General lessons from large-scale studies to identify human cancer predisposition genes.

There are now about 100 genes known to cause Mendelian inherited cancer syndromes, but these only explain a minor part of the familial clustering of the common cancers. The increased familial relative risk of cancer in the general population must largely involve genes of low- or moderate-penetrance. Until recently, attempts to identify cancer predisposition genes with low penetrance had proved similarly unrewarding. However, in the past 2 years, developments in this area have been rapid. In particular, the 'common disease-common variant' model of predisposition has come to the fore. In this model, alleles of high frequency (typically > 10%) and low penetrance (typically < two-fold increased lifetime risk) contribute substantially to susceptibility to the common human diseases, including cancers. Many common risk alleles for cancer have been found by genome-wide association studies (GWASs) in the form of tagging SNPs, although identification of the disease-causing variants generally remains a difficult problem. The 'common disease-common variant' model has recently been criticized by proponents of a 'common disease-rare variant' model. In fact, the conflict between the models is false and a more continuous approach, bounded only by technical limitations and sample sizes, appears to be more appropriate. In this review, we summarize the general findings from cancer GWASs and their problems, and discuss the issues of finding rarer variants and other forms of cancer-predisposing variation, such as copy number polymorphisms.