The Use of Non-Variant Sites to Improve the Clinical Assessment of Whole-Genome Sequence Data.

Genetic testing, which is now a routine part of clinical practice and disease management protocols, is often based on the assessment of small panels of variants or genes. On the other hand, continuous improvements in the speed and per-base costs of sequencing have now made whole exome sequencing (WES) and whole genome sequencing (WGS) viable strategies for targeted or complete genetic analysis, respectively. Standard WGS/WES data analytical workflows generally rely on calling of sequence variants respect to the reference genome sequence. However, the reference genome sequence contains a large number of sites represented by rare alleles, by known pathogenic alleles and by alleles strongly associated to disease by GWAS. It's thus critical, for clinical applications of WGS and WES, to interpret whether non-variant sites are homozygous for the reference allele or if the corresponding genotype cannot be reliably called. Here we show that an alternative analytical approach based on the analysis of both variant and non-variant sites from WGS data allows to genotype more than 92% of sites corresponding to known SNPs compared to 6% genotyped by standard variant analysis. These include homozygous reference sites of clinical interest, thus leading to a broad and comprehensive characterization of variation necessary to an accurate evaluation of disease risk. Altogether, our findings indicate that characterization of both variant and non-variant clinically informative sites in the genome is necessary to allow an accurate clinical assessment of a personal genome. Finally, we propose a highly efficient extended VCF (eVCF) file format which allows to store genotype calls for sites of clinical interest while remaining compatible with current variant interpretation software.

Imputation reliability on DNA biallelic markers for drug metabolism studies.

BACKGROUND: Imputation is a statistical process used to predict genotypes of loci not directly assayed in a sample of individuals. Our goal is to measure the performance of imputation in predicting the genotype of the best known gene polymorphisms involved in drug metabolism using a common SNP array genotyping platform generally exploited in genome wide association studies. METHODS: Thirty-nine (39) individuals were genotyped with both Affymetrix Genome Wide Human SNP 6.0 (AFFY) and Affymetrix DMET Plus (DMET) platforms. AFFY and DMET contain nearly 900000 and 1931 markers respectively. We used a 1000 Genomes Pilot + HapMap 3 reference panel. Imputation was performed using the computer program Impute, version 2. SNPs contained in DMET, but not imputed, were analysed studying markers around their chromosome regions. The efficacy of the imputation was measured evaluating the number of successfully imputed SNPs (SSNPs). RESULTS: The imputation predicted the genotypes of 654 SNPs not present in the AFFY array, but contained in the DMET array. Approximately 1000 SNPs were not annotated in the reference panel and therefore they could not be directly imputed. After testing three different imputed genotype calling threshold (IGCT), we observed that imputation performs at its best for IGCT value equal to 50%, with rate of SSNPs (MAF > 0.05) equal to 85%. CONCLUSIONS: Most of the genes involved in drug metabolism can be imputed with high efficacy using standard genome-wide genotyping platforms and imputing procedures.

Fully non-homogeneous hidden Markov model double net: a generative model for haplotype reconstruction and block discovery.

OBJECTIVE: In the last decade, haplotype reconstruction in unrelated individuals and haplotype block discovery have riveted the attention of computer scientists due to the involved strong computational aspects. Such tasks are usually addressed separately, but recently, statistical techniques have permitted them to be solved jointly. Following this trend we propose a generative model that permits researchers to solve the two problems jointly. METHOD: The model inference is based on variational learning, which permits one to estimate quickly the model parameters while remaining robust even to local minima. The model parameters are then used to segment genotypes into blocks by thresholding a quantitative measure of boundary presence. RESULTS: Experiments on real data are presented, and state-of-the-art systems for haplotype reconstruction and strategies for block estimation are considered as comparison. CONCLUSIONS: The proposed method can be used for a fast and reliable estimation of haplotype frequencies and the relative block structure. Moreover, the method can be easily used as part of a more complex system. The threshold used for block discovery can be related to the quality-of-fit reached in the model learning, resulting in an unsupervised strategy for block estimation.