Next generation sequencing is producing ever increasing amounts of variation information in numerous genes related to disease. Many of these remain unclassified variants (UVs) and their disease relevance is not known. Researchers of many diseases including mismatch repair system (MMR) defects, specifically Lynch syndrome, have developed approaches to classify these UVs. The ability to group the cases either as disease-related or benign can be used in diagnosis as well as for patient management.
In this issue, Thompson and coworkers (Hum Mutat 34:201–210), representing the international Colon Cancer Family Registry, have developed an approach based on multifactorial likelihood analysis. The approach has previously been used to classify variants in BRCA1 and -2. The challenge with their Bayesian method is that for the parameters used for classification, input likelihood ratios (LRs) are needed based on well-known cases. They utilized information about microsatellite instability for 10 markers and somatic BRAF protein p.V600E variations for probands in the Colon Cancer Family Registry to derive LRs for tumor characteristics. Prior probabilities of pathogenicity were estimated for missense variants based on sequence conservation and bioinformatic predictions with MAPP and PolyPhen.
Likelihood ratios for segregation and penetrance estimates were used together with the LRs for tumor characteristics to calculate multifactorial likelihood ratio for the variants. In addition to missense substitutions, intronic variants were also investigated. The authors were able to classify 31 out of 54 variants as benign (9) or pathogenic (22). The analysis indicates that pathogenicity of variants can be predicted reliably once sufficient datasets are available. For most diseases, collaboration in large consortia are essential to collect, organize and distribute the necessary clinical and genetic information.