessment of patient prognosis. Therefore, in this study, histopathological image characterization may be a more convenient, low-cost, and accurate alternative strategy for predicting the molecular expression levels of PGF in HCC patients. One of the most compelling aspects of our study is the biological interpretability of the pathomics model, we performed differential gene enrichment analysis of inflammation-related genes, abundance of immune cell infiltration, and gene mutation profiles in the high pathology score group and low pathology score group, respectively. Our gene set enrichment analysis (GSEA) revealed significant differences in the coagulation and ribosome pathways between high and low PGF expression groups., which is worthy of consideration.Coagulation factors can enhance metastasis by increasing the adhesion of cancer cells and enhancing the activation of endothelial cells[ 37 ], while the coagulation cascade response plays a crucial role in shaping the tumor-immune microenvironment (TME)[ 38 ] .And PGF as a key gene related to coagulation and fibrinolysis in HCC can affect the occurrence and development of HCC through TME[ 17 , 39 ]. The ribosome pathway, on the other hand, is closely associated with protein synthesis and cell proliferation. Upregulation of PGF activates the mTOR pathway, which is involved in the regulation of multiple processes involved in ribosome biosynthesis and can control protein synthesis by promoting ribosome biosynthesis [ 40 – 42 ].Hyperactivation of ribosome biogenesis leads to hepatocellular transformation and plays a key role in the development of HCC[ 43 ].Inflammation is an important hallmark of cancer and exhibits an important role in both tumorigenesis and progression[ 23 ], In HCC, the relationship between tumors and the inflammatory response is even closer, as the outcome of hepatitis leading to cirrhosis and the eventual development of HCC is well known.In the differential analysis of inflammation-related genes, we found that a considerable number of inflammatory genes were expressed at higher levels in the high PS group compared with the low PS group, especially HBEGF was more significantly expressed. Increased local blood supply is one of the characteristics of HCC development. In a basic study of peripheral arterial disease (PAD), researchers caused an increase in local blood supply in the hind limbs of mice by inhibiting the relevant genes, and the expression of HBEGF and PGF was up-regulated, suggesting that HBEGF may cause alterations in local blood supply through angiogenesis or microenvironmental remodeling[ 44 ]. According to previous studies and the results of this paper, HBEGF and PGF may play a common role in the development of tumor lesions or some benign vascular diseases. In our analysis of immune cell infiltration in hepatocellular carcinoma we found that γδT cells infiltrated to a higher extent in the PS low expression group than in the high expression group. γδT cells are a subset of T cells that play a role in anti-tumor immunity, but their function is often impaired in HCC patients. He et al.[ 45 ] pointed out that γδT cells were involved in mediating anti-tumor responses and that their infiltration effects correlate with HCC progression and patient prognosis, whereas in the development of HCC, γδT cells demonstrate a decreasing degree of infiltration and diminished anti-tumor effects. In HCC patients, PGF regulates γδT cell activation and function through the MAPK-RAP1A signaling pathway, thereby promoting immune evasion and tumor progression [ 46 ]. This highlights the potential of targeting PGF and its downstream signaling pathways as a strategy for enhancing anti-tumor immunity in HCC. In addition, TP53 mutations are the most common type of genetic variation in HCC, with the largest mutation type being missense mutations, which are involved in the regulation of DNA repair, cell cycle arrest and apoptosis[ 47 ]. The high mutation rate of TP53 in our study (29% in the high PGF group and 25% in the low PGF group) underscores the importance of this gene in HCC pathogenesis and highlights the need for targeted therapies that can restore TP53 function or bypass its tumor-suppressive effects. Our findings have far-reaching clinical implications. The ability to noninvasively predict PGF expression levels using routine H&E staining images opens up new possibilities for personalized medicine in HCC. By combining pathomics with other holographic data, such as genomics and proteomics, clinicians can gain a more comprehensive understanding of tumor biology and develop treatment strategies accordingly. For example, patients with high PGF expression may benefit from therapies targeting angiogenic or coagulation pathways, while patients with TP53 mutations may require alternative approaches such as immunotherapy or targeted gene therapy. In addition, the identification of inflammation-related genes and patterns of immune cell infiltration associated with PGF expression provides new avenues for immunotherapy development. By targeting the molecular pathways and immune mechanisms affected by PGF, it may be possible to increase the efficacy of existing therapies and improve patient prognosis. However, several challenges remain, such as heterogeneity between different patient cohorts and possible confounding factors in retrospective studies, which may limit the generalizability of our model. Future studies should aim to validate our findings in larger multicenter cohorts and explore the effects of other clinical variables, such as comorbidities and treatment regimens, on PGF expression and HCC prognosis. Furthermore, although our pathomics model showed high predictive accuracy, there is still room for improvement. The relatively low AUC in the external validation cohort (0.740 compared to 0.811 in the training cohort) suggests the need to further optimize the model parameters and incorporate more features to improve its stability and predictive power. Advances in image analysis and machine learning algorithms, coupled with the integration of multi-omics data, are expected to further refine the model and expand its application. 5. Conclusion In conclusion, our study highlights the prognostic value of PGF expression in HCC and demonstrates the potential of pathomics as a non-invasive tool for predicting molecular expression and assessing patient prognosis. By integrating machine learning with bioinformatics analysis, we have uncovered novel insights into the molecular mechanisms underlying HCC progression, paving the way for more personalized and effective treatment strategies. As the field of pathomics continues to evolve, it is poised to play an increasingly important role in the era of precision medicine, offering new hope for patients with HCC and other malignancies. Declarations Funding The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. Author Contributions Long Chen and Xusheng Zhang contributed equally. Long Chen, Xusheng Zhang, and Bendong Chen participated in the conceptualization and design of this study. Long Chen and WeiHu Ma organized the database and statistical analysis. Long Chen, Bendong Chen, Lin Ding,Shicai Liang,Xuebo Wang, and KeJun Liu divided the work and participated in the drawing. Long Chen wrote the first manuscript. Bendong Chen and KeJun Liu participated in the revision of the manuscript. All authors read and agreed to the final manuscript and authorship arrangement. Consent for publication Not applicable to this study. Data Availability and Ethical Statement The dataset supporting the conclusions of this article is available in the [TCGA] repository, unique persistent identifier and hyperlink to dataset in https://portal.gdc.cancer.gov/projects/TCGA-LIHC.The external test cohort was sourced from the General Hospital of Ningxia Medical University and approved by the Research Ethics Committee of the General Hospital of Ningxia Medical University (Ethics Approval Number: KYLL-2025-0520), adhering to the ethical standards outlined in the Declaration of Helsinki (2013 revision). Informed Consent Statement Each participant in the study gave their consent. Conflicts of Interest We declare that there is no conflict of interest regarding the publication of this paper. Acknowledgement This study is especially grateful for the help provided by all the doctors from the Department of Pathology, General Hospital of Ningxia Medical University. Disclaimer: Since the four appendices in the article take up a lot of space, I would like to present them as supplementary files. If your journal would like to include the appendices in the main text, I can assist with the re-formatting. References Brown, Z.J., et al., Management of Hepatocellular Carcinoma: A Review. JAMA Surg, 2023. 158(4): p. 410-420. Yang, J.D., et al., A global view of hepatocellular carcinoma: trends, risk, prevention and management. Nat Rev Gastroenterol Hepatol, 2019. 16(10): p. 589-604. Byrd, K., et al., Role of Multidisciplinary Care in the Management of Hepatocellular Carcinoma. Semin Liver Dis, 2021. 41(1): p. 1-8. Llovet, J.M., et al., Hepatocellular carcinoma. Nat Rev Dis Primers, 2021. 7(1): p. 6. Fischer, C., et al., Anti-PlGF inhibits growth of VEGF(R)-inhibitor-resistant tumors without affecting healthy vessels. Cell, 2007. 131(3): p. 463-75. Xu, H.X., et al., Expression and prognostic significance of placental growth factor in hepatocellular carcinoma and peritumoral liver tissue. Int J Cancer, 2011. 128(7): p. 1559-69. Miao, Y.D., et al., Prognostic role of expression of angiogenesis markers in hepatocellular carcinoma: A bioinformatics analysis. World J Gastroenterol, 2022. 28(30): p. 4221-4226. Albonici, L., et al., Multifaceted Role of the Placental Growth Factor (PlGF) in the Antitumor Immune Response and Cancer Progression. Int J Mol Sci, 2019. 20(12). Simbrunner, B., et al., Placental growth factor levels neither reflect severity of portal hypertension nor portal-hypertensive gastropathy in patients with advanced chronic liver disease. Dig Liver Dis, 2021. 53(3): p. 345-352. Shen, B., et al., Identification and Analysis of Immune-Related Gene Signature in Hepatocellular Carcinoma. Genes (Basel), 2022. 13(10). Liu, K. and J. Hu, Classification of acute myeloid leukemia M1 and M2 subtypes using machine learning. Comput Biol Med, 2022. 147: p. 105741. Nishio, M., et al., Homology-Based Image Processing for Automatic Classification of Histopathological Images of Lung Tissue. Cancers (Basel), 2021. 13(6). Banna, G.L., et al., The Promise of Digital Biopsy for the Prediction of Tumor Molecular Features and Clinical Outcomes Associated With Immunotherapy. Front Med (Lausanne), 2019. 6: p. 172. Beck, A.H., et al., Systematic analysis of breast cancer morphology uncovers stromal features associated with survival. Sci Transl Med, 2011. 3(108): p. 108ra113. Yu, K.H., et al., Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features. Nat Commun, 2016. 7: p. 12474. Shamai, G., et al., Artificial Intelligence Algorithms to Assess Hormonal Status From Tissue Microarrays in Patients With Breast Cancer. JAMA Netw Open, 2019. 2(7): p. e197700. Yan, Z., et al., Prognostic significance of TNFRSF4 expression and development of a pathomics model to predict expression in hepatocellular carcinoma. Heliyon, 2024. 10(11): p. e31882. Vivian, J., et al., Toil enables reproducible, open source, big biomedical data analyses. Nat Biotechnol, 2017. 35(4): p. 314-316. Wang, X., et al., Weakly Supervised Deep Learning for Whole Slide Lung Cancer Image Analysis. IEEE Trans Cybern, 2020. 50(9): p. 3950-3962. Chen, L., et al., Histopathological image and gene expression pattern analysis for predicting molecular features and prognosis of head and neck squamous cell carcinoma. Cancer Med, 2021. 10(13): p. 4615-4628. Zeng, H., et al., Integration of histopathological images and multi-dimensional omics analyses predicts molecular features and prognosis in high-grade serous ovarian cancer. Gynecol Oncol, 2021. 163(1): p. 171-180. Liang, Y., et al., Identification and Validation of a Novel Inflammatory Response-Related Gene Signature for the Prognosis of Colon Cancer. J Inflamm Res, 2021. 14: p. 3809-3821. Zhai, W.Y., et al., A Novel Inflammatory-Related Gene Signature Based Model for Risk Stratification and Prognosis Prediction in Lung Adenocarcinoma. Front Genet, 2021. 12: p. 798131. Baek, M.Y., et al., Clinical outcomes of patients with a single hepatocellular carcinoma less than 5 cm treated with transarterial chemoembolization. Korean J Intern Med, 2019. 34(6): p. 1223-1232. Saednia, K., et al., Quantitative digital histopathology and machine learning to predict pathological complete response to chemotherapy in breast cancer patients using pre-treatment tumor biopsies. Sci Rep, 2022. 12(1): p. 9690. Li, H., et al., Integrative Analysis of Histopathological Images and Genomic Data in Colon Adenocarcinoma. Front Oncol, 2021. 11: p. 636451. Huang, X.X., et al., Up-regulation of proproliferative genes and the ligand/receptor pair placental growth factor and vascular endothelial growth factor receptor 1 in hepatitis C cirrhosis. Liver Int, 2007. 27(7): p. 960-8. Ho, M.C., et al., Placenta growth factor not vascular endothelial growth factor A or C can predict the early recurrence after radical resection of hepatocellular carcinoma. Cancer Lett, 2007. 250(2): p. 237-49. Nagaoka, S., et al., The ratio of serum placenta growth factor to soluble vascular endothelial growth factor receptor-1 predicts the prognosis of hepatocellular carcinoma. Oncol Rep, 2010. 23(6): p. 1647-54. Shen, Y., et al., Role of stemness-related genes TIMP1, PGF, and SNAI1 in the prognosis of colorectal cancer through single-cell RNA-seq. Cancer Med, 2023. 12(10): p. 11611-11623. Bae, S.H., T. Hwang and M.R. Han, Unraveling the hypoxia modulating potential of VEGF family genes in pan-cancer. Genomics Inform, 2023. 21(4): p. e44. Gohrig, A., et al., Placental growth factor promotes neural invasion and predicts disease prognosis in resectable pancreatic cancer. J Exp Clin Cancer Res, 2024. 43(1): p. 153. Liao, H., et al., Preoperative Radiomic Approach to Evaluate Tumor-Infiltrating CD8(+) T Cells in Hepatocellular Carcinoma Patients Using Contrast-Enhanced Computed Tomography. Ann Surg Oncol, 2019. 26(13): p. 4537-4547. Xu, X., et al., Radiomic analysis of contrast-enhanced CT predicts microvascular invasion and outcome in hepatocellular carcinoma. J Hepatol, 2019. 70(6): p. 1133-1144. Kim, S., et al., Radiomics on Gadoxetic Acid-Enhanced Magnetic Resonance Imaging for Prediction of Postoperative Early and Late Recurrence of Single Hepatocellular Carcinoma. Clin Cancer Res, 2019. 25(13): p. 3847-3855. Acs, B., M. Rantalainen and J. Hartman, Artificial intelligence as the next step towards precision pathology. J Intern Med, 2020. 288(1): p. 62-81. Hill, C.N., et al., Deciphering the Role of the Coagulation Cascade and Autophagy in Cancer-Related Thrombosis and Metastasis. Front Oncol, 2020. 10: p. 605314. He, Q., J. Yang and Y. Jin, Immune infiltration and clinical significance analyses of the coagulation-related genes in hepatocellular carcinoma. Brief Bioinform, 2022. 23(4). Fan, M., et al., Establishment and verification of a prognostic model based on coagulation and fibrinolysis-related genes in hepatocellular carcinoma. Aging (Albany NY), 2024. 16(9): p. 7578-7595. Abetov, D.A., et al., Formation of mammalian preribosomes proceeds from intermediate to composed state during ribosome maturation. J Biol Chem, 2019. 294(28): p. 10746-10757. Gentilella, A., S.C. Kozma and G. Thomas, A liaison between mTOR signaling, ribosome biogenesis and cancer. Biochim Biophys Acta, 2015. 1849(7): p. 812-20. He, J., et al., Ribosome biogenesis protein Urb1 acts downstream of mTOR complex 1 to modulate digestive organ development in zebrafish. J Genet Genomics, 2017. 44(12): p. 567-576. Yang, X.M., et al., Nucleolar HEAT Repeat Containing 1 Up-regulated by the Mechanistic Target of Rapamycin Complex 1 Signaling Promotes Hepatocellular Carcinoma Growth by Dominating Ribosome Biogenesis and Proteome Homeostasis. Gastroenterology, 2023. 165(3): p. 629-646. Kant, S., et al., Neural JNK3 regulates blood flow recovery after hindlimb ischemia in mice via an Egr1/Creb1 axis. Nat Commun, 2019. 10(1): p. 4223. He, W., et al., Hepatocellular carcinoma-infiltrating gammadelta T cells are functionally defected and allogenic Vdelta2(+) gammadelta T cell can be a promising complement. Clin Transl Med, 2022. 12(4): p. e800. Li, H., et al., MAPK-RAP1A Signaling Enriched in Hepatocellular Carcinoma Is Associated With Favorable Tumor-Infiltrating Immune Cells and Clinical Prognosis. Front Oncol, 2021. 11: p. 649980. Zhang, X., Z. Fu and X. Zhang, TP53 Mutation Related and Directly Regulated lncRNA Prognosis Markers in Hepatocellular Carcinoma. Onco Targets Ther, 2021. 14: p. 4427-4437. Additional Declarations No competing interests reported. Supplementary Files AppendixTable.zip Appendix Table 1 Patients in the TCGA cohort were divided into high- and low-expression groups based on PGF expression levels, and a baseline information table was created.The results showed that there were no statistically significant differences in the distribution of clinical variables between the two groups, and they were comparable. Appendix Table 2 Patients were divided into high- and low-expression groups based on PGF expression levels in the external testing queue, and a baseline information table was created.The results showed that there were no statistically significant differences in the distribution of clinical variables between the two groups, and they were comparable. Appendix Table 3 The TCGA cohort was randomly divided into training and validation sets at a ratio of 7:3, and a baseline feature table was created. The analysis results showed that the analysis of variance between the two groups did not reveal any significant differences (P>0.05). Appendix Table 4 The TCGA cohort was divided into high PS and low PS groups based on their PS scores, and a baseline information table was created. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 04 Aug, 2025 Reviewers invited by journal 04 Aug, 2025 Editor invited by journal 09 Jul, 2025 Editor assigned by journal 25 Jun, 2025 Submission checks completed at journal 25 Jun, 2025 First submitted to journal 21 Jun, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6945731","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":495716576,"identity":"7a57c6c0-bbb3-4078-8a78-d9d366d9cb84","order_by":0,"name":"Long Chen","email":"","orcid":"","institution":"Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Long","middleName":"","lastName":"Chen","suffix":""},{"id":495716577,"identity":"a38351b3-fc1a-4af5-92ca-2cfec7b46331","order_by":1,"name":"Xusheng Zhang","email":"","orcid":"","institution":"Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xusheng","middleName":"","lastName":"Zhang","suffix":""},{"id":495716578,"identity":"a33c89b6-aa1c-49a4-b2f3-7ce1f8891de4","order_by":2,"name":"Kejun Liu","email":"","orcid":"","institution":"General Hospital of Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Kejun","middleName":"","lastName":"Liu","suffix":""},{"id":495716579,"identity":"9784eb18-5db3-4c0f-9da4-b3dcbeb93123","order_by":3,"name":"Weihu Ma","email":"","orcid":"","institution":"Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Weihu","middleName":"","lastName":"Ma","suffix":""},{"id":495716580,"identity":"a1767cf7-ea9a-40e6-9cfb-11d41b874a5e","order_by":4,"name":"Ling Ding","email":"","orcid":"","institution":"Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ling","middleName":"","lastName":"Ding","suffix":""},{"id":495716581,"identity":"384f3dda-9c1a-4432-ab06-f83211ea9c6e","order_by":5,"name":"Shicai Liang","email":"","orcid":"","institution":"Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Shicai","middleName":"","lastName":"Liang","suffix":""},{"id":495716582,"identity":"c73ee5b5-d269-4992-8a81-621e618dadd7","order_by":6,"name":"Xuebo Wang","email":"","orcid":"","institution":"Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xuebo","middleName":"","lastName":"Wang","suffix":""},{"id":495716583,"identity":"e72688c2-bc15-4cbc-a92e-7e31e3f07223","order_by":7,"name":"Bendong Chen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8UlEQVRIiWNgGAWjYDACCQYGZiDF2ADED6BiBkRrYYYpJV4LmwRRWvhnNx97XFBxR7Zfuv1adWHbtsQG9uZtEgw1d3BbcudYuvGMM8+MZ845U3Z7xpnbiQ08x8okGI49w6nFQCLHTJq37XDihhs5abd5KoBagCISjA2H8WjJ/ybN+w+ipZjHAKhF/g0hLTls0rwNIC3px5ghtvDg1yJxI81Mesaxw8YzZ+QwS/OcuW3cxpNWbJFwDLcW/hnJz6QLag7L9kukP/zM23Zbtp/98MYbH2pwa0ECPJDoYAMRCcRoYGBgf0CculEwCkbBKBhxAAAFDVc3prPbwQAAAABJRU5ErkJggg==","orcid":"","institution":"General Hospital of Ningxia Medical University","correspondingAuthor":true,"prefix":"","firstName":"Bendong","middleName":"","lastName":"Chen","suffix":""}],"badges":[],"createdAt":"2025-06-21 15:08:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6945731/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6945731/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88490235,"identity":"9fe3f018-01cd-4d46-97de-d7c1469c72d7","added_by":"auto","created_at":"2025-08-07 04:10:22","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":8394459,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFlowchart of this study to construct clinical molecular prediction model based on machine learning approach.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/22aa446d1e3097b0517b8b3b.png"},{"id":88491182,"identity":"c2a42f04-9872-4d32-b44e-e46df624c528","added_by":"auto","created_at":"2025-08-07 04:18:23","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":38608180,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSegmentation of H\u0026amp;E image with sub-images.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/26dd2a3c7fe250b11f07b78b.png"},{"id":88491894,"identity":"d9608557-7fc3-4642-b2a2-4b937d5034ec","added_by":"auto","created_at":"2025-08-07 04:26:22","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":5322037,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCorrelation analysis between PGF and HCC. \u003c/strong\u003eA.Comparison of PGF expression in tumor tissues and normal tissues in TCGA pairs of columns; B. Kaplan-Meier survival curves demonstrating the changes in survival between high and low PGF expression groups; C, D. One-way \u0026amp; multifactorial Cox regression analyses exploring whether high expression of PGF is an independent influencing factor for OS. E, F. Thery are interaction tests and subgroup analysis plots to analyze the interaction between the PGF high expression group and the low expression group in different subgroups of each covariate and the effect on patients' prognosis in each subgroup.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/5d023efd4ae153d639b2e3dc.png"},{"id":88491891,"identity":"a2db6573-1482-413c-b244-600d1827befd","added_by":"auto","created_at":"2025-08-07 04:26:22","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1062167,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConstruction of a machine learning-based prediction model for PGF expression level.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA. screening of pathological features by machine learning; B. constructing a prediction model by the seven pathological features screened.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/bbdf65c06b8da9c8b566480b.png"},{"id":88492489,"identity":"18b5b1e3-c493-447b-9003-8f1d3559cc4a","added_by":"auto","created_at":"2025-08-07 04:34:23","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":10214945,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEvaluation of the prediction model. \u003c/strong\u003eA, B and C are the ROC curves for the molecular prediction scores of the training, validation, and test cohorts, respectively; D, E and F are the calibration curves for the training, validation, and test cohorts, respectively; and G, F, and I are the DCA curves for the training, validation, and test cohorts, respectively.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/6a612cd1d69fb11c0113bcf5.png"},{"id":88491166,"identity":"259beda5-84e3-4032-9b40-dc17bf7de700","added_by":"auto","created_at":"2025-08-07 04:18:22","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2541927,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eValidation of the prediction model.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA, B and C represent the plots of variability in PS scores grouped by high and low PGF expression across the training cohort, validation cohort, and test cohort, respectively. D illustrates the KM curve plotted in the TCGA dataset, categorized by high and low PS.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/5c96a147b2d818174e390a12.png"},{"id":88491897,"identity":"fb6b93ed-bdd0-4f4c-bb24-e5a0fd46c461","added_by":"auto","created_at":"2025-08-07 04:26:22","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":3893886,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eValidation of the prediction model \u003c/strong\u003eA and B are univariate \u0026amp; multivariate Cox regression analyses related to PS to explore whether PS score is an independent influence on OS; C and D are interaction tests and subgroup analysis plots to analyze whether there is an interaction between high and low PS expression and each covariate as well as to analyze the effect of high and low PS expression on patients' prognosis in different subgroups of each covariate.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/85c4f6dc8ce1c08ffe90ff00.png"},{"id":88490249,"identity":"ec935c22-737c-46b8-a9aa-f5dd92fe21b7","added_by":"auto","created_at":"2025-08-07 04:10:22","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":13442889,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePathohistological mechanism analysis.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA and B are plots of enrichment analysis of differential genes between high and low PS subgroups as a result of model prediction; C is a plot of differential analysis of PS and inflammation-related genes as a result of model prediction; D is a plot of differential analysis of PS and immune cell abundance as a result of model prediction; and E is a plot of mutation analysis of genes between high and low PS subgroups as a result of model prediction.\u003c/p\u003e","description":"","filename":"image8.png","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/f3fb60a8b8893df05e208972.png"},{"id":88493079,"identity":"32613b6d-9c3b-4f61-8bcb-51969a321f22","added_by":"auto","created_at":"2025-08-07 04:43:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":76769284,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/ebcbd001-cee5-4331-8774-fc4fceacef51.pdf"},{"id":88490238,"identity":"e240998a-3079-473e-b072-621dcea0e8f6","added_by":"auto","created_at":"2025-08-07 04:10:22","extension":"zip","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":315432,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAppendix Table 1\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatients in the TCGA cohort were divided into high- and low-expression groups based on PGF expression levels, and a baseline information table was created.The results showed that there were no statistically significant differences in the distribution of clinical variables between the two groups, and they were comparable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAppendix Table 2\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatients were divided into high- and low-expression groups based on PGF expression levels in the external testing queue, and a baseline information table was created.The results showed that there were no statistically significant differences in the distribution of clinical variables between the two groups, and they were comparable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAppendix Table 3\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe TCGA cohort was randomly divided into training and validation sets at a ratio of 7:3, and a baseline feature table was created. The analysis results showed that the analysis of variance between the two groups did not reveal any significant differences (P\u0026gt;0.05).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAppendix Table 4\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe TCGA cohort was divided into high PS and low PS groups based on their PS scores, and a baseline information table was created.\u003c/p\u003e","description":"","filename":"AppendixTable.zip","url":"https://assets-eu.researchsquare.com/files/rs-6945731/v1/b126a594dbd6598f05a9e306.zip"}],"financialInterests":"No competing interests reported.","formattedTitle":"Exploratory Study on the Prognostic Value and Biological Interpretability of PGF-Related H\u0026E Pathomics Models in Hepatocellular Carcinoma","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eHepatocellular carcinoma(HCC) is the sixth most common malignant tumor and the fourth leading cause of cancer-related deaths worldwide[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. HCC poses an enormous health burden globally in terms of morbidity and mortality, particularly in China[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Although surgical resection and liver transplantation are effective treatments for patients with HCC, numerous clinical studies indicate that the 5-year recurrence rate of postoperative HCC is as high as 70%[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] .Therefore, there is an urgent need for more markers to assess the prognosis of patients with HCC.\u003c/p\u003e\u003cp\u003ePGF is a member of the vascular endothelial growth factor family. PGF not only promotes angiogenesis but also has a pro-inflammatory effect[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. It was found that the expression of PGF was significantly elevated in HCC tumor tissues compared to normal tissues[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Meanwhile, PGF is closely related to the pathophysiological process of liver diseases, primarily involved in regulating the activation of epithelial-mesenchymal transition, and plays an important role in the development of other tumors[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Additionally,, a competing risk model based on PGF has demonstrated strong predictive value for HCC prognosis, and PGF may serve as a new target for anti-tumor immunity in HCC[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eGiven the potential value of PGF in the development and prognosis of hepatocellular carcinoma, it is crucial to accurately detect its expression level. However, current conventional assays have some limitations, such as high price, high time cost, difficulty in specimen collection, and results affected by operator and antibody.H\u0026amp;E-stained sections are essential for clinical diagnosis and the most accessible image data, which have the advantages of being widely used, comprehensive information, stable, and inexpensive. However, there are limitations for pathologists to predict biomarker expression by observing H\u0026amp;E sections alone.\u003c/p\u003e\u003cp\u003ePathomics is a field of research based on artificial intelligence techniques, such as machine learning, which transforms pathology images into high-fidelity, high-throughput data that can be mined. This field encompasses quantitative features, including texture features, morphological features, margin gradient features, and biological properties, and aims to quantify aspects of pathological diagnosis, molecular expression, and disease prognosis[\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Previous studies have shown that machine learning can accurately recognize information in H\u0026amp;E images that is often not captured by pathologists through the human eye[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] .For example,Shamai et al.[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] successfully predicted the expression content of various molecular biomarkers, such as estrogen receptor, in breast cancer H\u0026amp;E sections with high accuracy using a machine learning approach. The pathomics model constructed by Zhaoyong Yan et al.[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] successfully predicted TNF receptor superfamily member 4 (TNFRSF4) expression, and the model had good predictive performance for molecular expression and revealed a significant association between high pathohistological scores and poorer OS.This study demonstrates that molecular prediction in H\u0026amp;E sections can be achieved using machine learning, overcoming the limitations of traditional methods. Hence,we aim to use machine learning to predict the expression level of PGF in H\u0026amp;E sections of HCC patients, enabling us to evaluate their prognosis. This approach will serve as a complementary tool for prognostic evaluation in HCC patients. Additionally, we will integrate bioinformatics analysis to explore the potential molecular mechanisms underlying Pathomics.\u003c/p\u003e"},{"header":"2. Methodology","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1. Patient cohort\u003c/h2\u003e\u003cp\u003eThe flow of this study is illustrated in (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The public database cohort used in this study was sourced from the publicly available TCGA dataset, and all data were de-identified and approved by the ethics committee for public research purposes. Conversely, the hospital cohort was obtained from the General Hospital of Ningxia Medical University and received approval from its Research Ethics Committee(KYLL-2025-0520).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec4\" class=\"Section3\"\u003e\u003ch2\u003e2.1.1. Public database queue\u003c/h2\u003e\u003cp\u003eFirst, 377 patients with HCC were selected from the TCGA database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://portal.gdc.cancer.gov/\u003c/span\u003e\u003cspan address=\"https://portal.gdc.cancer.gov/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The inclusion criteria were (a) patients with available H\u0026amp;E-stained section images from surgical resections. (b) Samples with primary solid tumor RNA-seq.(c) samples with a primary diagnosis and initial treatment of hepatocellular carcinoma. The exclusion criteria were(a) samples lacking survival status, survival time, or having a survival time of less than one month. (b) samples with missing clinical data,or substandard quality of pathological images. Ultimately, this study included a sample of 267 HCC cases in the TCGA cohort.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section3\"\u003e\u003ch2\u003e2.1.2. Hospital cohort\u003c/h2\u003e\u003cp\u003eIn addition, we included 80 HCC patients who underwent surgical treatment at the General Hospital of Ningxia Medical University from January 2016 to December 2018. Samples with hepatocellular carcinoma at both initial diagnosis and initial treatment were included, and samples with missing pathologic images or substandard pathologic image quality were excluded, resulting in 79 patients being included in the external validation cohort.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.2. survival analysis\u003c/h2\u003e\u003cp\u003eWe downloaded the HCC-related RNAseq data from the TCGA website, and analyzed the intergroup differences between the HCC tumor group and the normal group by Toil pipeline[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], and the model outputs the probability of predicting the gene expression level of Pathomics score after the modeling was completed, and compared the Pathomics score in the PGF high and low by Wilcoxon test. groupings with or without differences between the subgroups. Subsequently, by plotting the Kaplan-Meier survival curves, we demonstrated the survival differences between the high and low PGF expression groups as well as the survival differences between the high PS and low PS groups. In addition, we investigated whether high PGF expression as well as high PS expression were independent influences on overall survival (OS) using univariate and multivariate Cox regression analysis. Finally, we performed exploratory subgroup analyses by univariate COX regression to analyze the effects of PGF (high expression group vs. low expression group) and PS (high expression group vs. low expression group) on patients' prognosis in different subgroups of each covariate; the likelihood ratio test was used to analyze the interactions between PGF as well as PS and other covariates.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.3. Segmentation and processing of pathology images\u003c/h2\u003e\u003cp\u003eFirstly, image segmentation and processing were applied to the pathology images of the TCGA cohort. Given that the entire pathology image contains a significant amount of extraneous information, we used the OTSU algorithm (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://opencv.org/\u003c/span\u003e\u003cspan address=\"https://opencv.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) to divide the image into two parts: the irrelevant background and the relevant tissue region for the study[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. After removing the irrelevant background, the 40\u0026times; image was segmented into multiple subimages of 1024\u0026times;1024 pixels (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA), and the 20\u0026times; image was segmented into multiple subimages of 512\u0026times;512 pixels, which were then up-sampled to 1024\u0026times;1024 pixels. A final review was conducted by a pathologist to exclude subimages with poor image quality, such as contamination, blurriness, or blank areas exceeding 50%. Subsequently, 10 subimages from each segmented pathology image were randomly selected for further analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB) to reduce computational cost while ensuring sufficient data[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. The pathology images from the external validation queue were photographed and uploaded by specialized equipment and stored as 40\u0026times;magnification images for subsequent model validation.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.4. Image feature extraction\u003c/h2\u003e\u003cp\u003eWe utilized PyRadiomics, a Python-based tool (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pyradiomics.readthedocs.io/en/latest/\u003c/span\u003e\u003cspan address=\"https://pyradiomics.readthedocs.io/en/latest/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), to perform image normalization and extract relevant features, including first-order, second-order, and higher-order features, from the subimages. After extracting features from each of the 10 subimages per patient, the corresponding mean was calculated and used as the pathomics eigenvalue of each sample for subsequent data analysis.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.5. Feature screening and model construction\u003c/h2\u003e\u003cp\u003eThe TCGA cohort was randomly divided into training and validation sets in a 7:3 ratio. We employed the Maximum Relevance Minimum Redundancy (mRMR) algorithm and the Recursive Feature Elimination (RFE) algorithm to identify the optimal subset of features. Initially, the mRMR algorithm was used to filter out a set of relevant features, followed by the RFE algorithm to further refine the selection until the best subset of features was obtained. The RFE algorithm continued to identify relevant features vital for generating accurate models. Finally, we utilized the Gradient Boosting Machine (GBM) to assign importance scores to the selected pathomics features, thereby constructing a model capable of predicting the expression of PGF.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e2.6. Evaluation and validation of the model\u003c/h2\u003e\u003cp\u003eFor model evaluation, we assessed the predictive performance of the model by plotting ROC curves (Receiver Operating Characteristic curves) and calculating the respective AUC values (Area Under the ROC Curve, AUC) for the training, validation, and test sets. The calibration of the predictive models was assessed by plotting calibration curves and conducting Hosmer-Lemeshow goodness-of-fit tests. Decision curve analysis (DCA) was also performed to illustrate the clinical benefits of the predictive models. For model validation, we incorporated all pathology image data into the constructed pathomics model, calculated the Pathomics Score for each sample, and evaluated differences in PS between the training set, validation set, and the PGF high and low subgroups in the test set using the Wilcoxon test. These results were then visualized. Subsequently, the PS was integrated with clinical data and categorized into Low/High dichotomous variables PS after determining cutoff values using the survminer package.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e2.7. Pathohistological mechanism analysis\u003c/h2\u003e\u003cp\u003eTo explore the molecular mechanisms underlying Pathomics, we conducted Gene Set Enrichment Analysis (GSEA) using the KEGG and Hallmark gene sets to evaluate the enrichment of differentially expressed genes in high/low PS groups. Subsequently, we analyzed the differential expression of 200 inflammation-related genes between the high and low PS subgroups using the Wilcoxon rank-sum test, focusing on genes with a p-value of less than 0.05 for visualization[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. We uploaded the gene expression matrices of the HCC samples from the TCGA cohort to the ImmuCellAI database to calculate immune cell infiltration for each sample. We then analyzed the differences in immune cell infiltration in high/low PS groups using the Wilcoxon rank-sum test. Finally, we downloaded mutation data for TCGA-LIHC patients from the TCGA database and intersected it with 267 samples from the TCGA cohort to analyze the mutation data. We visualized the top 15 genes with the highest mutation frequency, assessing the relationship between the mutation profiles of TCGA-LIHC patients and the PS grouping model predictions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e2.8. Statistical Methods\u003c/h2\u003e\u003cp\u003eContinuous variables were compared between groups using the Wilcoxon test. The calibration of the pathohistological prediction model was assessed by plotting ROC and calibration curves and performing the Hosmer-Lemeshow goodness-of-fit test. Kaplan-Meier survival curves illustrated variations in survival rates across different groups, and the log-rank test evaluated the significance of survival rate differences between groups. Model performance was assessed through likelihood ratio test analysis. All analyses were conducted using R software version 4.1.0, with hypothesis testing being two-sided and a p-value of less than 0.05 indicating a significant difference.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.1. Correlation analysis of PGF with HCC\u003c/h2\u003e\u003cp\u003eIn both the TCGA cohort and the external validation cohort, we categorized patients into high- and low-expression groups based on the level of PGF expression and created baseline information tables for each cohort (\u003cspan refid=\"Sec21\" class=\"InternalRef\"\u003eAppendix Table\u0026nbsp;1\u003c/span\u003e、\u003cspan refid=\"Sec22\" class=\"InternalRef\"\u003eAppendix Table\u0026nbsp;2\u003c/span\u003e). The results showed no statistically significant differences in the distribution of clinical variables between the groups and were comparable.Analysis of variance showed that the PGF expression level in HCC tumor tissues was significantly higher than that in normal tissues (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). Kaplan-Meier curves showed that there was a significant difference in overall survival between patients with high and low PGF expression in the TCGA cohort (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001),the median survival time of patients in the low PGF expression group being 84.4 months, compared to only 46.2 months for those in the high PGF expression group(Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). In univariate analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC), high PGF expression emerged as a risk factor for overall survival (OS) in patients (HR\u0026thinsp;=\u0026thinsp;2.123, 95% CI: 1.368\u0026ndash;3.293, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). When included in multivariate analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD), high PGF expression remained a distinct risk factor for OS (HR\u0026thinsp;=\u0026thinsp;1.922, 95% CI: 1.217\u0026ndash;3.036, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005). In the interaction test analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE), there was an interaction between the expression content of PGF and the prognosis of patients in the three subgroups of whether or not they were treated with medication, whether or not they had undergone ablative embolization, and the level of blood AFP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The results of the subgroup analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF) showed that in subgroups undergoing drug therapy, ablative embolization therapy, and with AFP levels\u0026thinsp;\u0026gt;\u0026thinsp;400, elevated PGF did not correlate with patient prognosis. However, in subgroups not defined by these conditions, PGF expression levels could accurately reflect patient prognosis. While drug therapy and ablation embolization are commonly used clinical treatments that can affect patient survival time to varying degrees[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e],they also influence PGF expression levels, thereby complicating the assessment of HCC patient prognosis based on PGF expression. In the subgroup with AFP\u0026thinsp;\u0026gt;\u0026thinsp;400, PGF could not effectively predict the prognosis of HCC patients. This may be due to AFP's influence on PGF expression levels, though no detailed reports exist on the relationship between these factors in HCC patients domestically or internationally. This observation presents a novel perspective for exploring the impact of molecular levels on the survival prognosis of HCC patients.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.2. Pathology image feature extraction and model construction\u003c/h2\u003e\u003cp\u003eA total of 267*10 sub-images of 1024\u0026times;1024 pixels were obtained after segmenting and filtering the pathology images of the TCGA cohort. After image normalization, 1,488 features were extracted and finally the average value was taken to represent the histopathological features of this sample.[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. The TCGA cohort was randomly divided into training and validation sets with a 7:3 ratio, comprising 188 cases in the training set and 79 cases in the validation set. A baseline profile table was created, indicating that the analysis of variance for each variable showed no significant differences between the groups (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05) (\u003cspan refid=\"Sec23\" class=\"InternalRef\"\u003eAppendix Table\u0026nbsp;3\u003c/span\u003e). Subsequently, in the training set, we used the mRMR method to screen the top 30 image features with strong predictive efficacy. These features were further refined using the RFE method, resulting in a total of 7 features significantly correlated with the prediction of the PGF molecule for subsequent analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA).Finally, we calculated the importance scores of these 7 features in the model by the GBM algorithm (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB), and constructed the Pathomics model in the training set by the relevant features and their corresponding weights.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e3.3. Evaluation of the model\u003c/h2\u003e\u003cp\u003eThe results of our model efficacy evaluation showed that the HCC Pathomics model features had high predictive accuracy in both the training set (AUC\u0026thinsp;=\u0026thinsp;0.881) and the validation set (AUC\u0026thinsp;=\u0026thinsp;0.747) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA,B). In addition, in the external test cohort (AUC\u0026thinsp;=\u0026thinsp;0.74), the model had equally good predictive efficacy (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC).We then plotted calibration curves for the training, validation, and test datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD, E, and F), and the calibration curves and the Hosmer-Lemeshow goodness-of-fit test showed that the pathohistological prediction model's predictive probability of whether PGF was highly expressed was in good agreement with the true value (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05); the DCA curves (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eG, H, and I) demonstrated that the model had clinical utility.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e3.4.Prognostic value of the model\u003c/h2\u003e\u003cp\u003eWe divided the training, validation, and test sets into high expression and low expression groups based on PGF expression levels. By comparing the distribution of PS values across these three datasets, we observed that the PS of the PGF high expression group were significantly higher than those of the PGF low expression group (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA, B, C). The TCGA cohort was divided into high and low PS groups based on their PS. A baseline information table for each clinical variable was created (\u003cspan refid=\"Sec24\" class=\"InternalRef\"\u003eAppendix Table\u0026nbsp;4\u003c/span\u003e). The KM curves plotted (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eD) showed that the median survival time for the high PS group was 56.17 months, while it was 82.87 months for the low PS group. High PS was significantly associated with lower OS in HCC patients (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.006), highlighting the positive predictive value of predictive models in assessing patient prognosis.Further univariate and multivariate Cox regression analyses revealed that high PS was an independent risk factor for overall survival (OS) (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eA, B).Interaction test analysis showed (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eC, D) that there was no significant interaction between PS and the majority of subgroups, indicating that PS scores are reliable predictors of patient prognosis across different subgroups for each covariate.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e3.5. biological interpretability\u003c/h2\u003e\u003cp\u003eIn the KEGG gene set ( Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eA), signaling pathways such as RIBOSOME and COAGUATION demonstrated significant differences between the high/low PS groups. In the Hallmark gene set ( Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eB), the COAGUATION signaling pathway likewise demonstrated obvious differences.. In the differential analysis of inflammation-related genes between the high and low PS subgroups, we found (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eC) that the expression of genes such as \u003cem\u003eHBEGF\u003c/em\u003e was significantly elevated in the high PS group. In the immune cell infiltration analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eD), γδT cells were significantly different in the high and low PS subgroups (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01). Finally, Differential analysis of somatic gene mutations in the high/low PS groups showed that the mutation rate of the \u003cem\u003eTP53\u003c/em\u003e (29% vs. 25%) gene differed from that of the low PS group in the high PS group, and the mutation rate was higher than 20% in both the high and low PS groups(Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eE, F).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eWe developed a pathohistological model based on PGF expression levels for predicting the prognosis of HCC patients by PS, and validated its clinical effects as well as explored the biological interpretability behind it by an external test cohort. The main findings of this study were 1) PGF expression level had an independent prognostic value in HCC (HR\u0026thinsp;=\u0026thinsp;1.922, 95% CI: 1.217\u0026ndash;3.036, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005);2) a HE pathohistochemistry model consisting of seven pathological features was constructed to achieve non-invasive prediction of PGF (AUC\u0026thinsp;=\u0026thinsp;0.811) and was externally test-cohort validation (AUC\u0026thinsp;=\u0026thinsp;0.740); 3) the biological interpretability of the pathohistological model revealed that PS mapping PGF expression levels was associated with signaling pathways such as RIBOSOME and COAGUATION, inflammation-related genes such as \u003cem\u003eHBEGF\u003c/em\u003e, immune cells such as γδT cells, as well as mutations in \u003cem\u003eTP53\u003c/em\u003e genes.\u003c/p\u003e\u003cp\u003ePGF is a member of the pro-angiogenic VEGF family and is directly or indirectly involved in promoting immune escape and tumor metastasis, playing a significant role in tumor development[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Our study found a strong association between PGF and poor prognosis in HCC, indicating that the PGF molecule is an independent risk factor for the survival time of HCC patients. Furthermore, upregulation of PGF has been observed in cirrhotic patients[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. And there is increasing evidence that overexpression of PGF is associated with early recurrence of HCC, and PGF may be an important prognostic indicator for HCC[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].Additionally, the research by Miao et al.[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] demonstrated a significant difference in overall survival (OS) between high and low levels of PGF expression in HCC patients. PGF is not only valuable in hepatocellular carcinoma but also in other tumors, such as colorectal cancer, renal cell carcinoma, and pancreatic cancer, as it can reflect the prognosis of patients to a certain extent[\u003cspan additionalcitationids=\"CR31\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e] .\u003c/p\u003e\u003cp\u003eThe development of pathomics models represents a major leap forward in the integration of histopathology and molecular biology.Currently, with the rapid development of the field of Pathomics, its applications have expanded from the histological level, including tissue composition and cell types, to molecular pathology and disease prognosis. By utilizing various data mining algorithms based on machine learning methods, high-dimensional data can be transformed into predictive or prognostic information[\u003cspan additionalcitationids=\"CR34\" citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Recent seminal findings have shown the wide application of machine learning methods in Pathomics, demonstrating high value in metastasis detection in lymph nodes, \u003cem\u003eKi67\u003c/em\u003e scoring in breast cancer, Gleason grading in prostate cancer, and tumor-infiltrating lymphocyte scoring in melanoma[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. We have successfully constructed a H\u0026amp;E pathomics model from cases in the TCGA database using a machine learning approach to achieve non-invasive prediction of PGF expression levels. Most importantly, in an external validation cohort consisting of cases from our hospital, the model still demonstrated excellent prediction accuracy, even though the two cohorts were extremely different in terms of ethnicity, region, and so on. This demonstrates the stability and generalizability of H\u0026amp;E pathology images for molecular prediction, which can help us realize the possibility of molecular expression prediction from a pathology perspective. This non-invasive prediction may reach a newer height with the advancement of image analysis and processing tools. Meanwhile, models constructed from multicenter and large sample sizes will demonstrate higher predictive accuracy, thus providing a more accurate assessment of patient prognosis. Therefore, in this study, histopathological image characterization may be a more convenient, low-cost, and accurate alternative strategy for predicting the molecular expression levels of PGF in HCC patients.\u003c/p\u003e\u003cp\u003eOne of the most compelling aspects of our study is the biological interpretability of the pathomics model, we performed differential gene enrichment analysis of inflammation-related genes, abundance of immune cell infiltration, and gene mutation profiles in the high pathology score group and low pathology score group, respectively. Our gene set enrichment analysis (GSEA) revealed significant differences in the coagulation and ribosome pathways between high and low PGF expression groups., which is worthy of consideration.Coagulation factors can enhance metastasis by increasing the adhesion of cancer cells and enhancing the activation of endothelial cells[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e], while the coagulation cascade response plays a crucial role in shaping the tumor-immune microenvironment (TME)[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e] .And PGF as a key gene related to coagulation and fibrinolysis in HCC can affect the occurrence and development of HCC through TME[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. The ribosome pathway, on the other hand, is closely associated with protein synthesis and cell proliferation. Upregulation of PGF activates the mTOR pathway, which is involved in the regulation of multiple processes involved in ribosome biosynthesis and can control protein synthesis by promoting ribosome biosynthesis [\u003cspan additionalcitationids=\"CR41\" citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e].Hyperactivation of ribosome biogenesis leads to hepatocellular transformation and plays a key role in the development of HCC[\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e].Inflammation is an important hallmark of cancer and exhibits an important role in both tumorigenesis and progression[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], In HCC, the relationship between tumors and the inflammatory response is even closer, as the outcome of hepatitis leading to cirrhosis and the eventual development of HCC is well known.In the differential analysis of inflammation-related genes, we found that a considerable number of inflammatory genes were expressed at higher levels in the high PS group compared with the low PS group, especially \u003cem\u003eHBEGF\u003c/em\u003e was more significantly expressed. Increased local blood supply is one of the characteristics of HCC development. In a basic study of peripheral arterial disease (PAD), researchers caused an increase in local blood supply in the hind limbs of mice by inhibiting the relevant genes, and the expression of \u003cem\u003eHBEGF\u003c/em\u003e and PGF was up-regulated, suggesting that \u003cem\u003eHBEGF\u003c/em\u003e may cause alterations in local blood supply through angiogenesis or microenvironmental remodeling[\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. According to previous studies and the results of this paper, \u003cem\u003eHBEGF\u003c/em\u003e and PGF may play a common role in the development of tumor lesions or some benign vascular diseases. In our analysis of immune cell infiltration in hepatocellular carcinoma we found that γδT cells infiltrated to a higher extent in the PS low expression group than in the high expression group. γδT cells are a subset of T cells that play a role in anti-tumor immunity, but their function is often impaired in HCC patients. He et al.[\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e] pointed out that γδT cells were involved in mediating anti-tumor responses and that their infiltration effects correlate with HCC progression and patient prognosis, whereas in the development of HCC, γδT cells demonstrate a decreasing degree of infiltration and diminished anti-tumor effects. In HCC patients, PGF regulates γδT cell activation and function through the MAPK-RAP1A signaling pathway, thereby promoting immune evasion and tumor progression [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. This highlights the potential of targeting PGF and its downstream signaling pathways as a strategy for enhancing anti-tumor immunity in HCC. In addition, \u003cem\u003eTP53\u003c/em\u003e mutations are the most common type of genetic variation in HCC, with the largest mutation type being missense mutations, which are involved in the regulation of DNA repair, cell cycle arrest and apoptosis[\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. The high mutation rate of \u003cem\u003eTP53\u003c/em\u003e in our study (29% in the high PGF group and 25% in the low PGF group) underscores the importance of this gene in HCC pathogenesis and highlights the need for targeted therapies that can restore \u003cem\u003eTP53\u003c/em\u003e function or bypass its tumor-suppressive effects.\u003c/p\u003e\u003cp\u003eOur findings have far-reaching clinical implications. The ability to noninvasively predict PGF expression levels using routine H\u0026amp;E staining images opens up new possibilities for personalized medicine in HCC. By combining pathomics with other holographic data, such as genomics and proteomics, clinicians can gain a more comprehensive understanding of tumor biology and develop treatment strategies accordingly. For example, patients with high PGF expression may benefit from therapies targeting angiogenic or coagulation pathways, while patients with \u003cem\u003eTP53\u003c/em\u003e mutations may require alternative approaches such as immunotherapy or targeted gene therapy. In addition, the identification of inflammation-related genes and patterns of immune cell infiltration associated with PGF expression provides new avenues for immunotherapy development. By targeting the molecular pathways and immune mechanisms affected by PGF, it may be possible to increase the efficacy of existing therapies and improve patient prognosis. However, several challenges remain, such as heterogeneity between different patient cohorts and possible confounding factors in retrospective studies, which may limit the generalizability of our model. Future studies should aim to validate our findings in larger multicenter cohorts and explore the effects of other clinical variables, such as comorbidities and treatment regimens, on PGF expression and HCC prognosis. Furthermore, although our pathomics model showed high predictive accuracy, there is still room for improvement. The relatively low AUC in the external validation cohort (0.740 compared to 0.811 in the training cohort) suggests the need to further optimize the model parameters and incorporate more features to improve its stability and predictive power. Advances in image analysis and machine learning algorithms, coupled with the integration of multi-omics data, are expected to further refine the model and expand its application.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eIn conclusion, our study highlights the prognostic value of PGF expression in HCC and demonstrates the potential of pathomics as a non-invasive tool for predicting molecular expression and assessing patient prognosis. By integrating machine learning with bioinformatics analysis, we have uncovered novel insights into the molecular mechanisms underlying HCC progression, paving the way for more personalized and effective treatment strategies. As the field of pathomics continues to evolve, it is poised to play an increasingly important role in the era of precision medicine, offering new hope for patients with HCC and other malignancies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that no funds, grants, or other support were received during the preparation of this manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eLong Chen and Xusheng Zhang contributed equally. Long Chen, Xusheng Zhang, and Bendong Chen participated in the conceptualization and design of this study. Long Chen and WeiHu Ma organized the database and statistical analysis. Long Chen, Bendong Chen, Lin Ding,Shicai Liang,Xuebo Wang, and KeJun Liu divided the work and participated in the drawing. Long Chen wrote the first manuscript. Bendong Chen and KeJun Liu participated in the revision of the manuscript. All authors read and agreed to the final manuscript and authorship arrangement.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable to this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability and Ethical Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe dataset supporting the conclusions of this article is available in the [TCGA] repository, unique persistent identifier and hyperlink to dataset in https://portal.gdc.cancer.gov/projects/TCGA-LIHC.The external test cohort was sourced from the General Hospital of Ningxia Medical University and approved by the Research Ethics Committee of the General Hospital of Ningxia Medical University (Ethics Approval Number: KYLL-2025-0520), adhering to the ethical standards outlined in the Declaration of Helsinki (2013 revision).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInformed Consent Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEach participant in the study gave their consent.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflicts of Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe declare that there is no conflict of interest regarding the publication of this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study is especially grateful for the help provided by all the doctors from the Department of Pathology, General Hospital of Ningxia Medical University.\u003c/p\u003e\u003cp\u003eDisclaimer: Since the four appendices in the article take up a lot of space, I would like to present them as supplementary files. If your journal would like to include the appendices in the main text, I can assist with the re-formatting.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBrown, Z.J., et al., Management of Hepatocellular Carcinoma: A Review. JAMA Surg, 2023. 158(4): p. 410-420.\u003c/li\u003e\n\u003cli\u003eYang, J.D., et al., A global view of hepatocellular carcinoma: trends, risk, prevention and management. Nat Rev Gastroenterol Hepatol, 2019. 16(10): p. 589-604.\u003c/li\u003e\n\u003cli\u003eByrd, K., et al., Role of Multidisciplinary Care in the Management of Hepatocellular Carcinoma. Semin Liver Dis, 2021. 41(1): p. 1-8.\u003c/li\u003e\n\u003cli\u003eLlovet, J.M., et al., Hepatocellular carcinoma. Nat Rev Dis Primers, 2021. 7(1): p. 6.\u003c/li\u003e\n\u003cli\u003eFischer, C., et al., Anti-PlGF inhibits growth of VEGF(R)-inhibitor-resistant tumors without affecting healthy vessels. Cell, 2007. 131(3): p. 463-75.\u003c/li\u003e\n\u003cli\u003eXu, H.X., et al., Expression and prognostic significance of placental growth factor in hepatocellular carcinoma and peritumoral liver tissue. Int J Cancer, 2011. 128(7): p. 1559-69.\u003c/li\u003e\n\u003cli\u003eMiao, Y.D., et al., Prognostic role of expression of angiogenesis markers in hepatocellular carcinoma: A bioinformatics analysis. World J Gastroenterol, 2022. 28(30): p. 4221-4226.\u003c/li\u003e\n\u003cli\u003eAlbonici, L., et al., Multifaceted Role of the Placental Growth Factor (PlGF) in the Antitumor Immune Response and Cancer Progression. Int J Mol Sci, 2019. 20(12).\u003c/li\u003e\n\u003cli\u003eSimbrunner, B., et al., Placental growth factor levels neither reflect severity of portal hypertension nor portal-hypertensive gastropathy in patients with advanced chronic liver disease. Dig Liver Dis, 2021. 53(3): p. 345-352.\u003c/li\u003e\n\u003cli\u003eShen, B., et al., Identification and Analysis of Immune-Related Gene Signature in Hepatocellular Carcinoma. Genes (Basel), 2022. 13(10).\u003c/li\u003e\n\u003cli\u003eLiu, K. and J. Hu, Classification of acute myeloid leukemia M1 and M2 subtypes using machine learning. Comput Biol Med, 2022. 147: p. 105741.\u003c/li\u003e\n\u003cli\u003eNishio, M., et al., Homology-Based Image Processing for Automatic Classification of Histopathological Images of Lung Tissue. Cancers (Basel), 2021. 13(6).\u003c/li\u003e\n\u003cli\u003eBanna, G.L., et al., The Promise of Digital Biopsy for the Prediction of Tumor Molecular Features and Clinical Outcomes Associated With Immunotherapy. Front Med (Lausanne), 2019. 6: p. 172.\u003c/li\u003e\n\u003cli\u003eBeck, A.H., et al., Systematic analysis of breast cancer morphology uncovers stromal features associated with survival. Sci Transl Med, 2011. 3(108): p. 108ra113.\u003c/li\u003e\n\u003cli\u003eYu, K.H., et al., Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features. Nat Commun, 2016. 7: p. 12474.\u003c/li\u003e\n\u003cli\u003eShamai, G., et al., Artificial Intelligence Algorithms to Assess Hormonal Status From Tissue Microarrays in Patients With Breast Cancer. JAMA Netw Open, 2019. 2(7): p. e197700.\u003c/li\u003e\n\u003cli\u003eYan, Z., et al., Prognostic significance of TNFRSF4 expression and development of a pathomics model to predict expression in hepatocellular carcinoma. Heliyon, 2024. 10(11): p. e31882.\u003c/li\u003e\n\u003cli\u003eVivian, J., et al., Toil enables reproducible, open source, big biomedical data analyses. Nat Biotechnol, 2017. 35(4): p. 314-316.\u003c/li\u003e\n\u003cli\u003eWang, X., et al., Weakly Supervised Deep Learning for Whole Slide Lung Cancer Image Analysis. IEEE Trans Cybern, 2020. 50(9): p. 3950-3962.\u003c/li\u003e\n\u003cli\u003eChen, L., et al., Histopathological image and gene expression pattern analysis for predicting molecular features and prognosis of head and neck squamous cell carcinoma. Cancer Med, 2021. 10(13): p. 4615-4628.\u003c/li\u003e\n\u003cli\u003eZeng, H., et al., Integration of histopathological images and multi-dimensional omics analyses predicts molecular features and prognosis in high-grade serous ovarian cancer. Gynecol Oncol, 2021. 163(1): p. 171-180.\u003c/li\u003e\n\u003cli\u003eLiang, Y., et al., Identification and Validation of a Novel Inflammatory Response-Related Gene Signature for the Prognosis of Colon Cancer. J Inflamm Res, 2021. 14: p. 3809-3821.\u003c/li\u003e\n\u003cli\u003eZhai, W.Y., et al., A Novel Inflammatory-Related Gene Signature Based Model for Risk Stratification and Prognosis Prediction in Lung Adenocarcinoma. Front Genet, 2021. 12: p. 798131.\u003c/li\u003e\n\u003cli\u003eBaek, M.Y., et al., Clinical outcomes of patients with a single hepatocellular carcinoma less than 5 cm treated with transarterial chemoembolization. Korean J Intern Med, 2019. 34(6): p. 1223-1232.\u003c/li\u003e\n\u003cli\u003eSaednia, K., et al., Quantitative digital histopathology and machine learning to predict pathological complete response to chemotherapy in breast cancer patients using pre-treatment tumor biopsies. Sci Rep, 2022. 12(1): p. 9690.\u003c/li\u003e\n\u003cli\u003eLi, H., et al., Integrative Analysis of Histopathological Images and Genomic Data in Colon Adenocarcinoma. Front Oncol, 2021. 11: p. 636451.\u003c/li\u003e\n\u003cli\u003eHuang, X.X., et al., Up-regulation of proproliferative genes and the ligand/receptor pair placental growth factor and vascular endothelial growth factor receptor 1 in hepatitis C cirrhosis. Liver Int, 2007. 27(7): p. 960-8.\u003c/li\u003e\n\u003cli\u003eHo, M.C., et al., Placenta growth factor not vascular endothelial growth factor A or C can predict the early recurrence after radical resection of hepatocellular carcinoma. Cancer Lett, 2007. 250(2): p. 237-49.\u003c/li\u003e\n\u003cli\u003eNagaoka, S., et al., The ratio of serum placenta growth factor to soluble vascular endothelial growth factor receptor-1 predicts the prognosis of hepatocellular carcinoma. Oncol Rep, 2010. 23(6): p. 1647-54.\u003c/li\u003e\n\u003cli\u003eShen, Y., et al., Role of stemness-related genes TIMP1, PGF, and SNAI1 in the prognosis of colorectal cancer through single-cell RNA-seq. Cancer Med, 2023. 12(10): p. 11611-11623.\u003c/li\u003e\n\u003cli\u003eBae, S.H., T. Hwang and M.R. Han, Unraveling the hypoxia modulating potential of VEGF family genes in pan-cancer. Genomics Inform, 2023. 21(4): p. e44.\u003c/li\u003e\n\u003cli\u003eGohrig, A., et al., Placental growth factor promotes neural invasion and predicts disease prognosis in resectable pancreatic cancer. J Exp Clin Cancer Res, 2024. 43(1): p. 153.\u003c/li\u003e\n\u003cli\u003eLiao, H., et al., Preoperative Radiomic Approach to Evaluate Tumor-Infiltrating CD8(+) T Cells in Hepatocellular Carcinoma Patients Using Contrast-Enhanced Computed Tomography. Ann Surg Oncol, 2019. 26(13): p. 4537-4547.\u003c/li\u003e\n\u003cli\u003eXu, X., et al., Radiomic analysis of contrast-enhanced CT predicts microvascular invasion and outcome in hepatocellular carcinoma. J Hepatol, 2019. 70(6): p. 1133-1144.\u003c/li\u003e\n\u003cli\u003eKim, S., et al., Radiomics on Gadoxetic Acid-Enhanced Magnetic Resonance Imaging for Prediction of Postoperative Early and Late Recurrence of Single Hepatocellular Carcinoma. Clin Cancer Res, 2019. 25(13): p. 3847-3855.\u003c/li\u003e\n\u003cli\u003eAcs, B., M. Rantalainen and J. Hartman, Artificial intelligence as the next step towards precision pathology. J Intern Med, 2020. 288(1): p. 62-81.\u003c/li\u003e\n\u003cli\u003eHill, C.N., et al., Deciphering the Role of the Coagulation Cascade and Autophagy in Cancer-Related Thrombosis and Metastasis. Front Oncol, 2020. 10: p. 605314.\u003c/li\u003e\n\u003cli\u003eHe, Q., J. Yang and Y. Jin, Immune infiltration and clinical significance analyses of the coagulation-related genes in hepatocellular carcinoma. Brief Bioinform, 2022. 23(4).\u003c/li\u003e\n\u003cli\u003eFan, M., et al., Establishment and verification of a prognostic model based on coagulation and fibrinolysis-related genes in hepatocellular carcinoma. Aging (Albany NY), 2024. 16(9): p. 7578-7595.\u003c/li\u003e\n\u003cli\u003eAbetov, D.A., et al., Formation of mammalian preribosomes proceeds from intermediate to composed state during ribosome maturation. J Biol Chem, 2019. 294(28): p. 10746-10757.\u003c/li\u003e\n\u003cli\u003eGentilella, A., S.C. Kozma and G. Thomas, A liaison between mTOR signaling, ribosome biogenesis and cancer. Biochim Biophys Acta, 2015. 1849(7): p. 812-20.\u003c/li\u003e\n\u003cli\u003eHe, J., et al., Ribosome biogenesis protein Urb1 acts downstream of mTOR complex 1 to modulate digestive organ development in zebrafish. J Genet Genomics, 2017. 44(12): p. 567-576.\u003c/li\u003e\n\u003cli\u003eYang, X.M., et al., Nucleolar HEAT Repeat Containing 1 Up-regulated by the Mechanistic Target of Rapamycin Complex 1 Signaling Promotes Hepatocellular Carcinoma Growth by Dominating Ribosome Biogenesis and Proteome Homeostasis. Gastroenterology, 2023. 165(3): p. 629-646.\u003c/li\u003e\n\u003cli\u003eKant, S., et al., Neural JNK3 regulates blood flow recovery after hindlimb ischemia in mice via an Egr1/Creb1 axis. Nat Commun, 2019. 10(1): p. 4223.\u003c/li\u003e\n\u003cli\u003eHe, W., et al., Hepatocellular carcinoma-infiltrating gammadelta T cells are functionally defected and allogenic Vdelta2(+) gammadelta T cell can be a promising complement. Clin Transl Med, 2022. 12(4): p. e800.\u003c/li\u003e\n\u003cli\u003eLi, H., et al., MAPK-RAP1A Signaling Enriched in Hepatocellular Carcinoma Is Associated With Favorable Tumor-Infiltrating Immune Cells and Clinical Prognosis. Front Oncol, 2021. 11: p. 649980.\u003c/li\u003e\n\u003cli\u003eZhang, X., Z. Fu and X. Zhang, TP53 Mutation Related and Directly Regulated lncRNA Prognosis Markers in Hepatocellular Carcinoma. Onco Targets Ther, 2021. 14: p. 4427-4437.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcan","sideBox":"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcan/default.aspx","title":"BMC Cancer","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Hepatocellular carcinoma, Placental growth factor, Pathomics, Machine learning, Mechanism analysis","lastPublishedDoi":"10.21203/rs.3.rs-6945731/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6945731/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePURPOSE \u003c/strong\u003ePlacental growth factor (PGF) is implicated in hepatocellular carcinoma (HCC) progression, although its functional role remains unclear. This study aimed to develop a pathomics model for predicting PGF expression from H\u0026amp;E-stained HCC sections and investigate its prognostic relevance and molecular mechanisms.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMETHODS \u003c/strong\u003eRetrospective analysis utilized H\u0026amp;E images and clinical data from TCGA and an external cohort. Prognostic significance of PGF was assessed via survival analysis. Image segmentation employed the OTSU algorithm, followed by PyRadiomics-based feature extraction. Key features were selected using mRMR and RFE algorithms, with a gradient boosting machine (GBM) model constructed for PGF prediction. Model performance was validated through ROC, calibration, and decision curve analyses. Prognostic stratification, Cox regression, and subgroup analyses were conducted for high/low pathomics score (PS) groups. Bioinformatics approaches identified differentially expressed genes (DEGs) and immune infiltration patterns.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRESULTS \u003c/strong\u003ePGF expression independently predicted poor HCC prognosis (HR=1.922, 95% CI:1.217-3.036,\u003cem\u003eP\u003c/em\u003e=0.005). The pathomics model incorporating seven PGF-associated features demonstrated robust predictive performance (AUC=0.811; external validation AUC=0.740). High-PS patients exhibited significantly worse survival (HR=1.667, 95% CI:1.024–2.713, \u003cem\u003eP\u003c/em\u003e=0.040). DEGs in high-PS subgroups showed enrichment in ribosome and coagulation pathways, accompanied by upregulated \u003cem\u003eHBEGF\u003c/em\u003e (inflammation-related) and increased γδT cell infiltration. \u003cem\u003eTP53\u003c/em\u003emutations were prevalent (\u0026gt;20% mutation rate) in high-PS cases.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCONCLUSION \u003c/strong\u003ePGF serves as an independent prognostic biomarker in HCC. The developed pathomics model enables non-invasive PGF expression prediction through H\u0026amp;E image analysis. Mechanistically, PGF-associated molecular alterations involve inflammatory signaling, immune microenvironment remodeling, and frequent \u003cem\u003eTP53\u003c/em\u003e mutations, providing insights into HCC pathogenesis.\u003c/p\u003e","manuscriptTitle":"Exploratory Study on the Prognostic Value and Biological Interpretability of PGF-Related H\u0026amp;E Pathomics Models in Hepatocellular Carcinoma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-07 04:10:17","doi":"10.21203/rs.3.rs-6945731/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"301967743793495727962523589044352992161","date":"2025-08-05T03:08:41+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-04T05:26:56+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-07-09T04:32:43+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-25T05:45:40+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-06-25T05:44:37+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Cancer","date":"2025-06-21T15:01:36+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcan","sideBox":"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcan/default.aspx","title":"BMC Cancer","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f11ada67-529a-4ace-880b-ac3f6ae8b2c9","owner":[],"postedDate":"August 7th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-08-07T04:10:18+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-07 04:10:17","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6945731","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6945731","identity":"rs-6945731","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}