Development And Application Of Multi-Dimensional Pathogenicity Evaluation System For Somatic Mutations In Cancer

With the development of next-generation sequencing(NGS)technology,researchers are getting to know more and more about cancer genome.The prospective clinical genomic sequencing of patients' cancer is increasingly used as an important method for the cancer diagnosis and treatment routinely,and has been widely applicated in clinic research and auxiliary diagnosis.The precisely detecting clinically actionable somatic variations might provide an effective disease diagnosis and/or treatment.However,how to accurately analyze the somatic mutations with huge cancer genome sequencing data still has many challenges,especially for the non-bioinformatics researchers and clinicians.Therefore,it is necessary to develop an easy used computational tool for researchers and clinical geneticists to interpret somatic mutations in cancer systematically.Based on the extensive scientific literature reviews and guidelines of classifying cancer sequencing variants recommended by disease-focused expert groups,we summarized various influencing factors involved in the interpretation of somatic variations,established an evidence-based multiple-level evaluation system,and collected the available resources to create a local database.Leveraging basic annotation information,database information and databank information,we developed an interpretation tool,VIC(Variant Interpretation for Cancer),in Java programming language,and built a Web application platform.The major research work of this dissertation includes the following parts:1)Summarized the influencing factors to establish an evaluation system based on the scientific literature reviews and guidelines recommended by disease-focused expert groups.2)Established a local database based on the resources associated with cancer diagnosis and treatment from public databases and other various research sources for building the resources of our system.3)Formulated the detailed rules for rating evidence level and developed a tool(VIC)to systematically annotate tumorigenic genomic alterations.4)Designed and built a Web applications for the tool,based on Spring MVC architecture.5)Applied the tools to datasets from different sources for presenting its usage,including the dataset from the Catalog of Somatic Mutations in Cancer,the dataset of twenty-nine cases of small cell lung cancer downloaded from the cBioPortal website,the known pathogenicity dataset in the Database of Curated Mutations Knowledgebase,the case application dataset provided by the CGI,and the mutations from BRCA1/2 whole-exome sequencing data.In the COSMIC dataset,the number of "pathogenic/possible pathogenic" variants assessed by VIC is much lower than the number of "pathogenic" variants flagged in the COSMIC database.This result shows that clinical applications cannot rely solely on the result by a distinct tool.The predictions of the multiple tools make the assessment more reliable.By comparing the analytical results of twenty-nine small-cell lung cancer patients datasets and four CGI datasets using different tools,it shows that the multi-level assessment system involved multiple reference factors,and therefore the assessment result by VIC is more conservative than other tools in evaluation of the pathogenicity and the clinical availability for somatic mutations.Based on the analysis of the 1364 known pathogenicity variant sites in the DoCM database,we further compared the differences caused by different annotation tools and the biased differences of different evaluation systems.It is pointed out that the multidimensional evaluation system in this study has more applicable clinical value in interpreting the mutations for the disease treatment and diagnosis.In the thirty-five cases of BRCA1/2 whole-exome sequencing data,two mutations interpreted as "likely-pathogenicity" were found and could be used as reference in clinical applications.The evidence-based multiple-level evaluation system incorporates multi-dimensional information,such as biomarkers,biological signal pathway,mutation type and so on.By analyzing the test datasets,it demonstrated that the evaluation system can reliably classify somatic mutations in cancer.The tool,which was implemented in Java programming language,provides a new way for researchers and clinical geneticists to systematically assess pathogenicity for somatic mutations in cancer.Meanwhile,it has important value in promoting the application of NGS in both research and clinical settings.