DiGAS: Differential gene allele spectrum as descriptor in genetic studies

Abstract

Diagnosing subjects in complex genetic diseases is a very challenging task. Computational methodologies exploit information at genotype level by taking into account single nucleotide polymorphisms (SNP). They leverage the result of genome-wide association studies analysis to assign a statistical significance to each SNP. Recent methodologies extend such an approach by aggregating SNP significance at genetic level in order to identify genes that are related to the condition under study. However, such methodologies still suffer from the initial single-SNP analysis. Here, we present DiGAS, a tool for diagnosing genetic conditions by computing significance, by means of SNP information, but directly at the gene level. Such an approach is based on a generalized notion of allele spectrum, which evaluates the complete genetic alterations of the SNP set composing a gene at population level. Statistical significance of a gene is then evaluated by means of a differential analysis between the healthy and ill portions of the population. Tests, performed on well-established data sets regarding Alzheimers disease, show that DiGAS outperforms the state-of-the-art in distinguishing between ill and healthy subjects. HighlightsO_LIWe introduce a new generalized version of allele frequency spectrum. C_LIO_LIWe propose a methodology, called DiGAS, based on the new defined genomic information and independent from GWAS analysis that out-performs existing methods in distinguish healthy/ill subjects with a speed up of 5x. C_LIO_LIOn a reference Alzheimers disease genomic datasets, ADNI, DiGAS reaches F1 score up to 0.92. C_LIO_LIDiGAS methodology manages any type of genomic features, such as genes, exons, upstream/downstream regions. C_LI