1 item tagged with 'risk prediction models'.
An integrative web platform for user-friendly application of risk prediction models in hereditary cancer predisposition
Background and Objective: Predicting individual mutation and cancer risks is essential to assist genetic counsellors in clinical decision making for patients with a hereditary cancer predisposition. … Worldwide a variety of statistical models and empirical data for risk prediction have been developed and published for hereditary breast and ovarian cancer (HBOC), and hereditary non-polyposis colorectal cancer (HNPCC / Lynch syndrome, LS). However, only few models have so far been implemented in convenient and easy-to-use computer applications. We therefore aimed to develop user-friendly applications of selected HBOC and LS risk prediction models, and to make them available through the "Leipzig Health Atlas" (LHA), a web-based multifunctional platform to share research data, novel ontologies, models and software tools with the medical and scientific community. LHA is a project funded within the BMBF initiative "i:DSem – Integrative data semantics in system medicine". Methods and Results: We selected a total of six statistical models and empirical datasets relevant for HBOC and LS: 1) the Manchester Scoring System, 2) the "Mutation Frequency Explorer" of the German Consortium for Hereditary Breast and Ovarian Cancer (GC-HBOC), 3) an extended version of the Claus model, 4) MMRpredict, 5) PREMM1,2,6, and 6) PREMM5. The Manchester Scoring System allows calculation of BRCA1/2 mutation probabilities based on aggregated family history. The "Mutation Frequency Explorer" allows flexible assessment of mutation risks in BRCA1/2 and other genes for different sets of familial cancer histories based on a large dataset from the GC-HBOC. The extended Claus model (as implemented in the commercial predigree drawing software Cyrillic 2.1.3, which is no longer supported and no longer works on newer operating systems) predicts both mutation and breast cancer risks based on structured pedigree data. MMRpredict, PREMM 1,2,6, and PREMM 5 predict mutation risks in mismatch repair genes for patients from families suspected of having LS. All models were implemented using the statistical software "R" and the R-package "Shiny". "Shiny" allows the development of interactive applications by incorporating "R" with HTML and other web technologies. The Shiny apps are accessible on the website of the "Leipzig Health Atlas" (https://www.health-atlas.de) for registered researchers and genetic counselors. Conclusions: The risk prediction apps allow convenient calculation of mutation or cancer risks for an advice-seeking individual based on pedigree data or aggregated information on the familial cancer history. Target users should be specialized health professionals (physicians and genetic counselors) and scientists to ensure correct handling of the tools and careful interpretation of results.
Date Published: 8th Mar 2019
Publication Type: Proceedings
Created: 3rd Apr 2020 at 08:11, Last updated: 10th Jul 2020 at 13:10