Prediction And Analysis Of Cancer Synthetic Lethal Gene Pairs Based On Machine Learning And Statistical Inference

Cancer has seriously threatened human life and health.The global incidence and death burden of cancer are increasing rapidly.It is urgent to develop new safe and effective precision treatment strategies.Synthetic lethality(SL),as a new anti-cancer approach,is receiving increasing attention.It refers to the fact that damage to a single gene in a gene pair is not lethal to cells,while simultaneous damage to both genes can lead to cell death.The development of anticancer drugs based on the concept of synthetic lethality is expected to overcome the toxic side effects,drug resistance,and target limitations of traditional targeted therapy,and has become a safer and more effective precision medical strategy.It is particularly important to systematically obtain synthetic lethal genes for their potential applications.Therefore,this study first developed a prediction method based on machine learning and statistical inference to obtain abundant high-quality cancer specific synthetic lethal genes,providing high-quality data reference for subsequent research on anticancer drug development and other related fields.Furthermore,the prediction results were deeply analyzed and validated to fully explore the possible role of synthetic lethal genes in the occurrence and development of cancer and evaluate their clinical application potential.Finally,the cancer specific synthetic lethality database was constructed to provide data reference and analysis platform for related research.The main research contents of this article are as follows:1.To achieve cancer unbiased prediction of synthetic lethal genes,publicly validated synthetic lethal gene pairs were collected as training data,and machine learning was used to implement classification and novelty detection,solving the problem of negative sample missing in synthetic lethality prediction.By integrating data such as gene dependence,protein interaction networks,and functional similarity scores,synthetic lethality were comprehensively characterized.The deep mechanisms of synthetic lethality contained in the feature values were fitted using models.Finally,the results from other computational prediction studies were input for secondary prediction and the prediction results of different models were integrated,achieving high-precision cancer unbiased synthetic lethality prediction.A total of 49,669 potential synthetic lethal gene pairs were obtained,providing data reference for subsequent screening of cancer specific synthetic lethal gene pairs.2.Regarding the cancer specificity of synthetic lethality,further statistical inferences were made on the above machine learning results across 33 cancer types to obtain results for specific cancers.Using multi-omic data such as gene expression,mutation,and copy number variation,analysis was conducted based on the mutual exclusivity and differential expression of synthetic lethality,and two statistical inference strategies were designed to achieve cancer specific prediction of synthetic lethality.Finally,a total of 14,582 synthetic lethal gene pairs were obtained in 33 types of cancer,with significant differences in predicted results among different cancers,and most gene pairs only appeared in a certain type of cancer.These results can provide data support and methodological reference for subsequent related research.3.In order to understand the possible role of synthetic lethal genes in the occurrence and development of cancer and their potential values in prognosis,in-depth analysis was performed based on the above results,and biological experiments were conducted to verify the reliability of the prediction.The analysis was carried out from the perspectives of hot spot mutation,interaction network,function enrichment,drug sensitivity and prognosis.The results showed that APC,TP53 and TTN were hot spot mutations in many cancers,and they were closely related to the occurrence,development,and prognosis of cancer.The synthetic lethal genes with high frequencies mainly contribute to the cell cycle pathway and involved in DNA replication,ATPase activation and other functions.The drug analysis results showed that the sensitivity of cells with TP53 synthetic lethal partner mutations to MIRA-1 was significantly increased,which may be related to the inactivation of cells caused by synthetic lethality,revealing its potential application in cancer treatment.The patients with gene pair inactivation showed better prognosis in breast cancer and other cancers,which may be due to the occurrence of synthetic lethal effect caused by gene pair inactivation,thus inhibiting the activity of tumor cells.In addition,further experimental validation in colon cancer cells confirmed the existence of a synthetic lethal interaction between TP53 and USP1,proving the accuracy of the predicted results.4.To publish the analysis and prediction results of this study,CSSLDB(http://www.tmliang.cn/CSSL),a user-friendly cancer specific synthetic lethality database was developed,and it provided multiple analysis functions.Users can query the prediction results of specific genes and cancers in the database,and perform interaction network analysis,mutual exclusivity analysis,differential expression analysis,survival analysis,and drug sensitivity analysis on relevant gene pairs.The functions of CSSLDB were consistent with the key points of synthetic lethality research,which provided a powerful tool for related research.Taken together,to obtain high-quality synthetic lethal genes in various types of cancer,this study first applied machine learning methods for cancer unbiased synthetic lethality prediction,and then applied statistical inference strategies based on mutual exclusivity and differential expression in various cancers to achieve precise prediction,obtaining synthetic lethal interactions for specific cancers.Furthermore,the results were analyzed and validated from multiple perspectives,delving into the potential role and application potential of synthetic lethal genes in the occurrence and development of cancer.Finally,an online database was established to publish relevant research results and develop online analysis tools to provide a query and analysis platform for synthetic lethal gene pairs for relevant research.These results can provide data support and methodological references for research related to synthetic lethality,thereby promoting the development of precision medicine based on the concept of synthetic lethality.