Skip to main content
  • Research Article
  • Open access
  • Published:

Identification of candidate protective variants for common diseases and evaluation of their protective potential



Human polymorphisms with derived alleles that are protective against disease may provide powerful translational opportunities. Here we report a method to identify such candidate polymorphisms and apply it to common non-synonymous SNPs (nsSNPs) associated with common diseases. Our study also sought to establish which of the identified protective nsSNPs show evidence of positive selection, taking this as indirect evidence that the protective variant has a beneficial effect on phenotype. Further, we performed an analysis to quantify the predicted effect of each protective variant on protein function/structure.


An initial analysis of eight SNPs previously identified as associated with age-related macular degeneration (AMD), revealed that two of them have a derived allele that is protective against developing the disease. One is in the complement component 2 gene (C2; E318D) and the other is in the complement factor B gene (CFB; R32Q). Then, combining genomewide ancestral allele information with known common disease-associated nsSNPs from the GWAS catalog, we found 32 additional SNPs which have a derived allele that is disease protective. Out of the total 34 identified candidate protective variants (CPVs), we found that 30 show stronger evidence of positive selection than the protective variant in lipoprotein lipase (LPL; S447X), which has already been translated into gene therapy. Furthermore, 11 of these CPVs have a higher probability of affecting protein structure than the lipoprotein lipase protective variant (LPL; S447X).


We identify 34 CPVs from the human genome. Diseases they confer protection against include, but are not limited to, type 2 diabetes, inflammatory bowel disease, age-related macular degeneration, multiple sclerosis and rheumatoid arthritis. We propose that those 30 CPVs with evidence of stronger positive selection than the LPL protective variant, may be considered as priority candidates for therapeutic approaches. The next step towards translation will require testing the hypotheses generated by our analyses, specifically whether the CPV arose from a gain-of-function or a loss-of-function mutation.


Mutations associated with diseases are usually considered to be detrimental to health, increasing the risk of disease. However there are a growing number of reported missense rare mutations shown to be protective, lowering the risk of certain diseases and conditions [1, 2]. These are beginning to present valuable therapeutic opportunities [3]. If such a mutation gives rise to a selective advantage, and given enough time, we may expect it to rise in allele frequency to become a common variant in the population. Indeed a well documented example of such a common variant exists, the S447X variant of the lipoprotein lipase (LPL). One phenotypic benefit of this derived allele is its significantly lower risk of cardiovascular diseases and hypertension [4]. Furthermore recently this variant has been utilized therapeutically for the rare condition LPL-deficiency; using a gene therapy approach the derived variant was successfully administered first in mice [5, 6] and now in humans as alipogene tiparvovec [7].

In the light of this development, we sought to identify other common protective variants (CPVs) from non-synonymous (nsSNPs) that have the potential to lead to a translational opportunity. The number of common variants associated with disease continues to grow owing to the endeavour of genome wide association studies (GWAS). The first ever GWAS was carried out for age-related macular degeneration (AMD) [8] and more recently a collaborative GWAS has identified a total of 19 AMD susceptibility loci [9]. These findings led us to begin our search for CPVs in the context of AMD. To ascertain whether a nsSNP is protective (i.e. the respective derived allele is protective) we need to deduce which allele is derived and which is ancestral. For SNPs with a very low minor allele frequency (< 1%), in a majority of the cases it can be quite safely assumed that the rare allele is the derived allele. However for common variants the assumption that the rare allele is the derived allele can be erroneous. Indeed if a derived allele provided a protective function and gives individuals’ a selective advantage one might expect positive selection to sweep it to become the most common allele in the population. Thus for all common variants we must deduce which allele is derived and which is ancestral using the genome alignments with primate species.

The identification of such protective derived variants is of importance to drug development as they represent naturally occurring therapeutic opportunities. There is a high failure rate in pharmaceutical research and development, with less than 5% of molecules entering Phase I clinical trials actually being approved as safe and effective [10], and around a quarter of those reaching Phase II trials failing due to toxicity [10]. The fact that a protective variant has existed in a part of the human population without adverse effects gives confidence that a therapeutic approach targeting them is less likely to fail in clinical trials.

An understanding of the protein function of a therapeutic candidate is an important factor in helping prioritize and direct further research and ultimately in selecting the most promising targets for a drug development programme [10]. Even before the function and mechanism of action of a protein are characterised through carefully constructed in vitro and/or in vivo molecular biology studies, it is possible to assemble some knowledge of a variant’s function using bioinformatics approaches. Specifically, if a protective variant’s function gives a selective advantage compared to the ancestral variant then we can use approaches to detect evidence of positive natural selection [11]. Accordingly, we undertook an analysis to establish which of the identified nsSNPs show evidence of positive selection, using this as indirect evidence of the protective variant having a beneficial effect on phenotype through a change in protein function.

A change in protein function can be construed to have arisen from either a gain-of-function or a loss-of-function mutation, thus leading to different translational opportunities. A protective gain-of-function mutation may be harnessed through utilizing the derived allele in a gene-therapy approach, such as in the S447X example described above. A protective loss-of-function mutation may be harnessed by therapies that inhibit the function of the ancestral gene [12]. We therefore performed a further bioinformatics analysis to assess whether the identified CPVs were derived through a loss or a gain-of-function mutation.


Principal logic

Our work is based on the reasoning that every biallelic SNP is composed of an ancestral allele and a derived allele. If a SNP is associated with a disease then there are two possible situations: (1) either its derived allele is associated with increased risk (being more frequent in cases than controls), or (2) its derived allele is associated with decreased risk (being more frequent in controls than cases). In the latter case, if the SNP is functionally linked to the disease (and is not just a proxy) then it may be termed a protective variant. In this study we sought to identify such protective alleles; to increase the likelihood of it being functional we limited our search to non-synonymous SNPs. We do not confirm the functional impact of these variants, and hence we cannot conclude that they are truly protective. Our study therefore identifies SNPs that they are “potentially” protective and thus we term them as candidate protective variants (CPVs).

The resulting list of CPVs is then subjected to various bioinformatic analyses to initiate the collection of evidence supporting their protective function and thus paving the way for the experimental characterization necessary to confirm whether they are truly protective.

Ascertaining the ancestral and derived allele of SNPs

This ancestral allele information was inferred from whole genome alignments of 6 primate species (H. sapiens, P. troglodytes, G. gorilla, P. pygmaeus, M. mulatta and C. jacchus) released in Ensembl Compara 59 which implemented the Enredo-Pecan-Ortheus (EPO) alignment method [13]. We accessed this information from phase 1 of the 1000 Genomes project [14] by applying the tabix command from the SAM tools package [15] to the relevant chromosome variant call format (VCF) files.

Identifying CPVs from previously reported AMD genetic associations

To identify CPVs from nsSNPs for AMD we began with the 19 signals reported in the most recent comprehensive meta-analysis [9]. For each signal we used the UCSC genome browser to search nearby genes for nsSNPs associated with AMD [using genome assembly GRCh37/hg19] [16]. Furthermore we only included those nsSNPs which have been validated by meta-analysis.

Identifying CPVs from the GWAS catalog

To identify CPVs from the raw GWAS catalog we applied seven stages of filtering. Stage 1 involved selecting those associations labelled as missense. Stage 2 involved removing all associations in which the reported risk allele was missing (denoted by a “?” in the catalog). Stage 3 kept only those associations that have an ancestral allele with a high-confidence call. Stage 4 kept only those in which this ancestral allele matches the risk allele i.e. those which have a derived allele that is protective. Stage 5 kept only those associations in which an odds ratio was reported as opposed to a beta regression value (filtering out studies related to continuous traits such as height and heart rate, rather than binomial disease status). For stage 6 we manually scrutinized those studies in which associated genes were oriented on the negative strand, for such genes some studies report the risk allele as the allele on the positive strand whereas others report the coding allele on the negative strand. We removed those associations if it was not clear from the manuscript whether the risk allele referred to the positive or negative strand (note that the reported ancestral allele always refers to the allele on the positive strand). The seventh and final stage involved removing any duplicated associations, i.e. those reported in the catalog by more than one GWAS. The respective numbers of associations remaining after each of the above filter stages are presented in Additional file 1: Table S1.

Measuring selection by calculating integrated haplotype score (iHS)

For each CPV we estimated its integrated haplotype score (iHS) using the rehh package [17]. This was standardized using the iHS distribution of randomly selected SNPs whose derived allele frequency matches that (within 2.5%) of the nsSNP. A negative iHS score indicates that the derived allele has undergone recent positive selection [11] and thus supports the derived allele having a function beneficial for an individual’s fitness. For this calculation we considered only common SNPs (maf > 0.01) from the chromosomal region that flank either side of the nsSNP by 100 kb. All SNP data was taken from the 1000 Genomes Project, and by combining all racial groups as a representative sample (n = 1092) of the human population.

We sought to increase the confidence of the iHS score by, instead of taking a single iHS value for the single nsSNP, calculating a mean iHS from a set of SNPs that will also be subject to positive selection given that the nsSNP is. This set of SNPs were defined as follows: a SNP was included in the set if satisfied all three following conditions: (i) it was in high LD with the nsSNP (pairwise D’ > 0.9) [18]; (ii) it had a derived allele that is in-phase with the derived nsSNP allele and (iii) was evolutionary younger than the nsSNP (i.e. had a derived allele frequency less than the derived nsSNP allele frequency).

A graphical way of representing these three conditions was to employ a gene or phylogenetic tree. The set defined above is equivalent to considering all those SNPs that lie on the same branch and on all branches below the nsSNP. For this purpose we implemented the package GENETREE version 9.0 [19]. Note that this program only operates under perfect phylogeny (no recombination) meaning all pairwise D’ values need to equal 1. Thus we necessarily modified our data (where all D’ > 0.9) by pruning those rare haplotypes that have occurred from recombination.

Prediction of functional effect of variants

For 33 CPVs, we used PolyPhen-2 [20] to estimate the probability of the mutation being damaging. (One of the initial 34 CPVs (rs20541 in IL13) was not present in the PolyPhen-2 database and therefore is not included in this analysis.) For comparison we also included the probability for the LPL S447X variant. The PolyPhen-2 database (polyphen-2.2.2-whess-2011_12.sqlite) was downloaded using sqlite3. A mean probability was calculated if a mutation had multiple probabilities due to it having different effects on multiple transcripts (splice variants) of the same gene. It is important to note here that the term “damaging” used by PolyPhen-2 is meant to reflect that the mutation affects protein structure and therefore may affect function either in a gain- or loss-of-function manner. In addition the CPVs were evaluated using the Combined Annotation-Dependent Depletion (CADD) method which objectively integrates many diverse annotations to quantify the pathogenicity of human variants [21].


Identification of two independent CPVs for AMD

We brought together eight AMD-nsSNP associations from six chromosomal loci, each of which had been validated by meta-analysis (Table 1). By ascertaining the ancestral allele for each of these eight nsSNPs we identified that two have a derived allele that is protective; inheriting either one of these alleles reduces the risk of developing AMD. One is the rs9332739 SNP in the complement component 2 gene, C2; the derived allele of which changes its 318th amino acid from glutamic acid to aspartic acid (E318D). The other is the rs641153 SNP in the complement factor B gene, CFB; the derived, protective allele changes its 32nd amino acid from arginine to glutamine (R32Q).

Table 1 Non-synonymous SNPs associated with AMD validated by meta-analysis

It is interesting that these two nsSNP are in close physical proximity (≈10 kb apart) but are independent from one another, having low pairwise linkage disequilibrium between them (R2 = 0.002). It is not yet known what the level of epistasis between these two variants is upon AMD risk. Assuming no epistasis such that there is no departure from independence of effect sizes, an individual homozygous with both derived alleles (318D/318D and 32Q/32Q) is approximately 10 times less likely to develop AMD than an individual homozygous with both ancestral alleles (318E/318E and 32R/32R). This calculation is based upon the homozygote ORs reported in the relevant meta-analysis [22].

Although these SNPs are not in LD with each other, it is important to note that both these SNPs are in high LD with other nsSNPs in the region. Indeed E318D in C2 is in almost complete LD with another nsSNP in the locus, the L9H variant (rs4151667) in CFB (R2 = 0.95). This is one reason why the AMD association information alone is not sufficient to conclude with certainty that these missense changes are functionally connected to AMD pathogenesis. Further evidence is required to conclude this.

AMD CPVs show stronger evidence of recent positive selection than S447X in LPL

The fact that these AMD associated variants simply exist in the human population is testament to them not having fatal consequences or adverse effects. Thus a “natural experiment” has already been performed on these variants to test for adverse effects with zero incidences reported. This increases the appeal of them as candidates for therapeutic research and applicability.

To commence the collection of evidence supporting a functional link between these two nsSNPs and AMD pathogenesis we implemented a bioinformatics approach. Specifically we assessed the extent of recent positive selection around these nsSNPs using the integrated haplotype score (iHS) [11]. If an nsSNP has a derived allele that is protective against disease then a positive selection signal is expected given an increased reproductive fitness of those individuals inheriting such an allele. We used this method to compare the evidence for positive selection of the AMD variants against that of the previously translated LPL S447X variant.

The variant R32Q in CFB presented a mean iHS of −0.72 calculated from a SNP set comprising 13 SNPs (Fig. 1a and f). For comparison, the mean iHS pertaining to the S447X in LPL was −0.03, calculated from a set of 28 SNPs (Fig. 1d and f). The more negative mean iHS of R32Q in CFB suggested it has undergone more positive selection than S447X in LPL. Considering that S447X has a validated beneficial functional consequence we provisionally concluded that the CFB R32Q variant must also have a beneficial functional consequence. Inheriting the Q allele appears to have provided a selective advantage and increases an individual’s reproductive fitness. Similarly we found the E318D variant in C2 also gives rise to a more negative mean iHS than S447X in LPL. The SNP set consisting of 9 SNPs (Fig. 1b) gives a mean iHS of −0.55 (Fig. 1f). This provided support that this locus has undergone more recent positive selection than LPL S447X, implicating a potentially advantageous function with respect to an individual’s fitness.

Fig. 1
figure 1

Integrated haplotype score analysis of the SNP sets corresponding to two AMD candidate protective nsSNPs in C2 and CFB genes and one nsSNP protective against hypertension in LPL. High LD SNP sets, represented as genetic trees, for (a) CFB R32Q, (b) C2 E318D and (d) LPL S447X, where SNPs are numbered on the branches. The numbers at the bottom of each branch represent the number of haplotypes it is comprised of. SNPs on genetic trees are aligned against UCSC genome browser (c) and (e), displaying RefSeq genes in the respective regions (nsSNPs are designated with an asterisk *). The integrated haplotype score (iHS) for each of these SNPs is calculated, represented by the distributions shown in (f) for each nsSNP. Mean iHS is represented by a black circle (), median by a thick black line and the box represents the central 50% of the data. An outlier is represented by white circle ()

We repeated this analysis using only those individuals with European ancestry from the 1000 genomes project (n = 379). As with the results above using all individuals, both the R32Q and E318D variants gave rise to a more negative mean iHS than S447X. Here however the E318D in C2 presented the highest measure of selection (mean iHS = −0.74), followed by R32Q in CFB (mean iHS = −0.33).

Identification of CPVs for other common diseases

Next we applied the same reasoning to search the entire GWAS catalog for CPVs and identified a further 32 such variants, in addition to the two associated with AMD (Table 2 and Fig. 2). Out of the total 34 CPVs, 30 gave rise to a more negative mean iHS than LPL S447X. Four of the nsSNPs had a higher mean iHS than LPL S447X, and also had a positive iHS value suggesting that their derived alleles have been subject to negative selection. Taken together the number of CPVs with a mean iHS less than zero (k = 30) is significantly greater than expected by chance (P = 2.7 × 10−6; Binomial (34, 0.5) suggesting that overall, the derived allele of a CPV is more likely to have undergone positive selection than negative selection.

Table 2 Candidate protective nsSNPs
Fig. 2
figure 2

Integrated haplotype score distributions of the SNP sets corresponding to the 34 candidate protective nsSNPs. High LD SNP sets are computed for each candidate protective nsSNP, from which a distribution of the integrated haplotype scores is calculated and presented as boxplots. The distribution for LPL S447X is shown in green and those for the two AMD nsSNPs in magenta and cyan. The number identifying the nsSNP on the y-axis corresponds to the full description in Table 2

The nsSNP that had the most negative mean iHS was found in the interleukin 7 receptor (IL7R) gene, for which the ancestral allele codes for threonine and the derived allele codes for isoleucine (T244I). This nsSNP is associated with multiple sclerosis (with the ancestral allele more frequent in cases) and has been validated by meta-analysis [23]. The nsSNP that gave rise to the next most negative mean iHS was found in the collagen 11 alpha 1 (COL11A1) gene, associated with primary open-angle glaucoma. The ancestral allele codes for proline while the derived, protective allele codes for leucine. The third most negative was found in SLC30A8 and is associated with Type 2 diabetes. Its ancestral allele codes for arginine whereas the derived, protective allele codes for tryptophan. We repeated this analysis in the sample with European ancestry and comparable mean iHS results were observed.

For each CPV, the probability of it altering protein structure was calculated using PolyPhen-2 [20]. We found that 11 of the CPVs are predicted to have a higher probability of being damaging than the LPL S447X variant (Fig. 3a). Note that here the word “damaging” refers to altering protein stability and function. We subsequently compared these PolyPhen-2 scores to their CADD scores [21] and find significant positive correlation (r = 0.63; P = 6.2 × 10−5) between the two methods (Fig. 3b and Additional file 2: Table S2).

Fig. 3
figure 3

a Probabilities of the candidate protective variants altering their respective protein function. These probabilities are the Naïve Bayes posterior probability that the missense mutation is damaging based on the PolyPhen-2 HumDiv model. b Positive correlation between PolyPhen-2 and CADD scores


Naturally occurring genetic variants with protective effect against diseases represent a valuable potential resource for translational medicine. Recent advances in gene therapy have realized this by translating the protective effect of one such variant into an effective therapeutic strategy [7]. In this work we identified a total of 34 CPVs, having the potential to protect against an array of common diseases. Two of these were identified from an analysis of AMD genetic risk factors, then 32 more were identified through the GWAS catalogue. This is not an exhaustive list because it is unlikely that a recent GWAS has the power to detect all disease-associated nsSNPs, especially rare ones. In addition we only sought those nsSNPs that were directly reported in the GWAS catalogue and thus any causal nsSNP that is captured by a proxy SNP will have been also overlooked by our present method. Linkage disequilibrium incurs a further caveat to this study; for instance, it is possible that a nsSNP at a GWAS signal results in an amino acid change that has no effect on disease risk, but it is in high LD with a non-coding causal SNP elsewhere (e.g. in a regulatory region). Recent evidence does suggest however that coding variants are more likely to be causal than regulatory variants [24]. We also have to consider that there is a chance that a derived mutation may attain a high frequency due to drift, considering that the human race has undergone population bottlenecks [25]. Accordingly we labelled the variants identified in this study as candidate protective variants. In this study we limited our search to nsSNPs directly reported in the GWAS catalog, in order to maximize the likelihood of them being functional. However we note that our list is not a comprehensive catalog of all candidate protective variants; it is also possible that there are functional nsSNPs not reported in the catalog but which are in high LD with non-coding signals (lead SNPs) reported therein.

To advance from being a candidate to being confirmed as truly protective, these variants will need experimental validation based on characterisation of their mechanism of action. Also these candidate protective loci may be subjected to deeper genetic studies, sequencing larger samples of cases and controls to determine if any rare nonsense mutations are associated with the same disease. In this way one can build up an understanding of how different types of functional alterations (e.g. protein-truncation) affect disease risk. However the bioinformatics approaches similar to the one we undertook here build powerful avenues leading to discovery of protective variants for common diseases. We undertook an analysis to determine if the protective function of the derived allele has given it a selective advantage compared to the ancestral allele. We found 30 of the CSVs have a mean iHS that is more negative than the LPL S447X SNP. As the latter has already been successfully translated into gene therapy we conclude that this finding endorses these 30 variants as strong candidates for their protective role against disease. As these variants are still only candidates for being protective, collective further evidence of selection will be informative. The recently developed Singleton Density Score (SDS) [26] is able to detect selection that has occurred more recently than iHS and therefore would be valuable in expanding our knowledge of the selection dynamics that have acted upon these CPV loci. Finally with respect to selection we did not find any of the CPVs in a set of 86 regions of selection previously identified in the human genome [27]. However even though our 34 CPVs do not reach genome-wide significance individually, we do find that taken together their iHS scores represent a significant departure towards positive selection (P = 2.7 × 10−6).

The four CPVs that have a positive mean iHS indicate that they have undergone mild negative selection. Antagonistic pleiotropy could be one explanation for this; although they have a protective function with respect to the studied disease, the derived variant may increase the risk of another disease or present a selective disadvantage. Indeed the two nsSNPs in APOE (see Table 1) make up a set of haplotypes, of which one is known as ε4 and is well documented to be implicated in the pathogenesis of both AMD and Alzheimer’s disease (AD). Whereas the ε4 allele decreases an individual’s AMD risk, it increases AD risk. Thus knowledge of such antagonistic pleiotropy is very important as it represents an obstacle for therapeutic translation; whereas an intervention based on such a variant may achieve a reduction in the risk of developing one disease, it may increase the risk of another. Thus before any variant can be deemed truly protective such pleiotropic factors must be investigated, including those factors that affect selection such as how prevalent are the diseases involved and what are their ages of onset. Because the majority of the CPVs identified here are for relatively late onset diseases, which are not expected to engender strong selective pressures, the derived allele may have undergone positive selection due to providing another fitness advantage in younger individuals.

The two CPVs in CFB and C2, both protective against AMD, are of particular interest in that although they are close in physical distance (≈10kb apart) they are in linkage equilibrium with one another. If their protective effects are found to be independent (no epistasis) then an individual homozygous for the derived allele at both sites is about 10 times less likely to develop AMD than an individual homozygous for the ancestral allele at both sites. We also find these SNPs to be of interest because they show evidence of recent positive selection (Fig. 1) although the disease they protect against is a late-onset disease. We speculate therefore that these may also be protective against another disease, or offer another selective advantage. This speculation is backed by the fact that these two nsSNPs are in genes with function in the complement system, a key component of innate immunity.

Also of interest with respect to imminent translational research is the CPV identified in SLC30A8 with the derived allele predicted to be protective against Type 2 Diabetes (T2D). Recent evidence elsewhere has validated a direct functional role of this gene in T2D pathogenesis [12]. Specifically, deep sequencing of very large samples has confirmed that rare truncating, loss-of-function mutations are protective against T2D. The authors suggested the approach based on inhibiting the protein encoded by this gene as a therapeutic strategy for T2D prevention. If a protective variant is rare then it represents a particularly valuable therapeutic opportunity because a large fraction of the population is not subjected to its protective effect. Thus if translated, a rare protective variant has the potential to benefit more of the population than if the protective variant was common (assuming those already carrying the protective variant will not experience additional protection). Motivated by this, it will also be worth exploring non-coding rare protective alleles such as the two recently found to decrease AMD risk at the CFH locus (rs191281603 and rs148553336) [28].

Bringing together the research on SLC30A8 and LPL we highlight the fact that two distinct mechanisms, based on loss-of-function and gain-of-function mutation respectively, can both generate a protective variant. Determining which mechanism gave rise to a protective variant will be consequential in choosing the type of translational approach. As suggested for SLC30A8 with respect to T2D, if loss-of-function variants are observed to be protective then a therapeutic strategy aimed at inhibiting the gene product is appropriate. On the other hand, if gain-of-function variants are observed to be protective then a gene-therapy approach is likely to be the most appropriate, as employed by the LPL S447X-based therapy. It is also conceivable to administer the gain-of-function protein to the site of pathogenesis.


This study presents the first systematic analysis of the human genome undertaken to identify CPVs. We identified 34 such variants; the diseases they confer protection against include, but are not limited to T2D, inflammatory bowel disease, multiple sclerosis and rheumatoid arthritis. We propose that these represent highly promising translational targets, potentially accelerating the pathways to novel therapeutic strategies (e.g. gene therapy). Importantly, as the CPVs are naturally occurring they may substantially improve the effectiveness and safety of such therapeutic strategies.

To investigate the evolutionary selection upon these variants we developed a new method based on the previously proposed integrated haplotype score (iHS). Applying this we found that 30 CPVs show evidence of stronger positive selection than the LPL S447X protective variant, which has already been translated into gene therapy. We also carried out a bioinformatics analysis to ascertain the type and size of the effect that the CPV has on protein function. We found that 11 of the 34 CPVs are predicted to have a higher probability of being damaging than the LPL S447X mutation. The next step towards translation will require molecular experimental approaches to test the predictions generated by our analyses. This knowledge will in turn inform the type of translational approach to take: either a gene therapy-based approach (if protection is granted by a gain of beneficial function mutation) or molecular inhibition strategies (if protection is afforded by a loss of specific function).



Alzheimer’s Disease


age-related macular degeneration


Apolipoprotein E


Combined Annotation-Dependent Depletion


complement factor B


complement factor H


collagen 11 alpha 1


candidate protective variants




genome-wide association study


integrated haplotype score


interleukin 7 receptor




linkage disequilibrium


lipoprotein lipase


non-synonymous single nucleotide polymorphism


Singleton Density Score


type 2 diabetes


University of California at Santa Cruz


  1. Steinthorsdottir V, Thorleifsson G, Sulem P, Helgason H, Grarup N, Sigurdsson A, Helgadottir HT, Johannsdottir H, Magnusson OT, Gudjonsson SA, et al. Identification of low-frequency and rare sequence variants associated with elevated or reduced risk of type 2 diabetes. Nat Genet. 2014;

  2. Cohen J, Pertsemlidis A, Kotowski IK, Graham R, Garcia CK, Hobbs HH. Low LDL cholesterol in individuals of African descent resulting from frequent nonsense mutations in PCSK9. Nat Genet. 2005;37(2):161–5.

    Article  CAS  PubMed  Google Scholar 

  3. Sullivan D, Olsson AG, Scott R, Kim JB, Xue A, Gebski V, Wasserman SM, Stein EA. Effect of a monoclonal antibody to PCSK9 on low-density lipoprotein cholesterol levels in statin-intolerant patients: the GAUSS randomized trial. JAMA. 2012;308(23):2497–506.

    Article  CAS  PubMed  Google Scholar 

  4. Niu WQ, Qi Y. Meta-based association of the lipoprotein lipase gene S447X variant with hypertension and blood pressure variation. J Hum Hypertens. 2011;25(6):383–90.

    Article  CAS  PubMed  Google Scholar 

  5. Ross CJ, Liu G, Kuivenhoven JA, Twisk J, Rip J, van Dop W, Excoffon KJ, Lewis SM, Kastelein JJ, Hayden MR. Complete rescue of lipoprotein lipase-deficient mice by somatic gene transfer of the naturally occurring LPLS447X beneficial mutation. Arterioscler Thromb Vasc Biol. 2005;25(10):2143–50.

    Article  CAS  PubMed  Google Scholar 

  6. Rader DJ. Gain-of-function mutations and therapeutic implications - lipoprotein lipase S447X to the rescue. Arterioscl Throm Vas. 2005;25(10):2018–9.

    Article  CAS  Google Scholar 

  7. Gaudet D, Methot J, Dery S, Brisson D, Essiembre C, Tremblay G, Tremblay K, de Wal J, Twisk J, van den Bulk N, et al. Efficacy and long-term safety of alipogene tiparvovec (AAV1-LPLS447X) gene therapy for lipoprotein lipase deficiency: an open-label trial. Gene Ther. 2013;20(4):361–9.

    Article  CAS  PubMed  Google Scholar 

  8. Klein RJ, Zeiss C, Chew EY, Tsai JY, Sackler RS, Haynes C, Henning AK, SanGiovanni JP, Mane SM, Mayne ST, et al. Complement factor H polymorphism in age-related macular degeneration. Science. 2005;308(5720):385–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Fritsche LG, Chen W, Schu M, Yaspan BL, Yu Y, Thorleifsson G, Zack DJ, Arakawa S, Cipriani V, Ripke S, et al. Seven new loci associated with age-related macular degeneration. Nat Genet. 2013;45(4):433–9.

    Article  CAS  PubMed  Google Scholar 

  10. Plenge RM, Scolnick EM, Altshuler D. Validating therapeutic targets through human genetics. Nat Rev Drug Discov. 2013;12(8):581–94.

    Article  CAS  PubMed  Google Scholar 

  11. Voight BF, Kudaravalli S, Wen X, Pritchard JK. A map of recent positive selection in the human genome. PLoS Biol. 2006;4(3):e72.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Flannick J, Thorleifsson G, Beer NL, Jacobs SB, Grarup N, Burtt NP, Mahajan A, Fuchsberger C, Atzmon G, Benediktsson R, et al. Loss-of-function mutations in SLC30A8 protect against type 2 diabetes. Nat Genet. 2014;46(4):357–63.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Paten B, Herrero J, Beal K, Fitzgerald S, Birney E. Enredo and pecan: genome-wide mammalian consistency-based multiple alignment with paralogs. Genome Res. 2008;18(11):1814–28.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Genomes Project C, Abecasis GR, Altshuler D, Auton A, Brooks LD, Durbin RM, Gibbs RA, Hurles ME, GA MV. A map of human genome variation from population-scale sequencing. Nature. 2010;467(7319):1061–73.

    Article  Google Scholar 

  15. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R. The sequence alignment/map format and SAMtools. Bioinformatics. 2009;25(16):2078–9.

    Article  PubMed  PubMed Central  Google Scholar 

  16. Kent WJ, Sugnet CW, Furey TS, Roskin KM, Pringle TH, Zahler AM, Haussler D. The human genome browser at UCSC. Genome Res. 2002;12(6):996–1006.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Gautier M, Vitalis R. rehh: an R package to detect footprints of selection in genome-wide SNP data from haplotype structure. Bioinformatics. 2012;28(8):1176–7.

    Article  CAS  PubMed  Google Scholar 

  18. Lewontin RC. The interaction of selection and linkage. I General Considerations; Heterotic Models. Genetics. 1964;49(1):49–67.

    CAS  PubMed  PubMed Central  Google Scholar 

  19. Bahlo M, Griffiths RC. Inference from gene trees in a subdivided population. Theor Popul Biol. 2000;57(2):79–95.

    Article  CAS  PubMed  Google Scholar 

  20. Adzhubei IA, Schmidt S, Peshkin L, Ramensky VE, Gerasimova A, Bork P, Kondrashov AS, Sunyaev SR. A method and server for predicting damaging missense mutations. Nat Methods. 2010;7(4):248–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Kircher M, Witten DM, Jain P, O'Roak BJ, Cooper GM, Shendure J. A general framework for estimating the relative pathogenicity of human genetic variants. Nat Genet. 2014;46(3):310–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Thakkinstian A, McEvoy M, Chakravarthy U, Chakrabarti S, McKay GJ, Ryu E, Silvestri G, Kaur I, Francis P, Iwata T, et al. The association between complement component 2/complement factor B polymorphisms and age-related macular degeneration: a HuGE review and meta-analysis. Am J Epidemiol. 2012;176(5):361–72.

    Article  PubMed  Google Scholar 

  23. Zhang R, Duan L, Jiang Y, Zhang X, Sun P, Li J, Zhang M, Tang G, Wang X, Li X. Association between the IL7R T244I polymorphism and multiple sclerosis: a meta-analysis. Mol Biol Rep. 2011;38(8):5079–84.

    Article  CAS  PubMed  Google Scholar 

  24. Kichaev G, Yang WY, Lindstrom S, Hormozdiari F, Eskin E, Price AL, Kraft P, Pasaniuc B. Integrating functional data to prioritize causal variants in statistical fine-mapping studies. PLoS Genet. 2014;10(10):e1004722.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Li H, Durbin R. Inference of human population history from individual whole-genome sequences. Nature. 2011;475(7357):493–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Field Y, Boyle EA, Telis N, Gao Z, Gaulton KJ, Golan D, Yengo L, Rocheleau G, Froguel P, McCarthy MI, et al. Detection of human adaptation during the past 2000 years. Science. 2016;354(6313):760–4.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Grossman SR, Andersen KG, Shlyakhter I, Tabrizi S, Winnicki S, Yen A, Park DJ, Griesemer D, Karlsson EK, Wong SH, et al. Identifying recent adaptations in large-scale genomic data. Cell. 2013;152(4):703–13.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Fritsche LG, Igl W, Bailey JN, Grassmann F, Sengupta S, Bragg-Gresham JL, Burdon KP, Hebbring SJ, Wen C, Gorski M, et al. A large genome-wide association study of age-related macular degeneration highlights contributions of rare and common variants. Nat Genet. 2016;48(2):134–43.

    Article  CAS  PubMed  Google Scholar 

  29. Sofat R, Casas JP, Webster AR, Bird AC, Mann SS, Yates JR, Moore AT, Sepp T, Cipriani V, Bunce C, et al. Complement factor H genetic variant and age-related macular degeneration: effect size, modifiers and relationship to disease subtype. Int J Epidemiol. 2012;41(1):250–62.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Yuan D, Yang Q, Liu X, Yuan S, Xie P, Liu Q. Complement factor H Val62Ile variant and risk of age-related macular degeneration: a meta-analysis. Mol Vis. 2013;19:374–83.

    CAS  PubMed  PubMed Central  Google Scholar 

  31. Tong Y, Liao J, Zhang Y, Zhou J, Zhang H, Mao M. LOC387715/HTRA1 gene polymorphisms and susceptibility to age-related macular degeneration: a HuGE review and meta-analysis. Mol Vis. 2010;16:1958–81.

    CAS  PubMed  PubMed Central  Google Scholar 

  32. Thakkinstian A, McKay GJ, McEvoy M, Chakravarthy U, Chakrabarti S, Silvestri G, Kaur I, Li X, Attia J. Systematic review and meta-analysis of the association between complement component 3 and age-related macular degeneration: a HuGE review and meta-analysis. Am J Epidemiol. 2011;173(12):1365–79.

    Article  PubMed  Google Scholar 

  33. McKay GJ, Patterson CC, Chakravarthy U, Dasari S, Klaver CC, Vingerling JR, Ho L, de Jong PT, Fletcher AE, Young IS, et al. Evidence of association of APOE with age-related macular degeneration: a pooled analysis of 15 studies. Hum Mutat. 2011;32(12):1407–16.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. De Jager PL, Jia X, Wang J, de Bakker PI, Ottoboni L, Aggarwal NT, Piccio L, Raychaudhuri S, Tran D, Aubin C, et al. Meta-analysis of genome scans and replication identify CD6, IRF8 and TNFRSF1A as new multiple sclerosis susceptibility loci. Nat Genet. 2009;41(7):776–82.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Vithana EN, Khor CC, Qiao C, Nongpiur ME, George R, Chen LJ, Do T, Abu-Amero K, Huang CK, Low S, et al. Genome-wide association analyses identify three new susceptibility loci for primary angle closure glaucoma. Nat Genet. 2012;44(10):1142–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  36. Bouatia-Naji N, Bonnefond A, Cavalcanti-Proenca C, Sparso T, Holmkvist J, Marchand M, Delplanque J, Lobbens S, Rocheleau G, Durand E, et al. A variant near MTNR1B is associated with increased fasting plasma glucose levels and type 2 diabetes risk. Nat Genet. 2009;41(1):89–94.

    Article  CAS  PubMed  Google Scholar 

  37. Strange A, Capon F, Spencer CC, Knight J, Weale ME, Allen MH, Barton A, Band G, Bellenguez C, Bergboer JG, et al. A genome-wide association study identifies new psoriasis susceptibility loci and an interaction between HLA-C and ERAP1. Nat Genet. 2010;42(11):985–90.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Trevino LR, Yang W, French D, Hunger SP, Carroll WL, Devidas M, Willman C, Neale G, Downing J, Raimondi SC, et al. Germline genomic variants associated with childhood acute lymphoblastic leukemia. Nat Genet. 2009;41(9):1001–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Anderson CA, Boucher G, Lees CW, Franke A, D'Amato M, Taylor KD, Lee JC, Goyette P, Imielinski M, Latiano A, et al. Meta-analysis identifies 29 additional ulcerative colitis risk loci, increasing the number of confirmed associations to 47. Nat Genet. 2011;43(3):246–52.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Abnet CC, Freedman ND, Hu N, Wang Z, Yu K, Shu XO, Yuan JM, Zheng W, Dawsey SM, Dong LM, et al. A shared susceptibility locus in PLCE1 at 10q23 for gastric adenocarcinoma and esophageal squamous cell carcinoma. Nat Genet. 2010;42(9):764–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Yu XQ, Li M, Zhang H, Low HQ, Wei X, Wang JQ, Sun LD, Sim KS, Li Y, Foo JN, et al. A genome-wide association study in Han Chinese identifies multiple susceptibility loci for IgA nephropathy. Nat Genet. 2012;44(2):178–82.

    Article  CAS  Google Scholar 

  42. Yang W, Shen N, Ye DQ, Liu Q, Zhang Y, Qian XX, Hirankarn N, Ying D, Pan HF, Mok CC, et al. Genome-wide association study in Asian populations identifies variants in ETS1 and WDFY4 associated with systemic lupus erythematosus. PLoS Genet. 2010;6(2):e1000841.

    Article  PubMed  PubMed Central  Google Scholar 

  43. Wu C, Hu Z, He Z, Jia W, Wang F, Zhou Y, Liu Z, Zhan Q, Liu Y, Yu D, et al. Genome-wide association study identifies three new susceptibility loci for esophageal squamous-cell carcinoma in Chinese populations. Nat Genet. 2011;43(7):679–84.

    Article  CAS  PubMed  Google Scholar 

  44. Jostins L, Ripke S, Weersma RK, Duerr RH, McGovern DP, Hui KY, Lee JC, Schumm LP, Sharma Y, Anderson CA, et al. Host-microbe interactions have shaped the genetic architecture of inflammatory bowel disease. Nature. 2012;491(7422):119–24.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Kirino Y, Bertsias G, Ishigatsubo Y, Mizuki N, Tugal-Tutkun I, Seyahi E, Ozyazgan Y, Sacli FS, Erer B, Inoko H, et al. Genome-wide association analysis identifies new susceptibility loci for Behcet's disease and epistasis between HLA-B*51 and ERAP1. Nat Genet. 2013;45(2):202–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Schunkert H, Konig IR, Kathiresan S, Reilly MP, Assimes TL, Holm H, Preuss M, Stewart AF, Barbalic M, Gieger C, et al. Large-scale association analysis identifies 13 new susceptibility loci for coronary artery disease. Nat Genet. 2011;43(4):333–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Wallace C, Smyth DJ, Maisuria-Armer M, Walker NM, Todd JA, Clayton DG. The imprinted DLK1-MEG3 gene region on chromosome 14q32.2 alters susceptibility to type 1 diabetes. Nat Genet. 2010;42(1):68–71.

    Article  CAS  PubMed  Google Scholar 

  48. Berndt SI, Gustafsson S, Magi R, Ganna A, Wheeler E, Feitosa MF, Justice AE, Monda KL, Croteau-Chonka DC, Day FR, et al. Genome-wide meta-analysis identifies 11 new loci for anthropometric traits and provides insights into genetic architecture. Nat Genet. 2013;45(5):501–12.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  49. Scott LJ, Mohlke KL, Bonnycastle LL, Willer CJ, Li Y, Duren WL, Erdos MR, Stringham HM, Chines PS, Jackson AU, et al. A genome-wide association study of type 2 diabetes in Finns detects multiple susceptibility variants. Science. 2007;316(5829):1341–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Freilinger T, Anttila V, de Vries B, Malik R, Kallela M, Terwindt GM, Pozo-Rosich P, Winsvold B, Nyholt DR, van Oosterhout WP, et al. Genome-wide association analysis identifies susceptibility loci for migraine without Aura. Nat Genet. 2012;44(7):777–82.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Thorleifsson G, Magnusson KP, Sulem P, Walters GB, Gudbjartsson DF, Stefansson H, Jonsson T, Jonasdottir A, Stefansdottir G, Masson G, et al. Common sequence variants in the LOXL1 gene confer susceptibility to exfoliation glaucoma. Science. 2007;317(5843):1397–400.

    Article  CAS  PubMed  Google Scholar 

  52. Duerr RH, Taylor KD, Brant SR, Rioux JD, Silverberg MS, Daly MJ, Steinhart AH, Abraham C, Regueiro M, Griffiths A, et al. A genome-wide association study identifies IL23R as an inflammatory bowel disease gene. Science. 2006;314(5804):1461–3.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Sawcer S, Hellenthal G, Pirinen M, Spencer CC, Patsopoulos NA, Moutsianas L, Dilthey A, Su Z, Freeman C, Hunt SE, et al. Genetic risk and a primary role for cell-mediated immune mechanisms in multiple sclerosis. Nature. 2011;476(7359):214–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  54. Myouzen K, Kochi Y, Okada Y, Terao C, Suzuki A, Ikari K, Tsunoda T, Takahashi A, Kubo M, Taniguchi A, et al. Functional variants in NFKBIE and RTKN2 involved in activation of the NF-kappaB pathway are associated with rheumatoid arthritis in Japanese. PLoS Genet. 2012;8(9):e1002949.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Fingerlin TE, Murphy E, Zhang W, Peljto AL, Brown KK, Steele MP, Loyd JE, Cosgrove GP, Lynch D, Groshong S, et al. Genome-wide association study identifies multiple susceptibility loci for pulmonary fibrosis. Nat Genet. 2013;45(6):613–20.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  56. Siddiq A, Couch FJ, Chen GK, Lindstrom S, Eccles D, Millikan RC, Michailidou K, Stram DO, Beckmann L, Rhie SK, et al. A meta-analysis of genome-wide association studies of breast cancer identifies two novel susceptibility loci at 6q14 and 20q11. Hum Mol Genet. 2012;21(24):5373–84.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  57. Kozyrev SV, Abelson AK, Wojcik J, Zaghlool A, Linga Reddy MV, Sanchez E, Gunnarsson I, Svenungsson E, Sturfelt G, Jonsen A, et al. Functional variants in the B-cell gene BANK1 are associated with systemic lupus erythematosus. Nat Genet. 2008;40(2):211–6.

    Article  CAS  PubMed  Google Scholar 

  58. Bolton KL, Tyrer J, Song H, Ramus SJ, Notaridou M, Jones C, Sher T, Gentry-Maharaj A, Wozniak E, Tsai YY, et al. Common variants at 19p13 are associated with susceptibility to ovarian cancer. Nat Genet. 2010;42(10):880–4.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  59. Okada Y, Terao C, Ikari K, Kochi Y, Ohmura K, Suzuki A, Kawaguchi T, Stahl EA, Kurreeman FA, Nishida N, et al. Meta-analysis identifies nine new loci associated with rheumatoid arthritis in the Japanese population. Nat Genet. 2012;44(5):511–6.

    Article  CAS  PubMed  Google Scholar 

  60. Voight BF, Scott LJ, Steinthorsdottir V, Morris AP, Dina C, Welch RP, Zeggini E, Huth C, Aulchenko YS, Thorleifsson G, et al. Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat Genet. 2010;42(7):579–89.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  61. Zeggini E, Scott LJ, Saxena R, Voight BF, Marchini JL, Hu T, de Bakker PI, Abecasis GR, Almgren P, Andersen G, et al. Meta-analysis of genome-wide association data and large-scale replication identifies additional susceptibility loci for type 2 diabetes. Nat Genet. 2008;40(5):638–45.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  62. Nair RP, Duffin KC, Helms C, Ding J, Stuart PE, Goldgar D, Gudjonsson JE, Li Y, Tejasvi T, Feng BJ, et al. Genome-wide scan reveals association of psoriasis with IL-23 and NF-kappaB pathways. Nat Genet. 2009;41(2):199–204.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  63. Schumacher FR, Berndt SI, Siddiq A, Jacobs KB, Wang Z, Lindstrom S, Stevens VL, Chen C, Mondul AM, Travis RC, et al. Genome-wide association study identifies new prostate cancer susceptibility loci. Hum Mol Genet. 2011;20(19):3867–75.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references


The authors gratefully acknowledge the support from R & D Royal Wolverhampton NHS Trust, Wolverhampton, UK.


No funding was obtained for this study.

Availability of data and materials

The datasets analysed during the current study are available from the GWAS Catalog, (accessed 28/12/2013) and for the ancestral allele information from the 1000 Genomes Project .

Author information

Authors and Affiliations



JMB, LP and NH designed the study; JMB acquired the data and carried out the analysis; all authors contributed to the analysis and interpretation of data. All authors were involved in drafting the manuscript or revising it critically for important intellectual content, gave final approval of the version to be published and agreed to be accountable for all aspects related to the accuracy or integrity of any part of the work.

Corresponding author

Correspondence to Luminita Paraoan.

Ethics declarations

Ethics approval and consent to participate

No human or animal material has been directly used, as the study has used publicly available datasets.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Additional files

Additional file 1: Table S1.

Summary of filtering stages performed on GWAS catalog. (DOCX 59 kb)

Additional file 2: Table S2.

Comparison of methods (PolyPhen-2 and CADD) to predict functional effect of CPVs. (DOCX 92 kb)

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Butler, J.M., Hall, N., Narendran, N. et al. Identification of candidate protective variants for common diseases and evaluation of their protective potential. BMC Genomics 18, 575 (2017).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: