Review on Genetic Risk Score and Cancer Prediction Models

Objectives: In genome-wide association studies (GWASs), single-nucleotide polymorphisms (SNPs) that have been identified as cancer-associated loci are common, but they confer only small increases in risk. The question was whether combining multiple disease-related SNPs and the modest effects within Genetic Risk Score (GRS) may be useful in identifying subgroups that are at high risk of cancer. Methods: In this paper, we first reviewed articles that examined the predictability of GRS on cancer prediction models. Our data sources included a PubMed search of the literature published until February 2014. Secondly, we have calculated the GRS using the data example data with five SNPs related colorectal cancer (CRC) obtained from the Korean cancer prevention study II. Two approaches were used to calculate the GRS: a simple risk alleles count method (counted GRS) and a weighted method based on the genotype frequencies for each SNP and the effect sizes (allelic odds ratio or beta coefficient) from our study (weighted GRS). Results: Of 31 studies initially identified, 16 (135,110 participants) met the inclusion criteria. Among 16 articles, 7 studies were related to prostate cancer, 6 studies to breast cancer, and 3 studies to colon cancer and lung cancer. Fifteen studies except for one study concluded that in general, a genetic score may be helpful or useful in identifying the high risk group and particularly to determining the high risk individual among patients within a ‘‘gray zone’’ of cancer risk. The weighted GRS with age and sex (AUC=0.9333) had higher predictability on the CRC risk than the model with GRS alone (AUC=0.816). Conclusions: Although adding GRS improves prediction model performance, the clinical utility of these genetic risk models is limited. Nonetheless, the modelling suggests public health potential since it is possible to stratify the population into cancer risk categories, thereby informing targeted prevention and management.ope