The first successful GWAS was published in 2005 [4], and the following decade saw a flourishing and widespread application of the successful study design, leading to the identification of more than 1,000 risk loci in a variety of human complex traits, in more than 2,000 original publications. A GWAS is in simple terms a case-control association analysis, comparing the frequencies of genetic variants spread throughout the genome between two groups, patients and healthy controls: a GWAS is a scientific experiment, requiring a clear hypothesis and a well-defined phenotype and appropriate interpretation. The impact of any genetic association, wherein at multiple loci the allele frequency differs between cases and controls, must reflect the study design, as well as the population studied. Hence disease risk and disease severity, for example, are distinct questions answered in different ways. Risk loci (susceptibility loci) are determined as chromosomal regions (sometimes within single genes, susceptibility genes) where there is a statistically significant difference in the occurrence of particular variants observed in the patients (Fig. 8.1). Notably it may not be possible to always confidently assign a gene to an identified risk locus, and caution is therefore needed in making immediate biological interpretation of findings, not least because of the complexity of genetic interactions now recognised across the genome. Equally readers should be very sceptical of any study in the current era that adopts outdated single gene/variant analyses, unless it is apparent that appropriate validation cohorts are included in such candidate gene studies.

Given the high number of genetic variants tested (typically now around 1,000,000), statistical significance thresholds are stringently set by convention at P ≤ 5 × 10⁻⁸ (so-called genome-wide significance) to avoid false-positive findings (type 1 errors), and generally external validation of findings is sought as well. Inherent to the study design (association analysis), variants detected at risk loci must have a relatively high frequency to be detectable (i.e. they are ‘common’, typically with a frequency above 1–5 % in the general population), and being common they generally also exert a relatively low impact on disease risk (odds ratio typically below 1.5) [5]. The latter fact also implies that large collections of cases and controls have been required for the study design to be useful, preferably thousands, and the networks organised to recruit patients for DNA collection have promoted a collaborative, international working environment which should be considered a beneficial ‘side effect’ of GWAS [6]. For rare diseases like PSC, the statistical stringency and the low effect size of implied variants inevitably lead to false negatives (type 2 statistical errors), and this has to be kept in mind as a limitation of the data herein reviewed.

During the 1990s, liver disease genetics was dedicated to Mendelian traits, starting with the identification of genes for hyperbilirubinemias and Wilson’s disease [7–12] and a strong subsequent focus on cholestasis and hemochromatosis [13–23]. The interpretation of the gene findings in these studies has greatly influenced the thinking of susceptibility genes also in non-Mendelian (i.e. complex) diseases like PSC. This is important to be aware of, since the genetics as determined by GWAS represent fundamentally different mechanisms of causality. In Mendelian diseases, there is an approximately 1:1 relationship between genetic aberrations and disease traits (the genetic variants ‘cause disease’ frequently as mutations are structurally damaging to protein function). This being said, given the time taken to mechanistically understand even Mendelian diseases, it is relevant to reflect that disease penetrance and clinical phenotype are often not so easily explained by a single mutation-single effect model, even for diseases as classic monogenic as hemochromatosis and Wilson’s disease.

For GWAS findings, contextual factors (environment, gene-gene interactions, etc.) nevertheless play a considerably greater role than in Mendelian genetics (Fig. 8.2), making it inappropriate to assume susceptibility genes as causal (the disease-associated genetic variants in GWAS ‘do not cause disease’). This distinction between Mendelian genetics and ‘GWAS genetics’ is underlined by the fact that the overall contribution of genetics to complex traits like PSC is limited [24]. GWAS outcomes, even by mathematical extrapolations, are likely to represent a minor fraction (probably less than one third) of the susceptibility to complex traits [25, 26], and PSC so far makes up less than 10 % of the overall disease liability [27].

The relative importance of genetic influences in PSC is also evident in studies of heritability. There are no formal twin studies as for many other diseases, but registry data from Sweden makes a hazard ratio estimate of 11.1 in siblings of PSC patients (complicated by the lack of an ICD10 code for precise case identification) [28]. Such an estimate places PSC at the same degree of heritability as in most other complex autoimmune and immune-mediated conditions. In these diseases, heritability estimates mostly range near a relative sibling risk of ~10 on most instances. Notably, this number is very low compared to Mendelian conditions where relative sibling risk (depending on the penetrance of involved genetic variants) ranges from several hundreds to several thousands [29], further underlining how our thinking on the outcomes of genetic studies in complex diseases like PSC must be different from that of monogenic traits (Fig. 8.3).

As can be seen in Fig. 8.1, the genetic findings on chromosome 6 in PSC are several orders of magnitude stronger than those found in any other region. Throughout the rest of the genome, a number of weaker and less significant associations can be found. The important point about this overall genetic architecture in PSC is the fact that it resembles the genetic architecture of prototypical autoimmune diseases, e.g. type 1 diabetes, rheumatoid arthritis and multiple sclerosis. Prior to the genetic studies, it had been questioned whether PSC could be an autoimmune disease, particularly given the strong male predominance (two thirds of the patients are male) and lack of efficacy of immunosuppressive therapy. However, autoimmune diseases with a male predominance do exist (e.g. ankylosing spondylitis), and alongside other features observed (autoantibodies [30] and clonality of T cell receptors [31]), genetics clearly positions PSC as an inherently autoimmune condition, albeit one perhaps because of its biliary localisation that does not respond to classical immunosuppression. In many aspects, this global observation is one of the major outcomes of the genetic studies. The model contrasts that of other models of PSC development (e.g. toxic bile acid injury, gut leakage of bacterial components due to IBD [32]) whilst is compatible with models involving the cross-homing or cross-reactivity of lymphocytes between the bowel and the liver [33, 34].