Hunting human disease genes: lessons from the past, challenges for the future

Abstract

The concept that a specific alteration in an individual's DNA can result in disease is central to our notion of molecular medicine. The molecular basis of more than 3,500 Mendelian disorders has now been identified. In contrast, the identification of genes for common disease has been much more challenging. We discuss historical and contemporary approaches to disease gene identification, focusing on novel opportunities such as the use of population extremes and the identification of rare variants. While our ability to sequence DNA has advanced dramatically, assigning function to a given sequence change remains a major challenge, highlighting the need for both bioinformatics and functional approaches to appropriately interpret these data. We review progress in mapping and identifying human disease genes and discuss future challenges and opportunities for the field.

Fig. 1

Mendelian diseases of known molecular basis. The x-axis shows the time in years from 1988 to present. The left-hand y-axis indicates the cumulative number of Mendelian diseases for which a molecular basis is identified and the right-hand y-axis expresses this as a percentage of the approximately 7,000 Mendelian disorders that have been described. The first disease Mendelian disease gene to be cloned was the CFTR transporter involved in cystic fibrosis in 1989. Following the release of the first draft of the human genome sequence in 2001, the rate of discovery of Mendelian disease genes increased dramatically. As of November 2012, the molecular basis of 3,650 Mendelian diseases, or slightly more than 50 % of all Mendelian diseases, is known. Data are adapted from (McKusick ; Antonarakis and Beckmann ; Hamosh et al. ; Pearson et al. ; Online Mendelian Inheritance in Man 2012)

Fig. 2

Theoretical example of filtering steps used to limit number of variants identified by exome sequencing. The three examples indicate two affected individuals who are related to different degrees. Comparing the exomes of siblings, 1st cousins or 2nd cousins will limit candidate variants to the roughly 1/2th, 1/8th or 1/32nd of the genome, respectively, that is shared by two such individuals. Additional discrete filtering and bioinformatics steps can further reduce the number of candidate variants. Examples assume that all but 2 % of variants will be identified in public databases such as dbSNP and 1000 Genomes Project and that approximately 20 % of novel variants will be predicted to be deleterious to the function of the encoded protein

Fig. 3

Sequencing the extremes to identify rare variants involved in common disease. The individuals from the high and low ends of the population distribution of a trait are chosen for study. Either candidate genes or whole exomes or genomes are sequenced and the number of rare variants in a given gene is compared between the high and low groups. An excess of rare sequence variants in a given gene in the high versus low (or vice versa) group provides evidence for a role of that gene in the phenotype under study. See text for examples of where this has been used