Development of a general logistic model for disease risk prediction using multiple SNPs.

Human diseases are usually linked to multiloci genetic alterations, including single-nucleotide polymorphisms (SNPs). Methods to use these SNPs for disease risk prediction (DRP) are of clinical interest. DRP algorithms explored by commercial companies to date have tended to be complex and led to controversial prediction results. Here, we present a general approach for establishing a logistic model-based DRP algorithm, in which multiple SNP risk factors from different publications are directly used. In particular, the coefficient β of each SNP is set as the natural logarithm of the reported odds ratio, and the constant coefficient β is comprehensively determined by the coefficient and frequency of each SNP and the average disease risk in populations. Furthermore, homozygous SNP is considered a dummy variable, and the SNPs are updated (addition, deletion and modification) if necessary. Importantly, we validated this algorithm as a proof of concept: two patients with lung cancer were identified as the maximum risk cases from 57 Chinese individuals. Our logistic model-based DRP algorithm is apparently more intuitive and self-evident than the algorithms explored by commercial companies, and it may facilitate DRP commercialization in the era of personalized medicine.