Background: Radical concurrent chemoradiotherapy (CCRT) is frequently used as the first-line treatment for patients with locally advanced esophageal cancer. Unfortunately, some patients respond poorly. To predict response to radical concurrent chemoradiotherapy in pre-treatment patients with esophageal squamous carcinoma (ESCC), and compare the predicting efficacies of radiomics features of primary tumor with or without regional lymph nodes, we developed a radiomics-clinical model based on the positioning CT images. Finally, SHapley Additive exPlanation (SHAP) was used to explain the models. Methods: This retrospective study enrolled 105 patients with medically inoperable and/or unresectable ESCC who underwent radical concurrent chemoradiotherapy (CCRT) between October 2018 and May 2023. Patients were classified into responder and non-responder groups with RECIST standards. The 11 recently admitted patients were chosen as the validation set, previously admitted patients were randomly split into the training set (n = 70) and the testing set (n = 24). Primary tumor site (GTV), the primary tumor and the uninvolved lymph nodes at risk of microscopic disease (CTV) were identified as Regions of Interests (ROIs). 1762 radiomics features from GTV and CTV were respectively extracted and then filtered by statistical differential analysis and Least Absolute Shrinkage and Selection Operator (LASSO). The filtered radiomics features combined with 13 clinical features were further filtered with Mutual Information (MI) algorithm. Based on the filtered features, we developed five models (Clinical Model, GTV Model, GTV-Clinical Model, CTV Model, and CTV-Clinical Model) using the random forest algorithm and evaluated for their accuracy, precision, recall, F1-Score and AUC. Finally, SHAP algorithm was adopted for model interpretation to achieve transparency and utilizability. Results: The GTV-Clinical model achieves an AUC of 0.82 with a 95% confidence interval (CI) of 0.76–0.99 on testing set and an AUC of 0.97 with a 95% confidence interval (CI) of 0.84–1.0 on validation set, which are significantly higher than those of other models in predicting ESCC response to CCRT. The SHAP force map provides an integrated view of the impact of each feature on individual patients, while the SHAP summary plots indicate that radiomics features have a greater influence on model prediction than clinical factors in our model. Conclusion: GTV-Clinical model based on texture features and the maximum diameter of lesion (MDL) may assist clinicians in pre-treatment predicting ESCC response to CCRT.

Background: Tumor microenvironment plays pivotal roles in carcinogenesis, cancer development and metastasis. Composition of cancer immune cell subsets can be inferred by deconvolution of gene expression profile accurately. Compositions of the cell types in cancer microenvironment including cancer infiltrating immune and stromal cells have been reported to be associated with the cancer outcomes markers for cancer prognosis. However, rare studies have been reported on their association with the response to preoperative radiotherapy for rectal cancer. Methods: In this paper, we deconvoluted the immune/stromal cell composition from the gene expression profiles. We compared the composition of immune/stromal cell types in the RT responsive versus nonresponsive for rectal cancer. We also compared the peripheral blood immune cell subset composition in the stable diseases versus progressive diseases of rectal cancer patients with fluorescence-activated cell sorting from our institution. Results: Compared with the non-responsive group, the responsive group showed higher proportions of CD4 ⁺ T cell (0.1378 ± 0.0368 vs. 0.1071 ± 0.0373, p = 0.0215), adipocytes, T cells CD4 memory resting, and lower proportions of CD8 ⁺ T cell (0.1798 ± 0.0217 vs. 0.2104 ± 0.0415, p = 0.0239), macrophages M2, and preadipocytes in their cancer tissue. The responsive patients showed a higher ratio of CD4 ⁺ /CD8 ⁺ T cell proportions (mean 0.7869 vs. 0.5564, p = 0.0210). Consistently, the peripheral blood dataset showed higher proportion of CD4 ⁺ T cells and higher ratio of CD4 ⁺ /CD8 ⁺ T cells, and lower proportion of CD8 ⁺ T cells for favorable prognosis. We validated these results with a pooled dataset of GSE3493 and GSE35452, and more peripheral blood data, respectively. Finally, we imported these eight cell features including eosinophils and macrophage M1 to Support Vector Machines and could predict the pre-radiotherapy responsive versus non-responsive with an accuracy of 76%, ROC AUC 0.77, 95% confidential interval of 0.632–0.857, better than the gene signatures. Conclusions: Our results showed that the proportions of tumor-infiltrating subsets and peripheral blood immune cell subsets can be important immune cell markers and treatment targets for outcomes of radiotherapy for rectal cancer.

Esophageal fistula (EF), a rare and potentially fatal complication, can be better managed with predictive models for personalized treatment plans in esophageal cancers. We aim to develop a clinical-deep learning radiomics model for effectively predicting the occurrence of EF. The study involved esophageal cancer patients undergoing radiotherapy or chemoradiotherapy. Arterial phase enhanced CT images were used to extract handcrafted and deep learning radiomic features. Along with clinical information, a 3-step feature selection method (statistical tests, Least Absolute Shrinkage and Selection Operator, and Recursive Feature Elimination) was used to identify five feature sets in training cohort for constructing random forest EF prediction models. Model performance was compared and validated in both retrospective and prospective test cohorts. One hundred seventy five patients (122 in training and 53 in test cohort)were retrospectively collected from April 2018 to June 2022. An additional 27 patients were enrolled as a prospective test cohort from June 2022 to December 2023. Post-selection in the training cohort, five feature sets were used for model construction: clinical, handcrafted radiomic, deep learning radiomic, clinical-handcrafted radiomic, and clinical-deep learning radiomic. The clinical-deep learning radiomic model excelled with AUC of 0.89 (95% Confidence Interval: 0.83–0.95) in the training cohort, 0.81 (0.65–0.94) in the test cohort, and 0.85 (0.71–0.97) in the prospective test cohort. Brier-score and calibration curve analyses validated its predictive ability. The clinical-deep learning radiomic model can effectively predict EF in patients with advanced esophageal cancer undergoing radiotherapy or chemoradiotherapy.

Pancreatic adenocarcinoma (PAAD) is a pancreatic disease with considerable mortality worldwide. Because of a lack of obvious symptoms at the early stage, most PAAD patients are diagnosed at the terminal stage and prognosis is usually poor. In this study, we firstly obtained RNA sequencing data of 181 patients with PAAD from The Cancer Genome Atlas (TCGA) database to identify early diagnostic biomarkers for PAAD. Survival-related mRNAs were identified using a weighted gene co-expression network analysis (WGCNA), and then a linear prognostic model of seven long non-coding RNAs (lncRNAs) was established using univariate and multivariate Cox proportional hazards regression analyses, which is verified using a time-dependent receiver operating characteristic (ROC) curve analysis. Finally, according to the survival analysis, we constructed a survival-related competing endogenous RNA (ceRNA) network. Our results showed that: (1) The upregulated genes related to cell cycle-related pathway (including homologous recombination, DNA replication and mismatch repair) in PAAD can increase the proliferation ability of cancer cells; (2) The 7-lncRNA signature can predict the overall survival (OS) of PAAD patients; and (3) The key mRNAs and lncRNAs are involved in mutual regulation in the ceRNA network.

Metastatic colorectal cancer is commonly treated with oxaliplatin. However, patients may develop resistance to treatment over time, and currently, there are no validated predictive biomarkers for this resistance. A differential analysis of the transcriptome and DNA methylome of colorectal cancer cell lines, classified by their varying IC50 values for oxaliplatin, revealed that genes associated with resistance were enriched for interferon regulatory factor pathways. In contrast, sensitive genes showed enrichment for transcription, amino acid metabolism, development, and binding motifs of c-MyC:Max, AP1 and others. In univariate Cox’s proportional hazard model analysis, it was found that UBE2H expression is linked to shorter survival time in the TCGA dataset and was further validated across five GEO datasets of colorectal cancer. The transcription of UBE2H, along with its gene body methylation, and copy number variation was found to be higher in resistant cell lines compared to sensitive ones. Additionally, UBE2H levels were higher in cancer samples than in control samples, while the promoter methylation is lower in cancer samples than in control samples. In groups with high UBE2H, there was an increased infiltration of eight cell types, including CD8+ T cells and type 2 T helper cells. Conversely, only T helper 17 cells showed reduced infiltration. Moreover, UBE2H expression was positively correlated with checkpoint inhibitors, CTLA4 and PDCD1, along with immune regulatory genes, such as FOXP3, IL10, IGFB1, CD274, and LAG3, etc. Analysis of single cell RNA-sequencing data revealed that UBE2H expression is elevated in undifferentiated and proliferative cells located at the base of intestinal crypts in normal colon tissue. Our findings suggest that UBE2H could serve as a resistance marker to oxaliplatin, as it is associated with methylation, the presence of proliferative and undifferentiated cancer cells, and immune exhaustion. The proposed analytical pipeline may also be applicable to other cancers and diseases.

This paper proposes an efficient ensemble system to tackle the protein secondary structure prediction problem with neural networks as base classifiers. The experimental results show that the multi-layer system can lead to better results. When deploying more accurate classifiers, the higher accuracy of the ensemble system can be obtained.