Genetic risk score for ovarian cancer based on chromosomal-scale length variation

Abstract Introduction Twin studies indicate that a substantial fraction of ovarian cancers should be predictable from genetic testing. Genetic risk scores can stratify women into different classes risk. Higher treated or screened for cancer, which reduce cancer death rates. However, current do not work well. We developed score based on variations in the length chromosomes. Methods evaluated this using data collected by The Cancer Genome Atlas. synthesized dataset 414 who had serous carcinoma and 4225 no form cancer. characterized each woman 22 numbers, representing chromosome their germ line DNA. used gradient boosting machine to build classifier predict whether been diagnosed with Results chromosomal-scale variation could such highest 20% 160x (95% confidence interval 50x-450x) compared lowest 20%. we an area under curve receiver operating characteristic 0.88 0.86–0.91). Conclusion A DNA provides effective means predicting will develop