Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

The sequencing and interpretation of the genome obtained from a Serbian individual

  • Wazim Mohammed Ismail ,

    Contributed equally to this work with: Wazim Mohammed Ismail, Kymberleigh A. Pagel

    Roles Writing – original draft, Writing – review & editing

    Affiliation Department of Computer Science, Indiana University, Bloomington, Indiana, United States of America

  • Kymberleigh A. Pagel ,

    Contributed equally to this work with: Wazim Mohammed Ismail, Kymberleigh A. Pagel

    Roles Methodology, Writing – original draft, Writing – review & editing

    Affiliation Department of Computer Science, Indiana University, Bloomington, Indiana, United States of America

  • Vikas Pejaver,

    Roles Methodology, Writing – original draft, Writing – review & editing

    Affiliation Department of Computer Science, Indiana University, Bloomington, Indiana, United States of America

  • Simo V. Zhang,

    Roles Methodology, Writing – original draft, Writing – review & editing

    Affiliation Department of Computer Science, Indiana University, Bloomington, Indiana, United States of America

  • Sofia Casasa,

    Roles Writing – original draft

    Affiliation Department of Biology, Indiana University, Bloomington, Indiana, United States of America

  • Matthew Mort,

    Roles Writing – review & editing

    Affiliation Institute of Medical Genetics, Cardiff University, Cardiff, United Kingdom

  • David N. Cooper,

    Roles Writing – review & editing

    Affiliation Institute of Medical Genetics, Cardiff University, Cardiff, United Kingdom

  • Matthew W. Hahn,

    Roles Writing – review & editing

    Affiliations Department of Computer Science, Indiana University, Bloomington, Indiana, United States of America, Department of Biology, Indiana University, Bloomington, Indiana, United States of America

  • Predrag Radivojac

    Roles Supervision, Writing – original draft, Writing – review & editing

    Affiliation College of Computer and Information Science, Northeastern University, Boston, Massachusetts, United States of America

The sequencing and interpretation of the genome obtained from a Serbian individual

  • Wazim Mohammed Ismail, 
  • Kymberleigh A. Pagel, 
  • Vikas Pejaver, 
  • Simo V. Zhang, 
  • Sofia Casasa, 
  • Matthew Mort, 
  • David N. Cooper, 
  • Matthew W. Hahn, 
  • Predrag Radivojac


Recent genetic studies and whole-genome sequencing projects have greatly improved our understanding of human variation and clinically actionable genetic information. Smaller ethnic populations, however, remain underrepresented in both individual and large-scale sequencing efforts and hence present an opportunity to discover new variants of biomedical and demographic significance. This report describes the sequencing and analysis of a genome obtained from an individual of Serbian origin, introducing tens of thousands of previously unknown variants to the currently available pool. Ancestry analysis places this individual in close proximity to Central and Eastern European populations; i.e., closest to Croatian, Bulgarian and Hungarian individuals and, in terms of other Europeans, furthest from Ashkenazi Jewish, Spanish, Sicilian and Baltic individuals. Our analysis confirmed gene flow between Neanderthal and ancestral pan-European populations, with similar contributions to the Serbian genome as those observed in other European groups. Finally, to assess the burden of potentially disease-causing/clinically relevant variation in the sequenced genome, we utilized manually curated genotype-phenotype association databases and variant-effect predictors. We identified several variants that have previously been associated with severe early-onset disease that is not evident in the proband, as well as putatively impactful variants that could yet prove to be clinically relevant to the proband over the next decades. The presence of numerous private and low-frequency variants, along with the observed and predicted disease-causing mutations in this genome, exemplify some of the global challenges of genome interpretation, especially in the context of under-studied ethnic groups.


The genetic variation between individuals accounts for much of observed human diversity and has the potential to provide information on phenotypic outcomes of clinical consequence. Studies of genetic variation provided by individual genome sequences have revealed that this variation differs both within and between populations, and also varies considerably depending upon the population [1]. Moreover, characterization of genetic variation of individuals from multiple populations has revealed a correlation between genetic and geographic distances, and has become relevant for determining genetic ancestry and geographic origin [26]. Therefore, the characterization of genetic variation has been of major interest for diverse research fields, including medical, biological and anthropological sciences [210].

Sequencing of the first human genomes revealed that most genetic variation is derived from single nucleotide variants (SNVs), although insertions and deletions (indels) account for the majority of the variant nucleotides [11]. The increased accessibility of DNA sequencing has contributed to individual efforts from a range of distinct populations. To date, individual genomes from American [11, 12], Han Chinese [13], Russian [14], Khoisan [15], Bantu [15], Japanese [16], German [17], Gujarati Indian [18], Estonian [19], Pakistani [20] and Mongolian [21] populations have been sequenced and analyzed, among many others [1].

Larger-scale efforts to characterize human genetic variation have demonstrated that individuals from different populations carry particular combinations of rare and low-frequency variants. The 1000 Genomes Project Consortium has estimated that 86% of all variants are confined to a single continental group and that about 10% of variants observed in a population are private to that population [1]. Population-specific variants have the potential to be of both functional and biomedical importance [7, 2224]. Furthermore, evidence of biologically meaningful population-specific variation [25] emphasizes the need for ethnically relevant reference genomes, as has been performed, for example, for the Korean population [26]. Although we are not claiming to have introduced a new reference genome here, it is nevertheless important to expand our sequencing efforts across diverse populations, particularly those that have not been previously studied [10, 27].

In this paper, we describe the sequencing of the first genome of an individual of Serbian origin, a member of a relatively small population in Central to Southeastern Europe. We identify tens of thousands of novel genetic variants in this individual, more than a hundred of which map to protein-coding regions and several hundred of which reside in close proximity to gene coding regions. The extent of observed genetic variation allowed comparisons with extant European populations and reaffirms support for the hypothesis of close correspondence between genetic and geographic distances [2]. These results contribute to ongoing efforts to understand human genetic variation and its geographic distribution, as well as placing the Serbian genome within the context of the broader European population structure. Testing for Neanderthal introgression in the genome, we find evidence to suggest gene flow from Neanderthal to an ancestral pan-European genome, with the Serbian genome being placed within the range of other European populations. After variant annotation, we assess the burden of potentially pathogenic variation present in this genome and identify variants of putative clinical and pharmacogenetic relevance. Finally, we draw conclusions pertaining to the phenotypic consequences and biomedical interpretation of individually sequenced genomes.

Materials and methods

Donor information

The individual whose genome was sequenced and analyzed is a male of Serbian descent. The data, both derived and raw, are publicly available through the Personal Genome Project website [28], participant ID: hu3BDC4B.

Sample collection and DNA sequencing

Two milliliters of saliva were self-collected by the donor and stored using the DNA Genotek Oragene DISCOVER (OGR-500) sample collection kit. Extraction of DNA from the sample and subsequent sequencing were performed at the BGI (Shenzhen, China) on an Illumina HiSeq 2000 sequencer, using standard protocols. To minimize the likelihood of systematic bias in sampling, two libraries were prepared with an insert size of 500 bp each, with paired-end reads of length 90 bp. Sequencing was then carried out in four lanes for each library to ensure at least 30-fold coverage.

Read mapping and variant calling

Single Nucleotide Variants (SNVs) and indels were called using four different pipelines through a combination of two read mappers and two variant callers. The GRCh37 human genome was used as the reference genome to map the paired-end reads. The two read mappers used were BWA-MEM [29] and Bowtie2 [30]. The two variant callers were GATK [31] and Platypus [32]. The GATK pipeline included additional read and variant processing steps such as duplicate removal using Picard tools [33], base quality score recalibration, indel realignment, and genotyping and variant quality score recalibration using GATK, all used according to GATK best practice recommendations [34, 35].

As described later in the Results, variants identified using the BWA+ GATK pipeline were used for all downstream analysis. Variants in the intersection of all four pipelines (two read mappers and two variant callers) were considered to be confidently identified, where the intersection is defined as variant calls for which the chromosome, position, reference, and alternate fields in the VCF files were identical. All variant calls were subsequently annotated with information from NCBI RefSeq using ANNOVAR [36]. We estimated the amount of novel variation expected to be observed from the first individual in a previously uncharacterized population utilizing the 1000 Genomes Project Phase 3 VCF files [37]. To do this, we carried out a leave-one-population-out procedure; i.e., we excluded one of the 26 populations at a time and for each individual in the excluded population, calculated the fraction of variants not seen in any of the individuals from the remaining 25 populations. The calculated fractions of novel variants were used to understand the expected novelty when sequencing an individual from a new population, given a sample of a particular size of previously sequenced individuals from different populations.

Structural variants (SVs) were called using Structural Variation Engine (SVE) and FusorSV [38]. SVE is an execution engine for an ensemble of SV calling algorithms containing BreakDancer [39], BreakSeq2 [40], cnMOPS [41], CNVnator [42], DELLY [43], GenomeSTRiP [44, 45], Hydra [46], and LUMPY [47]. The Docker image of SVE was used to run all the stages with default parameters. All but GenomeSTRiP completed without errors. The Docker image of FusorSV was then used to merge the results from the remaining seven SV callers, using the default fusion model. SVint [48] was used to subsequently annotate the structural variants. Scripts and documentation for parameters used to run all the pipelines described in this study were added to the Personal Genome Project website, participant ID hu3BDC4B.

Principal component analysis

Principal component analysis (PCA) was carried out using the smartpca program from EIGENSOFT (v6.0.1;, on the Serbian genome combined with the SNV data (600,841 loci) from Lazaridis et al. [3]. Only the subset of European individuals from their curated fully public dataset was used, reducing the original set of 1,964 individuals to 260. A projection to the first two principal components was used to establish the correspondence between genetic and geographic distance in our results.

Neanderthal introgression

To test for Neanderthal introgression in the Serbian genome, we computed D-statistics [49, 50] using this genome and the dataset from Lazaridis et al. [9]. This dataset includes 294 ancient individuals (only one of which was used here) and a diverse set of 2,068 present-day humans, genotyped on the Affymetrix Human Origins array. Both the archaic and modern genotype data were provided in the PACKEDANCESTRYMAP format, and were combined using the mergeit program from EIGENSOFT (v6.1.2; The merged dataset, in total, contains 2,362 samples genotyped at 621,799 SNV loci. Upon request, we completed the consent form and obtained approval from David Reich’s laboratory before using this dataset. Some individuals from the study of Lazaridis et al. [9] could not be included due to consent issues relating to data distribution.

We next genotyped the Serbian genome against these predefined SNVs using GATK HaplotypeCaller and following the GATK best practices recommendations [34, 35]. We converted the resulting VCF files to the EIGENSTRAT format using VCFtools (v0.1.12a, [51]), and integrated the Serbian genotype with the modern and ancient datasets. Finally, we ran qpDstat from AdmixTools (default setting, v701) to calculate D-statistics and to test for Neanderthal gene flow into the Serbian genome [50].

Burden of pathogenic variation

Variants of putative clinical significance were identified using genotype-phenotype databases as well as computational variant-effect prediction. Manually curated genotype-phenotype databases, such as the Human Gene Mutation Database (HGMD) [52], ClinVar [53] and PharmGKB [54], annotate variants with a known relationship to phenotype [52, 55]. Clinical Annotations from PharmGKB were compared against dbSNP v142 rsIDs [56] obtained using the script in ANNOVAR and avsnp142. Variants identified by GATK were compared against HGMD and ClinVar to identify potentially disease-causing and disease-associated mutations.

All variants in protein-coding regions were extracted and inputted to the MutPred suite of tools [5760]. The remaining variation observed in the proband was interrogated using CADD [61]. For disease and gene ontology associations, the hypergeometric test in WebGestalt was used with Benjamini-Hochberg correction for multiple hypothesis-testing [62]. The background set that was used for these analyses included all protein-coding genes from the human reference genome. For the significance of an ontology term to be confirmed, at least five genes were required to be associated with it.


Effect of genotyping software

The choice of computational tools and their parameters in processing raw sequencing reads can significantly impact the resulting genome and the entirety of subsequent analysis [63, 64]. To understand the uncertainty of variant identification in our subject, we evaluated two different read mappers, BWA-MEM [29] and Bowtie2 [30], and two different variant callers, GATK [31] and Platypus [32].

The results from four different platforms are compared and contrasted in Fig 1. The SNV calling shows good concordance between both read mappers and variant callers, with a large proportion of variants identified by either platform being identified by all platforms. Using the BWA-MEM mapper (which we refer to simply as “BWA” from now on), for example, 2,991,390/3,280,434 = 91.2% of SNVs identified by GATK were also identified by Platypus and 89.1% of SNVs identified by Platypus were also identified by GATK (Fig 1). Indel calling, on the other hand, is less reliable, with 401,082/627,519 = 63.9% variants identified by GATK also identified by Platypus and only 66.7% of variants identified by Platypus being also identified by GATK. The influence of read mappers was markedly lower; i.e., using the GATK variant caller, we found that 95.1% of SNVs and 89.3% of indels identified with BWA were also identified with Bowtie2, and 98.3% SNVs and of 97.6% of indels identified with Bowtie2 were also identified with BWA. Smaller percentages of overlap were observed for Platypus. Based on the results observed in this work (Table A in S1 File) and the extent of usage of these tools in resequencing human genomes, we selected BWA+ GATK as our main platform.

Fig 1. Venn diagrams showing the total numbers of identified variants using two read mappers (BWA [29], Bowtie2 [30]) and two variant callers (GATK [31], Platypus [32]).

Identification of genetic variants

The genome of a Serbian individual was sequenced according to the protocols described in the Materials and Methods, with all 22 autosomes having similar coverage and the X and Y chromosome having approximately half this coverage. The genome sequencing and mapping achieved an average read depth of 34.7, with 98.3% of GRCh37 reference bases having coverage of 10-fold or more and 89.4% having coverage of 20-fold or more. The number of zero-depth positions were 7,649,443 (0.3%). The coverage distribution is shown in the Supporting Information (S1 Fig).

Using the BWA+ GATK pipeline, we identified a total of 3,908,814 variants (83.9% SNVs, 16.1% indels; Fig 1) in the Serbian genome, of which 2,195,638 (56.2%) were heterozygous with one non-reference allele, 23,095 (0.6%) were heterozygous with two non-reference alleles, and 1,690,081 (43.2%) were homozygous for a non-reference allele. The reported variants passed all quality filters of GATK (marked as “PASS”) and were subsequently mapped to GRCh37 human reference genomic regions using ANNOVAR [36]. It is important to mention that ANNOVAR considers all heterozygous positions with both alternative alleles as two different variants. Mechanisms by which heterozygous alternative alleles can arise include sequencing errors and highly variable sites, some of which are tri-allelic because of rare mutational events [65]. Therefore, the resulting genome contains a total of 3,931,909 variants, of which 2,940,042 (74.8%) were identified by all four platforms and are considered to be confident identifications. Unsurprisingly, the majority of identified variants were found to reside in the more expansive and less evolutionarily constrained intergenic and intronic regions (Table 1).

Table 1. Summary of identified variants using BWA+ GATK.

Variants not present in gnomAD [66] are listed as novel and variants identified by all four genotyping platforms are listed as confident.

To identify novel variation, we compared the identified variants against the Genome Aggregation Database (gnomAD) [66]. We found that 1.5% (60,153) all variants and 0.4% (12,439) of confident variants were not present in gnomAD. We shall refer to these variants as “novel” and “confident novel” variants, respectively. The breakdown of all variants and novel variants with respect to genomic location is shown in Tables 1 and 2. The percentage of novel variants varied across categories, comprising 0.9% (80) of nonsynonymous variants, 0.4% of synonymous variants, 0.7% (145) of exonic variants, 1.5% (20,531) of intronic variants, and 1.6% (34,779) of intergenic variants. We found that 45.0% (9,328/20,739) of the exonic variants were nonsynonymous, whereas 50.1% (10,381/20,739) were synonymous. Similar fractions were observed for the confident variants (44.1% vs. 52.3%). Of the 3,871,756 GATK variants that are also observed in the gnomAD database, 3,805,264 (95%) of these variants are annotated to have allele frequency greater than 1% in gnomAD and 3,676,638 (95%) with allele frequency greater than 5%. The proportion of novel variation in the Serbian individual is at the lower end of the distribution compared to 1000 Genomes Project participants (S6 Fig), consistent with a significantly larger size of gnomAD that currently integrates 15,708 whole-genomes and 125,748 exomes.

Table 2. Summary of identified exonic variants using BWA+GATK.

Variants not present in gnomAD [66] are listed as novel and variants identified by all four platforms are listed as confident.

Using SVE and FusorSV, we identified 848 deletions and 3 duplications, which include the most confident calls generated by FusorSV after merging call-sets from seven different SV-callers using the default fusion model. The numbers of structural variants called by individual SV-callers are reported in (Table B in S1 File). The deletions in the Serbian genome have a length distribution (S7 Fig) similar to the deletions in the 27 deep-coverage samples of the 1000 Genomes Project reported by FusorSV [38]. The lengths of the three duplications are 313101, 362391 and 471821 bp. We used SVint to annotate the functional impact of the structural variants. The genes that overlap with the identified structural variants are listed in S1 File Tables C and D.

Genetic variation and geographic distance

The projection of the Serbian individual to the first and second principal components against European groups from [3] confirms that individuals from the same geographic region cluster together (Fig 2). We clearly distinguish clusters of major populations composed of individuals from the same region, approximately mirroring a map of Europe. The PCA plot demonstrates that the genetic ancestry of the Serbian individual analyzed in the present study corresponds to its geographic distance from other populations. It is positioned in close proximity of the Croatian, Bulgarian, and Hungarian populations.

Fig 2. Principal component analysis (PCA) plot showing the proximity of the genome sequenced in this study to other European genomes.

As observed in previous studies [2, 3], genomic distance correlates with geographic distance.

A somewhat surprising finding is the similarity of distances between the Serbian individual and other mostly Slavic populations (Russian, Belarus, Ukrainian) relative to distances to various Central, Western, and Southern European groups (Czech, French, English, Albanian, Greek). The average Euclidean distance and variance between the Serbian individual and each of the available populations in the two-dimensional space of major PCA components is as follows: Croatian (0.016826 ± 0.010526), Bulgarian (0.033603 ± 0.000225), Hungarian (0.037121 ± 0.000177), Czech (0.053687 ± 0.000033), Albanian (0.058875 ± 0.000117), Ukrainian (0.064328 ± 0.000062), Belarusian (0.069803 ± 0.000043), Greek (0.071108 ± 0.00062), Tuscan (0.0736441 ± 0.000028), French (0.083077 ± 0.000159), English (0.084570 ± 0.000142), Norwegian (0.092721 ± 0.00088), Russian (0.095968 ± 0.000079), Estonian (0.098421 ± 0.000046), Finnish (0.108523 ± 0.000154), Sicilian (0.120370 ± 0.000481), Spanish (0.134602 ± 0.000776), Ashkenazi (0.156692 ± 0.000538). The three closest individuals to the Serbian genome were of Croatian ancestry (0.0038, 0.0046, and 0.0108).

We note that combining the Serbian individual with the set of 260 European individuals from Lazaridis et al. [3] caused 50 formerly biallelic sites to become triallelic (no monoallelic sites became triallelic). The triallelic sites were removed from the analysis, leaving 600,791 sites in the analysis. The smartpca program was applied to the 261-by-600,791 genotype matrix.

Gene flow with Neanderthals

Comparisons between Neanderthals and modern humans have previously revealed evidence of gene flow from Neanderthals to Europeans [49, 50, 67, 68]. To test whether the Serbian genome shares an excess of alleles with the Neanderthal genome, we integrated the Serbian genotype with a published panel of ancient and modern humans (Materials and Methods). We calculated D-statistics as a formal test for gene flow based on a four-taxon phylogeny, D(P1, P2, P3, O), where Pi (i ∈ {1, 2, 3}) are populations and O is an outgroup. Given a scenario where gene flow is absent, the derived alleles of P3 are expected, with equal likelihood, to match those of P1 and P2; i.e., D = 0. Alternatively, either P1 or P2 could share alleles with P3 more often than not, in which case D deviates from zero.

We computed D(Yoruba, Serbian, Altai, Chimpanzee) for testing for gene flow between Neanderthals (“Altai”) and the given Serbian genome. We expected a positive D value, given previous evidence that Neanderthals exchanged more alleles with Europeans than with Africans. The test returned a D value of 0.0241 ± 0.004476, which significantly deviated from zero (Z-score = 5.39; Table 3), suggesting gene flow between Neanderthal and the lineage leading to the Serbian genome. To validate this result, we also ran the test for other European populations (Table 3). D-statistics calculated for Croatian, French, Greek and Russian genomes were comparable to our result, all falling within the expected range of values reported in previous studies [49, 67, 68].

Table 3. Testing gene flow with Neanderthals.

The results show the D-statistic (D), its standard error (SE) and Z-score (Z) for the test using the set of populations P1, P2, and P3, with Chimpanzee as an outgroup (O). The last two columns show ABBA vs. BABA counts over the four genomes (P1, P2, P3, O).

We further attempted to ensure that the calculated D-statistics were unbiased. To do this, we repeated the analysis by replacing Yoruba with Mbuti, as some of the Yoruba samples could have had some recent European admixture. The calculation for D(Mbuti, Serbian, Altai, Chimpanzee) yielded a D value of 0.0186 ± 0.004763 (Z-score = 3.99; Table 3), consistent with our results using the Yoruba samples. We next checked whether the Serbian individual has reference biases in genotyping that could have inflated the D value. We performed D-statistics tests in the form of D(other European population, Serbian, Mbuti, hg19ref) and chose Croatian, French, Greek and Russian as the “other European population”. We obtained no test results indicating the bias of Serbian genotypes toward the reference (Croatian: 0.0054 ± 0.004183; French: 0.0038 ± 0.004078; Greek: 0.0090 ± 0.004182; Russian: 0.0074 ± 0.004192).

Analysis of medically relevant variants

The sequenced genome contains 2,343 genetic variants that are present in HGMD by virtue of their having been previously associated with a risk of disease; the proportions of variants within each effect category are shown in Table 4. Several homozygous variants, manually annotated as disease-causing (DM) are observed in the genome, shown in Table 5. Of these, one is a youth-onset phenotype, Factor XIII deficiency, associated with homozygosity for the disease-causing allele (NM_000129.3:c.-19+12C>A) in the proband’s genome. The disease phenotypes associated with these homozygous mutations typically become apparent in childhood, and therefore their occurrence in a healthy adult is indicative of variable penetrance. The other homozygous disease-causing variants result in phenotypes that have not yet been observed in either the individual or in their family history; perhaps reflecting either low expressivity or late-onset. Observed heterozygous disease-causing mutations are primarily childhood-onset without presentation in the individual, although they may represent recessive conditions; thus, their failure to manifest may not necessarily be indicative of poor reporting or curation quality. Next, we identified several variants with pathogenic annotation in the ClinVar database, an open-access alternative to HGMD [53]. These variants are either low-confidence or without known family history; more details are available in the Supporting Information (S1 File).

Table 4. Amount of disease-causing and potentially disease-relevant variation in the Serbian genome.

Identified variants were searched against HGMD and broken down into the phenotypic categories of HGMD. Variants were broken down into exonic and noncoding as well as homozygous and heterozygous.

Table 5. Disease-causing variants observed in the proband.

The table summarizes the analysis of five homozygous variants form the sequenced genome that are listed by HGMD as disease-causing.

We also identified several variants of potential pharmacogenetic relevance using PharmGKB. Variants in PharmGKB are assigned Clinical Annotation Levels of Evidence from variants with preliminary evidence (Level 4) to high confidence variant-drug combinations with medically endorsed integration into health systems (Level A1). The genome contains a single variant with a high-confidence annotation (Level 1B): rs2228001, associated with toxicity and adverse drug reaction to cisplatin, a chemotherapeutic agent. A further 17 variants were annotated with moderate evidence to impact the dosage, efficacy, metabolism and/or toxicity of drugs for diverse phenotypes including chronic hepatitis C, organ transplantation rejection, glaucoma, depression, schizophrenia, asthma, epilepsy and HIV infections, as well as several chemotherapy drugs.

Pathogenicity prediction.

In addition to known disease-associated variants, we identified missense variants predicted to be pathogenic by MutPred2 [57]. Of the 11,206 missense variants called by GATK, 9,329 passed all quality filters (annotated as ‘PASS’). Of these, 9,305 variants were unambiguously mapped to the correct protein isoforms and hence were amenable for prediction by MutPred2. Based on a score threshold of 0.8 (estimated 5% false positive rate), 95 missense variants were predicted to be ‘pathogenic.’

Of these, 14 variants were found in the homozygous state and 81 were found in the heterozygous state. Genes for these variants were enriched in GO terms related to peptidase activity S8 Fig). A similar analysis for disease associations revealed that the subject may be at risk for cardiovascular disorders (Table I in S1 File).

Next, we applied computational predictors on the remaining protein coding variation with the MutPred family of tools. First, we assessed the pathogenicity of 180 nonsense and frameshifting insertion and deletion variants with MutPred-LOF [58]. From this set, we identified a total of 7 variants with scores above the 0.5 score threshold (corresponding to a 5% false positive rate) (Table E in S1 File). Next, we assessed 279 non-frameshifting insertion and deletion variants with MutPred-Indel and identified 12 variants described in (Table F in S1 File. Finally, we assessed the pathogenicity of the 90 SNV splicing variants with MutPred Splice [59]. Of these, 28 of the variants scored at least 0.6 and were therefore classified as a “Splice Affecting Variant” by MutPred Splice. One of these variants is predicted to cause loss of natural 3’ splice sites, two variants are predicted to interrupt cryptic 3’ splice sites, and three variants are predicted to disrupt cryptic 5’ splice sites, described in the Supporting Information (Table G in S1 File).

To ensure assessment of the complete variome of the proband, we utilized CADD v1.3 [61] to evaluate all noncoding variants. To do this, we utilized a scaled C-score cutoff of 20 to identify the 1% most damaging variants. In total, we found 16 UTR variants, 1,630 intronic variants, 3,911 intergenic variants, 80 regulatory variants, 839/533 upstream/downstream variants, and 9 variants annotated as “noncoding_change.” All of these were predicted to be deleterious. The noncoding variants with the highest C-scores are described in the Supporting Information (Table H in S1 File).


This work describes the first whole-genome sequencing of a Serbian individual. Ancestry analysis positioned the Serbian individual in closest proximity to the Croatian population, consistent with its Southern Slavic ancestry [69]. Our analyses further support the hypothesis of gene flow between Neanderthal and pan-European ancestral populations, with the level of introgression into the Serbian genome being within the range observed in other European populations. Previous genetic studies involving Slavic populations employed mitochondrial, Y-chromosome and SNV-panel data to investigate the relationship between geographic, genetic and linguistic distances [69, 70]. Consistent with this work, our analyses expand the scope beyond Slavic populations and further contribute to the understanding of human genetic variation and its geographic distribution.

In contrast to studies using genotyping arrays [2, 3, 69, 70], the availability of whole-genome sequences presents the opportunity for a high-resolution individualized analysis. To this end, we found that the sequenced genome contains a significant number of previously unobserved variants, which emphasizes the importance of continued sequencing of a large number of individuals, especially from previously uncharacterized ethnic groups. Subsequent sequencing of other Serbian individuals could provide further insight into these novel variants; e.g., whether they are private to the population or to the individual. Such results would in turn contribute important information regarding variants that are currently considered to be rare, with implications for improved variant interpretation. Furthermore, new algorithms and reduced sequencing costs will have the potential to provide higher-quality analysis of structural variants. Our analysis also found a number of variants of clinical and pharmacogenomic significance that might extend beyond an individual’s disease risks to facilitate possible future medical interventions although conclusions are limited without validation and knowledge of allele frequencies in the Serbian population [71, 72]. Such variants might contribute to better outcomes in studies of disease penetrance, mechanistic understanding of population risks, and database curation.

Recent advances in high-throughput sequencing and reduced costs of genotyping have greatly facilitated whole-genome data generation, and have become key to understanding both human phenotypes and early human history [2, 3]. However, modern technology and cost structure continue to pose challenges in determining and interpreting one’s genome [73]. Variation in read mapping and variant calling contribute to the uncertainty of interpretation with different software packages, identifying different sets of variants. We found that inter-software discrepancies ranged from relatively small for SNVs to considerable for insertions and deletions, especially for structural variants. Therefore, variant and genome interpretation demand caution, since thousands of SNVs and tens of thousands of indels may simply constitute genotyping errors [74, 75].

It is worth mentioning that in addition to the technical aspects of genome sequencing, an important aspect of genome interpretation concerns psychosocial uncertainty due to phenotypic and privacy-associated risks [76]. The geographic distance analysis in this study has provided evidence that supports the individual’s own sense of Serbian ancestry; however, the finding of multiple predicted youth-onset pathogenic mutations in a healthy individual provides cautionary lessons for predictive medicine.

Supporting information

S2 Fig. Read depth across SNV and insertion/deletion variants across the four pipelines.


S3 Fig. Size of insertion/deletion variants across the four pipelines.


S4 Fig. Number of heterozygous and homozygous variants across the four pipelines.


S5 Fig. Size of insertion/deletion variants in the Serbian genome.


S6 Fig. Proportion of novel variants in thousand genomes project participants.


S7 Fig. Length distribution of deletions called by FusorSV.


S8 Fig. GO terms enriched in the set of 81 genes that harbored the 95 missense variants predicted to be pathogenic.


S1 File. Annotation descriptions and tables of variants scored as pathogenic.



We thank Dr. Chuanchao Wang for his suggestions and discussion on testing for Neanderthal admixture in the Serbian genome. We also thank the Associate Editor and anonymous reviewers for their insightful comments that have improved the precision and quality of the paper. This work was supported by the Center for Bioinformatics Research at Indiana University, and was originally carried out as part of a class project for INFO-I590.


  1. 1. Consortium GP, Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, et al. A global reference for human genetic variation. Nature. 2015;526(7571):68–74.
  2. 2. Novembre J, Johnson T, Bryc K, Kutalik Z, Boyko AR, Auton A, et al. Genes mirror geography within Europe. Nature. 2008;456(7218):98–101. pmid:18758442
  3. 3. Lazaridis I, Patterson N, Mittnik A, Renaud G, Mallick S, Kirsanow K, et al. Ancient human genomes suggest three ancestral populations for present-day Europeans. Nature. 2014;513(7518):409–413. pmid:25230663
  4. 4. Pagani L, Lawson DJ, Jagoda E, Morseburg A, Eriksson A, Mitt M, et al. Genomic analyses inform on migration events during the peopling of Eurasia. Nature. 2016;538(7624):238–242. pmid:27654910
  5. 5. Montinaro F, Busby GB, Gonzalez-Santos M, Oosthuitzen O, Oosthuitzen E, Anagnostou P, et al. Complex ancient genetic structure and cultural transitions in southern African populations. Genetics. 2017;205(1):303–316. pmid:27838627
  6. 6. House GL, Hahn MW. Evaluating methods to visualize patterns of genetic differentiation on a landscape. Mol Ecol Resour. 2018;18(3):448–460. pmid:29282875
  7. 7. Burchard EG, Ziv E, Coyle N, Gomez SL, Tang H, Karter AJ, et al. The importance of race and ethnic background in biomedical research and clinical practice. N Engl J Med. 2003;348(12):1170–1175. pmid:12646676
  8. 8. Gibson G, Muse SV. A primer of genome science. 3rd ed. Sunderland, Mass.: Sinauer Associates; 2009.
  9. 9. Lazaridis I, Nadel D, Rollefson G, Merrett DC, Rohland N, Mallick S, et al. Genomic insights into the origin of farming in the ancient Near East. Nature. 2016;536(7617):419–424. pmid:27459054
  10. 10. Manrai AK, Funke BH, Rehm HL, Olesen MS, Maron BA, Szolovits P, et al. Genetic misdiagnoses and the potential for health disparities. N Engl J Med. 2016;375(7):655–665. pmid:27532831
  11. 11. Levy S, Sutton G, Ng PC, Feuk L, Halpern AL, Walenz BP, et al. The diploid genome sequence of an individual human. PLoS Biol. 2007;5(10):e254. pmid:17803354
  12. 12. Wheeler DA, Srinivasan M, Egholm M, Shen Y, Chen L, McGuire A, et al. The complete genome of an individual by massively parallel DNA sequencing. Nature. 2008;452(7189):872–876. pmid:18421352
  13. 13. Wang J, Wang W, Li R, Li Y, Tian G, Goodman L, et al. The diploid genome sequence of an Asian individual. Nature. 2008;456(7218):60–65. pmid:18987735
  14. 14. Chekanov NN, Boulygina ES, Beletskiy AV, Prokhortchouk EB, Skryabin KG. Individual genome of the Russian male: SNP calling and a de novo assembly of unmapped reads. Acta Naturae. 2010;2(3):122–126. pmid:22649659
  15. 15. Schuster SC, Miller W, Ratan A, Tomsho LP, Giardine B, Kasson LR, et al. Complete Khoisan and Bantu genomes from southern Africa. Nature. 2010;463(7283):943–947. pmid:20164927
  16. 16. Fujimoto A, Nakagawa H, Hosono N, Nakano K, Abe T, Boroevich KA, et al. Whole-genome sequencing and comprehensive variant analysis of a Japanese individual using massively parallel sequencing. Nat Genet. 2010;42(11):931–936. pmid:20972442
  17. 17. Suk EK, McEwen GK, Duitama J, Nowick K, Schulz S, Palczewski S, et al. A comprehensively molecular haplotype-resolved genome of a European individual. Genome Res. 2011;21(10):1672–1685. pmid:21813624
  18. 18. Kitzman JO, Mackenzie AP, Adey A, Hiatt JB, Patwardhan RP, Sudmant PH, et al. Haplotype-resolved genome sequencing of a Gujarati Indian individual. Nat Biotechnol. 2011;29(1):59–63. pmid:21170042
  19. 19. Lilleoja R, Sarapik A, Reimann E, Reemann P, Jaakma U, Vasar E, et al. Sequencing and annotated analysis of an Estonian human genome. Gene. 2012;493(1):69–76. pmid:22138481
  20. 20. Azim MK, Yang C, Yan Z, Choudhary MI, Khan A, Sun X, et al. Complete genome sequencing and variant analysis of a Pakistani individual. J Hum Genet. 2013;58(9):622–626. pmid:23842039
  21. 21. Bai H, Guo X, Zhang D, Narisu N, Bu J, Jirimutu J, et al. The genome of a Mongolian individual reveals the genetic imprints of Mongolians on modern human populations. Genome Biol Evol. 2014;6(12):3122–3136. pmid:25377941
  22. 22. Nakatsuka N, Moorjani P, Rai N, Sarkar B, Tandon A, Patterson N, et al. The promise of discovering population-specific disease-associated genes in South Asia. Nat Genet. 2017;49(9):1403–1407. pmid:28714977
  23. 23. Smyth N, Ramsay M, Raal FJ. Population specific genetic heterogeneity of familial hypercholesterolemia in South Africa. Curr Opin Lipidol. 2018;29(2):72–79. pmid:29369830
  24. 24. Lencz T, Yu J, Palmer C, Carmi S, Ben-Avraham D, Barzilai N, et al. High-depth whole genome sequencing of an Ashkenazi Jewish reference panel: enhancing sensitivity, accuracy, and imputation. Hum Genet. 2018;137(4):343–355. pmid:29705978
  25. 25. Guda K, Veigl ML, Varadan V, Nosrati A, Ravi L, Lutterbaugh J, et al. Novel recurrently mutated genes in African American colon cancers. Proc Natl Acad Sci U S A. 2015;112(4):1149–1154. pmid:25583493
  26. 26. Cho YS, Kim H, Kim HM, Jho S, Jun J, Lee YJ, et al. An ethnically relevant consensus Korean reference genome is a step towards personal reference genomes. Nat Commun. 2016;7:13637. pmid:27882922
  27. 27. Popejoy AB, Fullerton SM. Genomics is failing on diversity. Nature. 2016;538(7624):161–164. pmid:27734877
  28. 28. Ball MP, Thakuria JV, Zaranek AW, Clegg T, Rosenbaum AM, Wu X, et al. A public resource facilitating clinical use of genomes. Proc Natl Acad Sci U S A. 2012;109(30):11920–11927. pmid:22797899
  29. 29. Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics. 2009;25(14):1754–1760. pmid:19451168
  30. 30. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods. 2012;9(4):357–359. pmid:22388286
  31. 31. McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 2010;20(9):1297–1303. pmid:20644199
  32. 32. Rimmer A, Phan H, Mathieson I, Iqbal Z, Twigg SRF, Consortium W, et al. Integrating mapping-, assembly- and haplotype-based approaches for calling variants in clinical sequencing applications. Nat Genet. 2014;46(8):912–918. pmid:25017105
  33. 33. Picard Tools;.
  34. 34. Van der Auwera GA, Carneiro MO, Hartl C, Poplin R, Del Angel G, Levy-Moonshine A, et al. From FastQ data to high confidence variant calls: the Genome Analysis Toolkit best practices pipeline. Curr Protoc Bioinformatics. 2013;43:11.10.1–11.10.33.
  35. 35. DePristo MA, Banks E, Poplin R, Garimella KV, Maguire JR, Hartl C, et al. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nat Genet. 2011;43(5):491–498. pmid:21478889
  36. 36. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res. 2010;38(16):e164. pmid:20601685
  37. 37. Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, Korbel JO, et al. A global reference for human genetic variation. Nature. 2015;526(7571):68–74. pmid:26432245
  38. 38. Becker T, Lee WP, Leone J, Zhu Q, Zhang C, Liu S, et al. FusorSV: an algorithm for optimally combining data from multiple structural variation detection methods. Genome Biol. 2018;19(1):38. pmid:29559002
  39. 39. Fan X, Abbott TE, Larson D, Chen K. BreakDancer: identification of genomic structural variation from paired-end read mapping. Curr Protoc Bioinformatics. 2014;45(1):15.6.1–15.6.11.
  40. 40. Lam HY, Mu XJ, Stutz AM, Tanzer A, Cayting PD, Snyder M. Nucleotide-resolution analysis of structural variants using BreakSeq and a breakpoint library. Nat Biotechnol. 2010;28(1):47–55. pmid:20037582
  41. 41. Klambauer G, Schwarzbauer K, Mayr A, Clevert DA, Mitterecker A, Bodenhofer U. cn.MOPS: mixture of Poissons for discovering copy number variations in next-generation sequencing data with a low false discovery rate. Nucleic Acids Res. 2012;40(9):e69. pmid:22302147
  42. 42. Abyzov A, Urban AE, Snyder M, Gerstein M. CNVnator: an approach to discover, genotype, and characterize typical and atypical CNVs from family and population genome sequencing. Genome Res. 2011;21(6):974–984. pmid:21324876
  43. 43. Rausch T, Zichner T, Schlattl A, Stutz AM, Benes V, Korbel JO. DELLY: structural variant discovery by integrated paired-end and split-read analysis. Bioinformatics. 2012;28(18):i333–i339. pmid:22962449
  44. 44. Handsaker RE, Korn JM, Nemesh J, McCarroll SA. Discovery and genotyping of genome structural polymorphism by sequencing on a population scale. Nat Genet. 2011;43(3):269–276. pmid:21317889
  45. 45. Handsaker RE, Van Doren V, Berman JR, Genovese G, Kashin S, Boettger LM. Large multiallelic copy number variations in humans. Nat Genet. 2015;47(3):296–303. pmid:25621458
  46. 46. Lindberg MR, Hall IM, Quinlan AR. Population-based structural variation discovery with Hydra-Multi. Bioinformatics. 2015;31(8):1286–1289. pmid:25527832
  47. 47. Layer RM, Chiang C, Quinlan AR, Hall IM. LUMPY: a probabilistic framework for structural variant discovery. Genome Biol. 2014;15(6):R84. pmid:24970577
  48. 48. SVint, a light-weight tool for annotating structure variants located outside the coding genome;.
  49. 49. Green RE, Krause J, Briggs AW, Maricic T, Stenzel U, Kircher M, et al. A draft sequence of the Neandertal genome. Science. 2010;328(5979):710–722.
  50. 50. Slatkin M, Racimo F. Ancient DNA and human history. Proc Natl Acad Sci U S A. 2016;113(23):6380–6387. pmid:27274045
  51. 51. Danecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA, et al. The variant call format and VCFtools. Bioinformatics. 2011;27(15):2156–2158. pmid:21653522
  52. 52. Stenson PD, Mort M, Ball EV, Evans K, Hayden M, Heywood S, et al. The Human Gene Mutation Database: towards a comprehensive repository of inherited mutation data for medical research, genetic diagnosis and next-generation sequencing studies. Hum Genet. 2017;136(6):665–677. pmid:28349240
  53. 53. Landrum MJ, Lee JM, M B, Brown G, Chao C, Chitipiralla S, et al. ClinVar: public archive of interpretations of clinically relevant variants. Nucleic Acids Res. 2016;44(D1):D862–D868. pmid:26582918
  54. 54. Klein TE, Chang JT, Cho MK, Easton KL, Fergerson R, Hewett M, et al. Integrating genotype and phenotype information: an overview of the PharmGKB project. Pharmacogenetics Research Network and Knowledge Base. Pharmacogenomics J. 2001;1(3):167–170. pmid:11908751
  55. 55. Whirl-Carrillo M, McDonagh EM, Hebert JM, Gong L, Sangkuhl K, Thorn CF, et al. Pharmacogenomics knowledge for personalized medicine. Clin Pharmacol Ther. 2012;92(4):414–417. pmid:22992668
  56. 56. Sherry ST, Ward MH, Kholodov M, Baker J, Phan L, Smigielski EM, et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res. 2001;29(1):308–311. pmid:11125122
  57. 57. Pejaver V, Urresti J, Lugo-Martinez J, Pagel KA, Lin GN, Nam HJ, et al. MutPred2: inferring the molecular and phenotypic impact of amino acid variants. bioRxiv 134981. 2017;.
  58. 58. Pagel KA, Pejaver V, Lin GN, Nam HJ, Mort M, Cooper DN, et al. When loss-of-function is loss of function: assessing mutational signatures and impact of loss-of-function genetic variants. Bioinformatics. 2017;33(14):i389–i398. pmid:28882004
  59. 59. Mort M, Sterne-Weiler T, Li B, Ball EV, Cooper DN, Radivojac P, et al. MutPred Splice: machine learning-based prediction of exonic variants that disrupt splicing. Genome Biol. 2014;15(1):R19. pmid:24451234
  60. 60. Pagel KA, Mort M, Cooper DN, Mooney SD, Radivojac P. Pathogenicity and functional effects of non-frameshifting insertion/deletion variation in the human genome. Unpublished;.
  61. 61. 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):310–315. pmid:24487276
  62. 62. Wang J, Duncan D, Shi Z, Zhang B. WEB-based GEne SeT AnaLysis Toolkit (WebGestalt): update 2013. Nucleic Acids Res. 2013;41(Web Server issue):77–83.
  63. 63. Hwang S, Kim E, Lee I, Marcotte EM. Systematic comparison of variant calling pipelines using gold standard personal exome variants. Sci Rep. 2015;5:17875. pmid:26639839
  64. 64. Zook J, McDaniel J, Parikh H, Heaton H, Irvine SA, Trigg L, et al. Reproducible integration of multiple sequencing datasets to form high-confidence SNP, indel, and reference calls for five human genome reference materials. bioRxiv 281006. 2018;.
  65. 65. Hodgkinson A, Eyre-Walker A. Human triallelic sites: evidence for a new mutational mechanism? Genetics. 2010;184(1):233–241. pmid:19884308
  66. 66. Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature. 2016;536(7616):285–291. pmid:27535533
  67. 67. Prufer K, Racimo F, Patterson N, Jay F, Sankararaman S, Sawyer S, et al. The complete genome sequence of a Neanderthal from the Altai Mountains. Nature. 2014;505(7481):43–49. pmid:24352235
  68. 68. Durand EY, Patterson N, Reich D, Slatkin M. Testing for ancient admixture between closely related populations. Mol Biol Evol. 2011;28(8):2239–2252. pmid:21325092
  69. 69. Kushniarevich A, Utevska O, Chuhryaeva M, Agdzhoyan A, Dibirova K, Uktveryte I, et al. Genetic heritage of the Balto-Slavic speaking populations: a synthesis of autosomal, mitochondrial and Y-chromosomal data. PLoS One. 2015;10(9):e0135820. pmid:26332464
  70. 70. Davidovic S, Malyarchuk B, Aleksic J, Derenko M, Topalovic V, Litvinov A, et al. Mitochondrial super-haplogroup U diversity in Serbians. Ann Hum Biol. 2017;44(5):408–418. pmid:28140657
  71. 71. Ramos E, Doumatey A, Elkahloun AG, Shriner D, Huang H, Chen G, et al. Pharmacogenomics, ancestry and clinical decision making for global populations. Pharmacogenomics J. 2014;14(3):217–222. pmid:23835662
  72. 72. Wright GEB, Carleton B, Hayden MR, Ross CJD. The global spectrum of protein-coding pharmacogenomic diversity. Pharmacogenomics J. 2018;18(1):187–195. pmid:27779249
  73. 73. van Nimwegen KJ, van Soest RA, Veltman JA, Nelen MR, van der Wilt GJ, Vissers LE, et al. Is the $1000 genome as near as we think? A cost analysis of next-generation sequencing. Clin Chem. 2016;62(11):1458–1464. pmid:27630156
  74. 74. Nielsen R, Paul JS, Albrechtsen A, Song YS. Genotype and SNP calling from next-generation sequencing data. Nat Rev Genet. 2011;12(6):443–451. pmid:21587300
  75. 75. Wall JD, Tang LF, Zerbe B, Kvale MN, Kwok PY, Schaefer C, et al. Estimating genotype error rates from high-coverage next-generation sequence data. Genome Res. 2014;24(11):1734–1739. pmid:25304867
  76. 76. Wang S, Jiang X, Singh S, Marmor R, Bonomi L, Fox D, et al. Genome privacy: challenges, technical approaches to mitigate risk, and ethical considerations in the United States. Ann N Y Acad Sci. 2017;1387(1):73–83. pmid:27681358