MASAN: a novel staging system for prognosis of patients with oesophageal squamous cell carcinoma

Laboratory blood parameters and machine learning for the prognosis of esophageal squamous cell carcinoma

Background

In contemporary study, the death of esophageal squamous cell carcinoma (ESCC) patients need precise and expedient prognostic methodologies.

Objective

To develop and validate a prognostic model tailored to ESCC patients, leveraging the power of machine learning (ML) techniques and drawing insights from comprehensive datasets of laboratory-derived blood parameters.

Methods

Three ML approaches, including Gradient Boosting Machine (GBM), Random Survival Forest (RSF), and the classical Cox method, were employed to develop models on a dataset of 2521 ESCC patients with 27 features. The models were evaluated by concordance index (C-index) and time receiver operating characteristics (Time ROC) curves. We used the optimal model to evaluate the correlation between features and prognosis and divide patients into low- and high-risk groups by risk stratification. Its performance was analyzed by Kaplan-Meier curve and the comparison with AJCC8 stage. We further evaluate the comprehensive effectiveness of the model in ESCC subgroup by risk score and KDE (kernel density estimation) plotting.

Results

RSF's C-index (0.746) and AUC (three-year AUC 0.761, five-year AUC 0.771) had slight advantage over GBM and the classical Cox method. Subsequently, 14 features such as N stage, T stage, surgical margin, tumor length, age, Dissected LN number, MCH, Na, FIB, DBIL, CL, treatment, vascular invasion, and tumor grade were selected to build the model. Based on these, we found significant difference for survival rate between low-(3-year OS 81.8%, 5-year OS 69.8%) and high-risk (3-year OS 25.1%, 5-year OS 11.5%) patients in training set, which was also verified in test set (all P < 0.0001). Compared with the AJCC8th stage system, it showed a greater discriminative ability which is also in good agreement with its staging ability.

Conclusion

We developed an ESCC prognostic model with good performance by clinical features and laboratory blood parameters.

Spatial analysis of stromal signatures identifies invasive front carcinoma-associated fibroblasts as suppressors of anti-tumor immune response in esophageal cancer

BACKGROUND

Increasing evidence indicates that the tumor microenvironment (TME) is a crucial determinant of cancer progression. However, the clinical and pathobiological significance of stromal signatures in the TME, as a complex dynamic entity, is still unclear in esophageal squamous cell carcinoma (ESCC).

METHODS

Herein, we used single-cell transcriptome sequencing data, imaging mass cytometry (IMC) and multiplex immunofluorescence staining to characterize the stromal signatures in ESCC and evaluate their prognostic values in this aggressive disease. An automated quantitative pathology imaging system determined the locations of the lamina propria, stroma, and invasive front. Subsequently, IMC spatial analyses further uncovered spatial interaction and distribution. Additionally, bioinformatics analysis was performed to explore the TME remodeling mechanism in ESCC. To define a new molecular prognostic model, we calculated the risk score of each patient based on their TME signatures and pTNM stages.

RESULTS

We demonstrate that the presence of fibroblasts at the tumor invasive front was associated with the invasive depth and poor prognosis. Furthermore, the amount of α-smooth muscle actin (α-SMA)⁺ fibroblasts at the tumor invasive front positively correlated with the number of macrophages (MØs), but negatively correlated with that of tumor-infiltrating granzyme B⁺ immune cells, and CD4⁺ and CD8⁺ T cells. Spatial analyses uncovered a significant spatial interaction between α-SMA⁺ fibroblasts and CD163⁺ MØs in the TME, which resulted in spatially exclusive interactions to anti-tumor immune cells. We further validated the laminin and collagen signaling network contributions to TME remodeling. Moreover, compared with pTNM staging, a molecular prognostic model, based on expression of α-SMA⁺ fibroblasts at the invasive front, and CD163⁺ MØs, showed higher accuracy in predicting survival or recurrence in ESCC patients. Regression analysis confirmed this model is an independent predictor for survival, which also identifies a high-risk group of ESCC patients that can benefit from adjuvant therapy.

CONCLUSIONS

Our newly defined biomarker signature may serve as a complement for current clinical risk stratification approaches and provide potential therapeutic targets for reversing the fibroblast-mediated immunosuppressive microenvironment.

Development and Validation of a New Staging System for Esophageal Squamous Cell Carcinoma Patients Based on Combined Pathological TNM, Radiomics, and Proteomics

OBJECTIVE

This study aimed to construct a new staging system for patients with esophageal squamous cell carcinoma (ESCC) based on combined pathological TNM (pTNM) stage, radiomics, and proteomics.

METHODS

This study collected patients with radiomics and pTNM stage (Cohort 1, n = 786), among whom 103 patients also had proteomic data (Cohort 2, n = 103). The Cox regression model with the least absolute shrinkage and selection operator, and the Cox proportional hazards model were used to construct a nomogram and predictive models. Concordance index (C-index) and the integrated area under the time-dependent receiver operating characteristic (ROC) curve (IAUC) were used to evaluate the predictive models. The corresponding staging systems were further assessed using Kaplan-Meier survival curves.

RESULTS

For Cohort 1, the RadpTNM4c staging systems, constructed based on combined pTNM stage and radiomic features, outperformed the pTNM4c stage in both the training dataset 1 (Train1; IAUC 0.711 vs. 0.706, p < 0.001) and the validation dataset 1 (Valid1; IAUC 0.695 vs. 0.659, p < 0.001; C-index 0.703 vs. 0.674, p = 0.029). For Cohort 2, the ProtRadpTNM2c staging system, constructed based on combined pTNM stage, radiomics, and proteomics, outperformed the pTNM2c stage in both the Train2 (IAUC 0.777 vs. 0.610, p < 0.001; C-index 0.898 vs. 0.608, p < 0.001) and Valid2 (IAUC 0.746 vs. 0.608, p < 0.001; C-index 0.889 vs. 0.641, p = 0.009) datasets.

CONCLUSIONS

The ProtRadpTNM2c staging system, based on combined pTNM stage, radiomic, and proteomic features, improves the predictive performance of the classical pTNM staging system.

ASO Author Reflections: A Multiomic Novel Staging System for Esophageal Squamous Cell Carcinoma Patients

The past eighth edition of the American Joint Committee on Cancer (AJCC)/International Union against Cancer (UICC) pathologic tumor-node-metastasis (pTNM) staging system for esophageal squamous cell carcinoma (ESCC) patients, which also is the gold standard of postoperative treatment decision-making, needs to be continuously improved. To improve the efficiency of the staging system, the proteomic data from Chinese ESCC patients was combined with preoperative radiomic data and pTNM data to establish the multiomic RadpTNM and ProtRadpTNM models and compare them with the traditional pTNM staging system. The results suggest that both the RadpTNM and ProtRadpTNM models are significantly better than the traditional pTNM staging system. Future prospective multicentered cohort studies in Asian and Caucasian patients with ESCC are warranted to evaluate the efficiency of the multiomic models.

A machine learning method for improving liver cancer staging

Liver cancer is a common malignant tumor, and its clinical stage is closely related to the clinical treatment and prognosis of patients. Currently, the BCLC staging system revised by the BCLC group of University of Barcelona is the globally recognized staging system for liver cancer. However, with the deepening of related research, the current staging system can no longer fully meet the clinical needs. In this work, we propose a novel machine learning method for constructing an automatic hepatocellular carcinoma staging model that incorporates far more clinical variables than any existing staging system. Our model is based on random survival forests, which generates a unique hazard function for each patient. B-splines are used to embed hazard functions into vectors in low-dimensional space and hierarchical clustering method groups similar patients to form staging cohorts. The resulting staging system significantly outperforms the BCLC system in terms of distinctiveness between patients in different stages.

Evidence of SARS-CoV-2 infection in gallbladder and aggravating cholecystitis to septic shock: a case report

Coronavirus disease 2019 (COVID-19) has threatened human health worldwide and could lead to multiple organs injury. However, the impact on the virus infecting the biliary system, especially the gallbladder, has remained unclear and no pathological evidence has been reported yet. A case of SARS-CoV-2 infection in a gallbladder with cholecystitis, which progressed rapidly to sepsis and required an emergency operation was investigated and reported. Clinical specimens of the COVID-19 patient including serum, oropharyngeal swabs, sputum, bile, abdominal drainage fluid, urine, stool, and gallbladder tissue were collected and tested for SARS-CoV-2 RNA using a quantitative polymerase chain reaction (qPCR) assay. Fresh normal gallbladder tissue and gangrenous gallbladder tissue were also collected for further research including hematoxylin and eosin (HE), immunohistochemistry (IHC), and immunofluorescent (IF) staining, and compared with the gallbladder from the COVID-19 patient. The bile, as well as the serum, oropharyngeal swabs, sputum, abdominal drainage fluid, urine, and rectal swabs were consecutively negative for SARS-CoV-2 RNA. The viral host receptor angiotensin-converting enzyme 2 (ACE2) was highly expressed in gallbladder epithelial cells, and viral nucleocapsid protein (NP) was visualized in the cytoplasm of gallbladder epithelial cells. Immune cells including CD2, CD3, CD4, CD8, CD20, CD38, CD68, and MPO were positive in gangrenous gallbladder tissues without SARS-CoV-2 infection, and were relatively downregulated in SARS-CoV-2 infective gallbladder tissue. This study provided evidence of SARS-CoV-2 infection in the gallbladder and verified that the gallbladder was one of the target organs that SARS-CoV-2 could attack and damage using ACE2 as a cell receptor. Due to the immune dysregulation involved, more vigilant management and early assessment is needed for COVID-19 patients with the comorbidity of cholecystitis.

Integrated single-cell transcriptome analysis reveals heterogeneity of esophageal squamous cell carcinoma microenvironment

The tumor microenvironment is a highly complex ecosystem of diverse cell types, which shape cancer biology and impact the responsiveness to therapy. Here, we analyze the microenvironment of esophageal squamous cell carcinoma (ESCC) using single-cell transcriptome sequencing in 62,161 cells from blood, adjacent nonmalignant and matched tumor samples from 11 ESCC patients. We uncover heterogeneity in most cell types of the ESCC stroma, particularly in the fibroblast and immune cell compartments. We identify a tumor-specific subset of CST1⁺ myofibroblasts with prognostic values and potential biological significance. CST1⁺ myofibroblasts are also highly tumor-specific in other cancer types. Additionally, a subset of antigen-presenting fibroblasts is revealed and validated. Analyses of myeloid and T lymphoid lineages highlight the immunosuppressive nature of the ESCC microenvironment, and identify cancer-specific expression of immune checkpoint inhibitors. This work establishes a rich resource of stromal cell types of the ESCC microenvironment for further understanding of ESCC biology.

A Transcription Factor-Based Risk Model for Predicting the Prognosis of Prostate Cancer and Potential Therapeutic Drugs

Background

Prostate cancer (PC) is one of the most critical cancers affecting men's health worldwide. The development of many cancers involves dysregulation or mutations in key transcription factors. This study established a transcription factor-based risk model to predict the prognosis of PC and potential therapeutic drugs.

Materials and Methods

In this study, RNA-sequencing data were downloaded and analyzed using The Cancer Genome Atlas dataset. A total of 145 genes related to the overall survival rate of PC patients were screened using the univariate Cox analysis. The Kdmist clustering method was used to classify prostate adenocarcinoma (PRAD), thereby determining the cluster related to the transcription factors. The support vector machine-recursive feature elimination method was used to identify genes related to the types of transcription factors and the key genes specifically upregulated or downregulated were screened. These genes were further analyzed using Lasso to establish a model. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were used for the functional analysis. The TIMER algorithm was used to quantify the abundance of immune cells in PRAD samples. The chemotherapy response of each GBM patient was predicted based on the public pharmacogenomic database, Genomics of Drug Sensitivity in Cancer (GDSC, http://www.cancerrxgene.org). The R package "pRRophetic" was applied to drug sensitivity (IC50) value prediction.

Results

We screened 10 genes related to prognosis, including eight low-risk genes and two high-risk genes. The receiver operating characteristic (ROC) curve was 0.946. Patients in the high-risk score group had a poorer prognosis than those in the low-risk score group. The average area under the curve value of the model at different times was higher than 0.8. The risk score was an independent prognostic factor. Compared with the low-risk score group, early growth response-1 (EGR1), CACNA2D1, AC005831.1, SLC52A3, TMEM79, IL20RA, CRACR2A, and FAM189A2 expressions in the high-risk score group were decreased, while AC012181.1 and TRAPPC8 expressions were increased. GO and KEGG analyses showed that prognosis was related to various cancer signaling pathways. The proportion of B_cell, T_cell_CD4, and macrophages in the high-risk score group was significantly higher than that in the low-risk score group. A total of 25 classic immune checkpoint genes were screened out to express abnormally high-risk scores, and there were significant differences. Thirty mutant genes were identified; in the high- and low-risk score groups, SPOP, TP53, and TTN had the highest mutation frequency, and their mutations were mainly missense mutations. A total of 36 potential drug candidates for the treatment of PC were screened and identified.

Conclusions

Ten genes of both high-and low-risk scores were associated with the prognosis of PC. PC prognosis may be related to immune disorders. SPOP, TP53, and TTN may be potential targets for the prognosis of PC.

Large-scale and high-resolution mass spectrometry-based proteomics profiling defines molecular subtypes of esophageal cancer for therapeutic targeting

Esophageal cancer (EC) is a type of aggressive cancer without clinically relevant molecular subtypes, hindering the development of effective strategies for treatment. To define molecular subtypes of EC, we perform mass spectrometry-based proteomic and phosphoproteomics profiling of EC tumors and adjacent non-tumor tissues, revealing a catalog of proteins and phosphosites that are dysregulated in ECs. The EC cohort is stratified into two molecular subtypes-S1 and S2-based on proteomic analysis, with the S2 subtype characterized by the upregulation of spliceosomal and ribosomal proteins, and being more aggressive. Moreover, we identify a subtype signature composed of ELOA and SCAF4, and construct a subtype diagnostic and prognostic model. Potential drugs are predicted for treating patients of S2 subtype, and three candidate drugs are validated to inhibit EC. Taken together, our proteomic analysis define molecular subtypes of EC, thus providing a potential therapeutic outlook for improving disease outcomes in patients with EC.

Using a machine learning approach to identify key prognostic molecules for esophageal squamous cell carcinoma

Background

A plethora of prognostic biomarkers for esophageal squamous cell carcinoma (ESCC) that have hitherto been reported are challenged with low reproducibility due to high molecular heterogeneity of ESCC. The purpose of this study was to identify the optimal biomarkers for ESCC using machine learning algorithms.

Methods

Biomarkers related to clinical survival, recurrence or therapeutic response of patients with ESCC were determined through literature database searching. Forty-eight biomarkers linked to recurrence or prognosis of ESCC were used to construct a molecular interaction network based on NetBox and then to identify the functional modules. Publicably available mRNA transcriptome data of ESCC downloaded from Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) datasets included GSE53625 and TCGA-ESCC. Five machine learning algorithms, including logical regression (LR), support vector machine (SVM), artificial neural network (ANN), random forest (RF) and XGBoost, were used to develop classifiers for prognostic classification for feature selection. The area under ROC curve (AUC) was used to evaluate the performance of the prognostic classifiers. The importances of identified molecules were ranked by their occurrence frequencies in the prognostic classifiers. Kaplan-Meier survival analysis and log-rank test were performed to determine the statistical significance of overall survival.

Results

A total of 48 clinically proven molecules associated with ESCC progression were used to construct a molecular interaction network with 3 functional modules comprising 17 component molecules. The 131,071 prognostic classifiers using these 17 molecules were built for each machine learning algorithm. Using the occurrence frequencies in the prognostic classifiers with AUCs greater than the mean value of all 131,071 AUCs to rank importances of these 17 molecules, stratifin encoded by SFN was identified as the optimal prognostic biomarker for ESCC, whose performance was further validated in another 2 independent cohorts.

Conclusion

The occurrence frequencies across various feature selection approaches reflect the degree of clinical importance and stratifin is an optimal prognostic biomarker for ESCC.

CD103+ tumor-infiltrating lymphocytes predict favorable prognosis in patients with esophageal squamous cell carcinoma

As an indispensable factor in preventing the recirculation of tissue lymphocytes to the lymphatic and blood systems, the integrin CD103 has enabled the characterization of lymphocyte populations in non-lymphoid tissues and organs. However, the expression, distribution, and clinical significance of CD103⁺ tumor-infiltrating lymphocytes (TILs) in esophageal squamous cell carcinoma (ESCC) remain unclear. In the present study, we included tumor and adjacent non-tumor tissue specimens from 198 patients with ESCC who had undergone surgical resection. Immunohistochemistry and immunofluorescence were used to detect CD103⁺ TIL distribution, as well as the co-expression of CD103 and T cell markers and functional molecules. Kaplan-Meier analysis and the Cox proportional hazards model were used to estimate the prognostic value of CD103⁺ TILs. The results showed that CD103⁺ TILs were predominantly located in adjacent non-tumor tissues compared with tumor tissues (P < 0.0001). Immunofluorescence double staining revealed that CD8⁺ T cells, but not CD4⁺ T cells, comprised the majority of CD103-expressing cells. Most of these CD103-expressing cells co-expressed CTLA-4 and granzyme B rather than the exhaustion marker PD-1. High density of intratumoral CD103⁺ TIL is associated with longer overall survival (OS) and disease-free survival (DFS) in both the internal (OS, P = 0.0004 and DFS, P = 0.0002) and external (OS, P = 0.038 and DFS, P = 0.12) cohorts. Multivariate Cox analysis showed the density of CD103⁺ TILs was an independent positive prognostic factor for OS (hazards ratio [HR] = 0.406; P = 0.0003 in the internal cohort; HR = 0.328, P = 0.01, in the external cohort) and DFS (HR = 0.385; P = 0.0002 in the internal cohort; HR = 0.270, P = 0.003, in the external cohort). Our findings indicate that CD103⁺ TILs might play an important role in the tumor microenvironment, and intratumoral CD103⁺ TILs could serve as a promising prognostic marker in ESCC.

A Novel Three-Gene Model Predicts Prognosis and Therapeutic Sensitivity in Esophageal Squamous Cell Carcinoma

To precisely predict the clinical outcome and determine the optimal treatment options for patients with esophageal squamous cell carcinoma (ESCC) remains challenging. Prognostic models based on multiple molecular markers of tumors have been shown to have superiority over the use of single biomarkers. Our previous studies have identified the crucial role of ezrin in ESCC progression, which prompted us to hypothesize that ezrin-associated proteins contribute to the pathobiology of ESCC. Herein, we explored the clinical value of a molecular model constructed based on ezrin-associated proteins in ESCC patients. We revealed that the ezrin-associated proteins (MYC, PDIA3, and ITGA5B1) correlated with the overall survival (OS) and disease-free survival (DFS) of patients with ESCC. High expression of MYC was associated with advanced pTNM-stage (P=0.011), and PDIA3 and ITGA5B1 were correlated with both lymph node metastasis (PDIA3: P < 0.001; ITGA5B1: P=0.001) and pTNM-stage (PDIA3: P=0.001; ITGA5B1: P=0.009). Furthermore, we found that, compared with the current TNM staging system, the molecular model elicited from the expression of MYC, PDIA3, and ITGA5B1 shows higher accuracy in predicting OS (P < 0.001) or DFS (P < 0.001) in ESCC patients. Moreover, ROC and regression analysis demonstrated that this model was an independent predictor for OS and DFS, which could also help determine a subgroup of ESCC patients that may benefit from chemoradiotherapy. In conclusion, our study has identified a novel molecular prognosis model, which may serve as a complement for current clinical risk stratification approaches and provide potential therapeutic targets for ESCC treatment.

Integration of gene interaction information into a reweighted random survival forest approach for accurate survival prediction and survival biomarker discovery

Accurately predicting patient risk and identifying survival biomarkers are two important tasks in survival analysis. For the emerging high-throughput gene expression data, random survival forest (RSF) is attracting more and more attention as it not only shows excellent performance on survival prediction problems with high-dimensional variables, but also is capable of identifying important variables according to variable importance automatically calculated within the algorithm. However, RSF still suffers from some problems such as limited predictive accuracy on independent datasets and limited biological interpretation of survival biomarkers. In this study, we integrated gene interaction information into a Reweighted RSF model (RRSF) to improve predictive accuracy and identify biologically meaningful survival markers. We applied RRSF to the prediction of patients with glioblastoma multiforme (GBM) and esophageal squamous cell carcinoma (ESCC). With a reconstructed global pathway network and an mRNA-lncRNA co-expression network as the prior gene interaction information, RRSF showed better overall predictive performance than RSF on three GBM and two ESCC datasets. In addition, RRSF identified a two-gene and three-lncRNA signature, which showed robust prognostic values and had high biological relevance to the development of GBM and ESCC, respectively.