Method for identifying cancer driver gene by using consensus prediction result

Abstract

The invention discloses a method for identifying a cancer driver gene by using a consensus prediction result. The method comprises the following steps: S1, receiving a mutation annotation format (MAF) file as input; S2, processing all the preprocessed input mutation data so as to respectively obtain a candidate driver gene list of each strategy; S3, on the basis of each differential driving gene list, obtaining a common driving gene list by using a rank integration method RobustRankAggreg; S4, evaluating the performance of the result by using Top-N-Precision and Top-N-nDCG, and carrying out KEGG pathway and gene ontology analysis on the common driver gene; S5, obtaining a consensus driving gene list by using an RAA algorithm; S6, employing the SuperExactTest and the Circos for organizing a visualization result. The method has certain superiority in driver gene prediction, although height difference exists between different driver gene identification strategies, not only can the most reliable driver gene be identified through cross analysis of results of each independent strategy, but also potential novel driver genes with undefined features can be found.

Description

technical field [0001] The invention relates to the technical field of cancer driver gene identification, in particular to a method for identifying cancer driver genes by using consensus prediction results. Background technique [0002] Next-generation sequencing technologies have identified millions of somatic mutations in human cancer cells, however, one of the major challenges in interpreting cancer genomes is how to efficiently distinguish driver mutations from guest mutations, under specific microenvironmental conditions in vivo, Driver mutations are causally related to oncogenes and are actively selected along the cancer developmental lineage, whereas guest mutations do not confer clonal growth advantages and are therefore not associated with tumor development. To address this issue, a variety of models based on unique Hypotheses and strategies to identify driver genes. [0003] Several studies have been reported to benchmark these approaches using consensus cancer dr...

AI Technical Summary

Problems solved by technology

[0002] Next-generation sequencing technologies have identified millions of somatic mutations in human cancer cells, however, one of the major challenges in interpreting cancer genomes is how to efficiently distinguish driver mutations from guest mutations, under specific microenvironmental conditions in vivo, Driver mutations are causally linked to oncogenes and are actively selected along the cancer developmental lineage, whereas guest mutations do not confer a clonal growth advantage and are therefore irrelevant to tumor development. To address this issue, a variety of unique-based Hypotheses and Strategies to Identify Driver Gene Approaches

[0003] Several studies have been reported to benchmark these methods using consensus cancer driver genes derived from individual models, and Collin et al. proposed an evaluation framework based on the inclusion of precision, agreement, and mean log-fold change (MLFC ) to benchmark several existing models, Matan et al. also benchmarked available methods by using measures such as precision and recall, Edward et al. Dividing driver gene calling methods into four subtypes, Denis et al. provided the most comprehensive benchmark of 21 driver gene prediction methods and proposed a Borda-based integration method ConsensusDriver, but in existing methods for identifying driver genes The reliability of the integrated results needs to be improved, so a method for identifying cancer driver genes using consensus prediction results is proposed to solve the above problems

Images