Based on complex networks and machine learning methods, this paper studies the mining of colorectal cancer treatment genes, and innovatively combines a variety of feature extraction and comparative analysis methods, from gene network features, gene attribute features, network and attribute integration The three aspects of characteristics comprehensively excavate the genetic characteristics, and demonstrate the feasibility of the study through comparative analysis from different perspectives. Constructing a colorectal cancer gene network, analyzing the changes in the network structure during the development of colorectal cancer, and mining the network characteristics of genes are the first issues to be studied in this paper. The analysis of the network structure compares the changes in the network structure of the driver genes in the Normal network and the Tumor network and the edge mechanism of the driver genes, and the distribution of the eigenvalues of the driver genes and non-driver genes in the Tumor network. During the development of colorectal cancer, the network structure of the gene has changed significantly, and the prediction results based on the network structure show better prediction results than the non-network structure. These findings are feasible for the research direction of this paper. The argument is carried out, and the relevant analysis results are also given in the article. However, the research method of the thesis is based on network research, so comparing structural features with non-structural features only shows that structural features have a good classification ability, and cannot directly explain that modeling using gene networks is better than not using gene networks. Finally, based on the random forest, the optimized classification is improved to reveal the important factors affecting the diagnosis of colorectal cancer, and then to identify the true potential colorectal cancer driver genes, providing guidance for the clinical research of colorectal cancer and driving gene mining.

Background: Recently, the incidence of cholangiocarcinoma (CCA) has gradually increased. As CCA has a poor prognosis, the ideal survival rate is scarce for patients. The abnormal expressed tsRNAs may regulate the progression of a variety of tumors, and tsRNAs is expected to become a new diagnostic biomarker of cancer. However, the expression of tsRNAs is obscure and should be elucidated in CCA. Methods: High‐throughput RNA sequencing technology (RNA‐seq) was utilized to determine the overall expression profiles of tsRNAs in three pairs CCA and adjacent normal tissues and to screen the tsRNAs that were differentially expressed. The target genes of dysregulated tsRNAs were predicted and the biological effects and potential signaling pathways of these target genes were explored by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. Quantitative real‐time polymerase chain reaction (qRT‐PCR) was used to validate 11 differentially expressed tRFs with 12 pairs CCA and adjacent normal tissues. Results: High‐throughput RNA‐seq totally demonstrated 535 dysregulated tsRNAs, of which 241 tsRNAs were upregulated, such as tRF‐21‐YLKZKWE5D，tRF‐16‐9NF5W8B，tRF‐27‐78YLKZKWE52，tRF‐19‐RLXN48KP，tRF‐33‐IK9NJ4S2I7L7DV，tRF‐19‐F8DHXYIV, and 294 tsRNAs were downregulated (tRF‐20‐739P8WQ0, tRF‐34‐JJ6RRNLIK898HR, tRF‐17‐VL8RPY5, tRF‐23‐YP9LON4VDP, tRF‐39‐EH623K76IR3DR2I2, tRF‐17‐18YKISM, tRF‐19‐Q1Q89PJZ, etc.) in CCA compared with adjacent normal tissues (|log2 [fold change] | ≥ 1 and p value <0.05). GO and KEGG enrichment analyses indicated that the target genes of dysregulated tRFs (tRF‐34‐JJ6RRNLIK898HR, tRF‐38‐0668K87SERM492V, and tRF‐39‐0668K87SERM492E2) were mainly enriched in the Notch signaling pathway, Hippo signaling pathway, cAMP signaling pathway and in growth hormone synthesis, secretion and action, etc. qRT‐PCR result showed that tRF‐34‐JJ6RRNLIK898HR/tRF‐38‐0668K87SERM492V/tRF‐39‐0668K87SERM492E2 was downregulated ( p = 0.021), and tRF‐20‐LE2WMK81 was upregulated in CCA ( p = 0.033). Conclusion: Differentially expressed tRFs in CCA are enriched in many pathways associated with neoplasms, which may impact the tumor progression and have potential to be diagnostic biomarkers and therapeutic targets of CCA.

Helicobacter pylori (H. pylori) is one of the globally recognized causative factors of gastric cancer (GC). Currently, no definite therapy and drugs for H. pylori-related GC have been widely acknowledged although H. pylori infection could be eradicated in early stage. Inflammation and immune response are spontaneous essential stages during H. pylori infection. H pylori may mediate immune escape by affecting inflammation and immune response, leading to gastric carcinogenesis. As an important component of transcriptome, non-coding RNAs (ncRNAs) have been proven to play crucial roles in the genesis and development of H. pylori-induced GC. This review briefly described the effects of ncRNAs on H. pylori-related GC from the perspective of inflammation and immune response, as well as their association with inflammatory reaction and immune microenvironment. We aim to explore the potential of ncRNAs as markers for the early diagnosis, prognosis, and treatment of H. pylori-related GC. The ncRNAs involved in H. pylori-related GC may all hold promise as novel therapeutic targets for immunotherapy.

Based on the massively parallel deep learning algorithm, this paper studies familial polyposis colorectal carcinogenesis, and proposes a semi-supervised multi-task survival analysis method based on deep learning, which transforms the survival analysis problem into multi-timepoint survival probability prediction. The multi-task learning model is composed of semi-supervised learning problems. We use semi-supervised loss and sorting loss to deal with data of censorship and the non-increasing probability of survival probability. It established a prognostic risk prediction model for familial polyposis colorectal cancer based on a semi-supervised logistic regression method and learns from supervised learning from five aspects of discriminating ability, interpretability, and clinical practicality. The method comparison expands the current understanding of the generalization capabilities of different models and provides a reference for the establishment of clinical prediction models. The effectiveness of this method was verified by external data and provided technical support for constructing a prognostic model with application value for multi-center real clinical data. This model has demonstrated better prediction performance than common models in the prognostic task of familial polyposis colorectal cancer, and successfully described the mechanism of action of predictors.