PMC:4289681 / 9442-29484 JSONTXT

{"target":"https://pubannotation.org/docs/sourcedb/PMC/sourceid/4289681","sourcedb":"PMC","sourceid":"4289681","source_url":"https://www.ncbi.nlm.nih.gov/pmc/4289681","text":"Material and Methods\n\nSNP Chip Data Sets\nThe analyses in this study were based on genetic data from 4,674 Greenlandic participants from three different cohorts. Of these individuals, 4,127 were participants of the Inuit Health in Transition (IHIT) cohort18 and/or the general population health survey (B99)19 from locations all over Greenland (Figure 1). The remaining 547 were from a cohort19 consisting of individuals who have Greenlandic ancestry and live in Denmark. In addition to data from the Greenlandic individuals, genetic data from 50 Danish individuals from the Inter99 cohort20 were included to represent Europeans. All but the 547 individuals from the cohort of Greenlanders living in Denmark were genotyped as a part of a recent disease study8 with the Illumina CardioMetaboChip21 (MetaboChip), which consists of 196,224 SNPs. About half of these SNPs are rare. We used the same chip to genotype the participants from the cohort of Greenlanders living in Denmark for this study. Additionally, Illumina genotyped the four original HapMap populations22 on the MetaboChip, and we used these genotypes to facilitate comparison between the Greenlandic population and other populations. From these data, we made two data sets, on which almost all analyses presented here are based.1. The full data set consisting of data from all 4,674 Greenlandic and all 50 Danish individuals.\n2. A restricted data set consisting of a subset of the Greenlandic individuals who are not closely related, who have no recent European ancestry (\u003c5% estimated European ancestry), and who are not recent migrants within Greenland. When this restricted data set was used in analyses, genetic data from this restricted subset of Greenlandic individuals were either combined with data from the 50 Danish individuals or data from 60 unrelated individuals with European ancestry, 44 unrelated Han Chinese individuals, 45 unrelated Japanese individuals, and 59 unrelated Yoruba individuals, all from HapMap.\nBelow is a detailed description of each of these data sets and what filters were applied to them.\n\nThe Full Data Set\nTo make this data set, we combined (1) genetic data from all genotyped participants of the IHIT cohort and the general population health survey (B99) from 15 locations in Greenland (Figure 1), (2) genetic data from all genotyped participants of the cohort of Greenlanders living in Denmark, and (3) genetic data from 50 Danish individuals from the Inter99 cohort. After merging the data sets, we removed all individuals who appeared in more than one cohort so that each individual was represented only once. We also removed all individuals with more than 2% missing genotypes among the SNPs with a minor allele frequency (MAF) above 1% and all individuals with misspecified or missing gender information. This left us with 4,127 Greenlandic individuals living in Greenland, 547 Greenlandic individuals living in Denmark, and 50 Danish individuals for a total of 4,724 individuals. From this data set we removed all SNPs with a MAF below 5% and/or more than 1% missingness, which left us with data from 92,362 SNPs, of which 92,151 were autosomal. The nonautosomal SNPs were only included in analyses where this is specifically stated.\n\nThe Full Data Set without LD\nSeveral of our analyses were based on Greenlandic allele frequencies corrected for European admixture. To create the data set we used to estimate these allele frequencies, we extracted the MetaboChip data from the 4,127 Greenlandic individuals living in Greenland from the full data set and then filtered out SNPs in strong linkage disequilibrium (LD) by retaining only one SNP for each pair of SNPs showing an r2 value greater than 0.3 in windows of 50 SNPs by using a step size of five SNPs. This left us with 63,911 SNPs.\n\nThe Restricted Greenlandic Data Set Combined with Danish Samples\nOn the basis of admixture proportions estimated with the autosomal SNPs from the full data set under the assumption of two ancestral populations (K = 2), we identified and removed all individuals with an estimated European ancestry proportion above 5%. From the remaining individuals, we extracted individuals from Qaanaaq, Upernavik villages, South Greenland villages (South villages), Tasiilaq villages, and Tasiilaq, which were the only Greenlandic sampling locations with more than 15 individuals left. This left us with 584 Greenlandic individuals with no or very little European ancestry. By applying ADMIXTURE23 with K = 4 and then RelateAdmix24 to these 584 Greenlandic individuals combined with the 50 Danish individuals, we then identified and removed 384 close relatives (Figure S1A, available online) to obtain a data set without closely related individuals. Finally, by principal-component analysis (PCA) of the remaining individuals, we identified and removed nine individuals who did not cluster with the rest of the individuals from the same overall region, i.e., the north (Qaanaaq), west (Upernavik villages), south (South villages), and east (Tasiilaq villages and Tasiilaq) (Figure S1B). The latter was intended to remove putative recent migrants between the different regions of Greenland. The identified putative migrants fit well with known recent migrations in Greenland. For example, it is well known that there are recent migrants between Qaanaaq and Upernavik. In fact, 9% of participants in Kullorsuaq (the northern-most settlements in the Upernavik district) stated that they were born in Avanersuaq (the North Greenlandic county where Qaanaaq is the main settlement), and 2% of participants in Avanersuaq were born in the Upernavik district. All in all, this left us with 191 individuals from five locations in Greenland. To form the restricted Greenlandic data set combined with Danish samples, we extracted data for these 191 individuals and for the 50 Danish individuals from the full data set. Subsequently, using a step size of ten SNPs, we removed strong LD from the data set by removing SNPs such that no pair of SNPs had r2 greater than 0.5 in windows of 100 SNPs. This filtering process resulted in a data set with 31,992 SNP sites.\n\nThe Restricted Greenlandic Data Set Combined with HapMap Samples\nGenotypes from the Greenlandic individuals included in the restricted Greenlandic data set combined with Danish individuals were merged with MetaboChip genotypes of the four original HapMap populations: CEU (Utah residents with ancestry from northern and western Europe from the CEPH collection), JPT (Japanese in Tokyo, Japan), CHB (Han Chinese in Beijing, China) and YRI (Yoruba in Ibadan, Nigeria). We removed the offspring of the HapMap trios such that all individuals in the data set were unrelated. Subsequently, all SNPs with a MAF below 5% and missingness above 1% were removed, leaving data for 102,559 SNP sites. No LD pruning was performed for this data set. We note that the D statistics estimated from this data set are consistent with the claim that the Greenlanders in this restricted data set have no European admixture (Figure S2).\n\nSequencing Data\nFor estimation of site-frequency spectra (SFSs) and sequence-data-based FST, we used the exome sequencing data generated by Moltke et al.8 from the 18 parents of nine trios of Greenlanders with no Danish ancestry from Qaanaaq (three trios), Tasiilaq villages (five trios), and Upernavik villages (one trio). Additionally, we downloaded 18 unrelated HapMap samples from four different populations (CEU, JPT, CHB, and YRI) from 1000 Genomes.25 For the exome data, the extended target region for Agilent SureSelect spanned 75 Mb. These regions were used for both the whole-genome sequencing data from 1000 Genomes and the exome sequencing data.\n\nAdmixture and PCA\nAdmixture proportions were estimated with the ADMIXTURE software23 with a range of K values (the assumed number of ancestral populations). For each K value, we ran ADMIXTURE 100 times with different seeds in order to evaluate convergence. For all values of K, ADMIXTURE converged to the same (largest) likelihood in more than 50% of these 100 runs. The estimated ancestry proportions for different K values were plotted together and colored in a manner that minimized the mean root-squared error between the different plots. PCAs were performed on the basis of the model presented in Patterson et al.,26 and the results were colored on the basis of location.\n\nEstimation of Admixture-Corrected Allele Frequencies\nSeveral of our analyses were based on Greenlandic allele frequencies corrected for European admixture. To estimate admixture-corrected allele frequencies, we first extracted the European ancestry proportions estimated by ADMIXTURE23 under the assumption of two ancestral populations (K = 2). To model the uncertainty associated with sampling individuals from a population, we parameterized the discrete distribution of admixture proportions, for each Greenlandic population, by using a mixture of a point mass at 0, a beta distribution, and a point mass at 1. We therefore had four parameters to estimate: the fraction of individuals with a European ancestry proportion of 0, the fraction of individuals with a European ancestry proportion of 1, and the two parameters of the beta distribution. We computed maximum-likelihood estimates of these parameters. Figure S3 shows the comparison between the average of the observed values and the expectation from the modeled distribution of admixture proportions. Finally, we used the discrete estimated distribution of admixture proportions for each population to compute the Greenlandic allele frequencies without the contribution of genetic admixture with Europeans. For each site, we computed the admixture-corrected Greenlandic allele frequencies f′G asfG′=∑b=1BfG−αbfD1−αbPr(αb)where fG is the admixed Greenlandic allele frequency, fD is the Danish allele frequency, and αb is the European ancestry proportion at the bth bin of the discrete distribution estimated as described above. An arbitrary number of bins (ten) was chosen for ease of calculation.\n\nSFS Estimation\nWe estimated the SFS from sequencing data from five populations (Greenlanders, CEU, CHB, JPT, and YRI) by using the full maximum-likelihood method from Nielsen et al.27 as implemented in ANGSD (Analysis of Next Generation Sequencing Data).28 Because we only had exome data from the Greenlanders, we only included data from the extended Agilent SureSelect exome target region for all five populations. Before estimating the SFSs, we discarded reads with a mapping quality below 30 and bases with a quality score below 20, which correspond to an error rate of 0.1% and 1%, respectively.\nWe also estimated the 2D SFS of the Greenlandic population and CHB and of the Greenlandic population and CEU. For these spectra, we only used sites from the extended target region where we had coverage for both populations.\nWhen estimating the SFSs with ANGSD, we chose to use the SAMtools genotype likelihood model.\n\nLD Estimation for Ancestral Populations\nLD is affected by admixture. Therefore, we developed a model that can accommodate admixture by first estimating the haplotype frequencies in each ancestral population from the observed genotypes and then calculating LD between pairs of SNPs from these haplotype frequencies.\nWe estimated the haplotype frequencies by using a maximum-likelihood approach: let G=(G1,G2, ...,Gn) be the genotypes of n individuals and Gi=(Gi1,Gi2) be the genotypes of individual i at the pair of SNP sites of interest. Further, assume that all n individuals have ancestry from one or more of K ancestral populations and that we know the admixture proportions αi=(αi1,αi2,…,αiK) for each individual i, and let α denote the vector (α1,α2,…,αn) (see previous section for a description of how we inferred the admixture proportions). Finally, let the frequency of haplotype j for the kth population be denoted as pjk, and let h = (h1, h2) be the unobserved pair of haplotypes for an individual, where the two haplotypes originate from the unobserved ancestral populations k1 and k2. Then, the likelihood of the ancestry-specific haplotype frequencies p=(pjk) given the observed genotypes and ancestry proportions can be written asL(p)=Pr(G|p,α)=∏i=1nPr(Gi|p,αi)=∏i=1n∑h∈H∑k1=1K∑k2=1KPr(Gi,h1,h2,k1,k2|p,αi)=∏i=1n∑h∈H∑k1=1K∑k2=1KPr(Gi|p,αi,h1,h2)Pr(h1,h2|p,k1,k2)Pr(k1,k2|αi)=∏i=1n∑h∈H∑k1=1K∑k2=1KPr(Gi|h1,h2)ph1k1ph2k2αik1αik2=∏i=1n∑h∈h(Gi)∑k1=1K∑k2=1Kph1k1ph2k2αik1αik2\nIn the above, we assume that the ancestral population of a haplotype is the same at the two SNP sites, given that these sites are in close proximity along the genome. Furthermore, we denote the set of all possible pairs of haplotypes by H, whereas we denote the set of all pairs of haplotypes that are consistent with the genotypes of individual i by h(Gi). The last rewriting step follows from the observation that Pr(Gi|h1,h2) is equal to 1 if the genotypes Gi are consistent with the haplotypes h1 and h2 (i.e., h belongs to h(Gi)) and that Pr(Gi|h1,h2) is equal to 0 otherwise.\nWe obtain maximum-likelihood estimates of the ancestry-specific haplotype frequencies by maximizing the above likelihood. This is done efficiently with the following expectation-maximization (EM) algorithm. First, random starting points are sampled from a uniform distribution. Then, EM iterations are performed until each new step does not change the parameters (we used a tolerance of 10−6), and each EM iteration for each haplotype frequency is given aspj∗k=12n∑i=1n∑h∈H(Gi)∑k1=1K∑k2=1Kph1k1ph2k2αik1αik2(Ij(h1)Ik(k1)+Ij(h2)Ik(k2))∑h∈H(Gi)∑k1=1K∑k2=1Kph1k1ph2k2αik1αik2(Ik(k1)+Ik(k2)),where I is the indicator function.\n\nInbreeding Estimation\nAdmixture also affects standard estimators of inbreeding coefficients, and we correct for this by allowing for admixture. We estimate the inbreeding coefficient F for each individual with a maximum-likelihood method that uses the estimated admixture proportions α=(α1,α2,…,αK) for the given individual along with the allele frequencies for the K source populations fs=(fs1,fs2,…,fsK) for each site s as estimated by ADMIXTURE.23 Let gs∈(0,1,2) be the individual’s genotype at site s, and let fs∗=∑k=1Kαkfsk be the probability of observing the minor allele at site s. The likelihood of F given the genotype data, D=(g1,g2,…,gS), is then given asL(F)=Pr(D|F)=∏s=1SPr(gs|F)withPr(gs|F)= {(1−fs∗)2(1−F)+(1−fs∗)F, gs=02(1−fs∗)fs∗(1−F),gs=1 fs∗fs∗(1−F)+fs∗F, gs=2\nThe maximum-likelihood estimate of F is then the F value that maximizes the above likelihood.\n\nTreeMix Analyses\nWe performed TreeMix29 analyses of allele frequencies estimated from two different data sets: the full data set without LD and the restricted Greenlandic data set combined with HapMap samples.\nFor the TreeMix analysis of the allele frequencies estimated from the full data set without LD, the allele frequencies were corrected for European ancestry before the analysis was performed. Then, 100 trees were generated with different seeds. Except for a few, they all had the same (highest) likelihood. In the cases where all trees did not have the same likelihood, the tree with the highest likelihood was used. Because this data set was pruned for LD, a window size of one SNP was used.\nTen trees were generated for the TreeMix analysis of the allele frequencies estimated from the restricted Greenlandic data set combined with HapMap samples. All gave the same likelihood. A window size of 100 SNPs was used for accommodating LD; however, increasing the window size to 500 or decreasing to 50 did not change the topology.\n\nD Statistics\nWe performed D-statistic-based tests on the SNP chip data from the restricted Greenlandic data set combined with HapMap samples. First, we estimated the allele frequency in each genotyped site separately for the population in each location. Then, we estimated the D statistics asD(H1,H2;H3,H4)=∑i=1M(fiH3−fiH4)(fiH1−fiH2)∑i=1M(fiH3+fiH4−2fiH3fiH4)(fiH1+fiH2−2fiH1fiH2),where H1, H2, H3, and H4 represent populations in the tree (((H1, H2), H3), H4), where H4 is the outgroup, M is the number of sites included, and fiH1 is the allele frequency for population H1 at site i.30 Only sites with information for all four populations were included. Z scores were obtained from the D statistics with SEs based on a “delete m jackknife for unequal m” procedure31 for 5 Mb regions weighted according to the number of SNPs in each block.\n\nInference of Ancestry Tract Lengths\nGenotypes from the full data set were phased with shapeit232 with the 1000 Genomes phased variant panel (Phase I v.3) as the reference panel. HapMap recombination rates (Phase II b37) were used as a proxy for the human genome genetic map. Local ancestry was inferred with RFMix33 for Qaanaaq and South Village individuals with an estimated global European ancestry proportion greater than 0.05 according to the ADMIXTURE analysis described above. We used two ancestral reference populations. As a proxy for the Inuit ancestral population, we used a reference panel (n = 46) composed of Greenlandic individuals with a global European ancestry proportion less than 0.05 (mean European ancestry = 0.0038; maximum European ancestry = 0.0436). We used the Danish samples (n = 46) as a reference panel to represent the European ancestral population. Local ancestry was inferred for all admixed Qaanaaq and South village individuals jointly with the RFMix admixture timing parameter G = 20, which corresponds to admixture occurring at least 500 years ago and a generation time of 25 years, and a window size of 0.1 cM. We allowed phase correction and used three iterations of the EM algorithm with reference panels included. To control for differences in population-level admixture proportions between Qaanaaq and South villages, we matched individuals according to ancestry proportion as closely as possible, resulting in a matched set of 40 individuals with a European ancestry proportion of at least 0.05 from each population. Length distributions of European admixture tracts were calculated for this matched set, summarized in 5 cM bins, and compared between the two populations.\nThe probability of observing at least one tract of length X cM in an individual, as a function of the time since admixture, can easily be approximated with the Markov approximation to tract lengths introduced by Pool and Nielsen.40 The tract lengths can be described by a two-state Markov process with transition rates λ1 = (1 − m)r(t − 1) and λ2 = mr(t − 1), t \u003e 1 from admixed to unadmixed DNA and from unadmixed to admixed DNA, respectively. Here, m is the admixture proportion, r is the recombination rate per base pair, and t is the admixture time in number of generations. The probability that site j initiates a run of at least k admixed sites is R = π2λ2(1 − λ1)k−1, where π2 is the stationary probability of the unadmixed state (1 − m). If R is small and the length of the genome, S, is large, the probability of observing no runs of length k is then approximately e−RS = exp[S(m − 1)mr(1 + (m − 1)r(t − 1))(k−1)(t − 1)]. With a genome size of S = 2.7 Gb and a recombination rate of 1.3 × 10−8 per base pair, the probability of observing no fragments larger than 39 cM is then equal to 0.9945 for an admixture fraction of m = 0.05 and an admixture time of t = 25. Equivalently, the probability of seeing at least one fragment of length 39 cM is ∼0.005 if the admixture time is 25 generations.\n\nEthical Considerations\nThe Greenlandic samples used in this study were donated by Greenlandic individuals as part of the general public health surveys presented in Jørgensen et al.,18 Bjerregaard et al.,19 and Bjerregaard.34 Ethics approval for genotyping the samples and using the genotype data for public health studies was received from the Commission for Scientific Research in Greenland as a part of the study by Moltke et al.8 The use of the genotype data for the present study has also received ethics approval from the Commission for Scientific Research in Greenland (project 2014-08, reference 2014-098017).","divisions":[{"label":"title","span":{"begin":0,"end":20}},{"label":"sec","span":{"begin":22,"end":7048}},{"label":"title","span":{"begin":22,"end":40}},{"label":"p","span":{"begin":41,"end":1988}},{"label":"label","span":{"begin":1290,"end":1292}},{"label":"p","span":{"begin":1293,"end":1387}},{"label":"label","span":{"begin":1388,"end":1390}},{"label":"p","span":{"begin":1391,"end":1988}},{"label":"p","span":{"begin":1989,"end":2086}},{"label":"sec","span":{"begin":2088,"end":3240}},{"label":"title","span":{"begin":2088,"end":2105}},{"label":"p","span":{"begin":2106,"end":3240}},{"label":"sec","span":{"begin":3242,"end":3795}},{"label":"title","span":{"begin":3242,"end":3270}},{"label":"p","span":{"begin":3271,"end":3795}},{"label":"sec","span":{"begin":3797,"end":6133}},{"label":"title","span":{"begin":3797,"end":3861}},{"label":"p","span":{"begin":3862,"end":6133}},{"label":"sec","span":{"begin":6135,"end":7048}},{"label":"title","span":{"begin":6135,"end":6199}},{"label":"p","span":{"begin":6200,"end":7048}},{"label":"sec","span":{"begin":7050,"end":7707}},{"label":"title","span":{"begin":7050,"end":7065}},{"label":"p","span":{"begin":7066,"end":7707}},{"label":"sec","span":{"begin":7709,"end":8385}},{"label":"title","span":{"begin":7709,"end":7726}},{"label":"p","span":{"begin":7727,"end":8385}},{"label":"sec","span":{"begin":8387,"end":10042}},{"label":"title","span":{"begin":8387,"end":8439}},{"label":"p","span":{"begin":8440,"end":10042}},{"label":"sec","span":{"begin":10044,"end":10960}},{"label":"title","span":{"begin":10044,"end":10058}},{"label":"p","span":{"begin":10059,"end":10643}},{"label":"p","span":{"begin":10644,"end":10867}},{"label":"p","span":{"begin":10868,"end":10960}},{"label":"sec","span":{"begin":10962,"end":13651}},{"label":"title","span":{"begin":10962,"end":11001}},{"label":"p","span":{"begin":11002,"end":11276}},{"label":"p","span":{"begin":11277,"end":12446}},{"label":"p","span":{"begin":12447,"end":13028}},{"label":"p","span":{"begin":13029,"end":13651}},{"label":"sec","span":{"begin":13653,"end":14526}},{"label":"title","span":{"begin":13653,"end":13674}},{"label":"p","span":{"begin":13675,"end":14432}},{"label":"p","span":{"begin":14433,"end":14526}},{"label":"sec","span":{"begin":14528,"end":15565}},{"label":"title","span":{"begin":14528,"end":14544}},{"label":"p","span":{"begin":14545,"end":14737}},{"label":"p","span":{"begin":14738,"end":15229}},{"label":"p","span":{"begin":15230,"end":15565}},{"label":"sec","span":{"begin":15567,"end":16407}},{"label":"title","span":{"begin":15567,"end":15579}},{"label":"p","span":{"begin":15580,"end":16407}},{"label":"sec","span":{"begin":16409,"end":19424}},{"label":"title","span":{"begin":16409,"end":16444}},{"label":"p","span":{"begin":16445,"end":18122}},{"label":"p","span":{"begin":18123,"end":19424}},{"label":"title","span":{"begin":19426,"end":19448}}],"tracks":[{"project":"2_test","denotations":[{"id":"25557782-23656981-2052260","span":{"begin":254,"end":256},"obj":"23656981"},{"id":"25557782-14527126-2052261","span":{"begin":306,"end":308},"obj":"14527126"},{"id":"25557782-14527126-2052262","span":{"begin":391,"end":393},"obj":"14527126"},{"id":"25557782-14634682-2052263","span":{"begin":588,"end":590},"obj":"14634682"},{"id":"25557782-25043022-2052264","span":{"begin":757,"end":758},"obj":"25043022"},{"id":"25557782-22876189-2052265","span":{"begin":793,"end":795},"obj":"22876189"},{"id":"25557782-14685227-2052266","span":{"begin":1063,"end":1065},"obj":"14685227"},{"id":"25557782-19648217-2052267","span":{"begin":4478,"end":4480},"obj":"19648217"},{"id":"25557782-24215025-2052268","span":{"begin":4512,"end":4514},"obj":"24215025"},{"id":"25557782-25043022-2052269","span":{"begin":7203,"end":7204},"obj":"25043022"},{"id":"25557782-20981092-2052270","span":{"begin":7505,"end":7507},"obj":"20981092"},{"id":"25557782-19648217-2052271","span":{"begin":7791,"end":7793},"obj":"19648217"},{"id":"25557782-17194218-2052272","span":{"begin":8328,"end":8330},"obj":"17194218"},{"id":"25557782-19648217-2052273","span":{"begin":8670,"end":8672},"obj":"19648217"},{"id":"25557782-22911679-2052274","span":{"begin":10225,"end":10227},"obj":"22911679"},{"id":"25557782-25420514-2052275","span":{"begin":10298,"end":10300},"obj":"25420514"},{"id":"25557782-19648217-2052276","span":{"begin":14101,"end":14103},"obj":"19648217"},{"id":"25557782-23166502-2052277","span":{"begin":14565,"end":14567},"obj":"23166502"},{"id":"25557782-22960212-2052278","span":{"begin":16152,"end":16154},"obj":"22960212"},{"id":"25557782-23910464-2052279","span":{"begin":16722,"end":16724},"obj":"23910464"},{"id":"25557782-19087958-2052280","span":{"begin":18352,"end":18354},"obj":"19087958"},{"id":"25557782-23656981-2052281","span":{"begin":19607,"end":19609},"obj":"23656981"},{"id":"25557782-14527126-2052282","span":{"begin":19629,"end":19631},"obj":"14527126"},{"id":"25557782-25043022-2052283","span":{"begin":19857,"end":19858},"obj":"25043022"}],"attributes":[{"subj":"25557782-23656981-2052260","pred":"source","obj":"2_test"},{"subj":"25557782-14527126-2052261","pred":"source","obj":"2_test"},{"subj":"25557782-14527126-2052262","pred":"source","obj":"2_test"},{"subj":"25557782-14634682-2052263","pred":"source","obj":"2_test"},{"subj":"25557782-25043022-2052264","pred":"source","obj":"2_test"},{"subj":"25557782-22876189-2052265","pred":"source","obj":"2_test"},{"subj":"25557782-14685227-2052266","pred":"source","obj":"2_test"},{"subj":"25557782-19648217-2052267","pred":"source","obj":"2_test"},{"subj":"25557782-24215025-2052268","pred":"source","obj":"2_test"},{"subj":"25557782-25043022-2052269","pred":"source","obj":"2_test"},{"subj":"25557782-20981092-2052270","pred":"source","obj":"2_test"},{"subj":"25557782-19648217-2052271","pred":"source","obj":"2_test"},{"subj":"25557782-17194218-2052272","pred":"source","obj":"2_test"},{"subj":"25557782-19648217-2052273","pred":"source","obj":"2_test"},{"subj":"25557782-22911679-2052274","pred":"source","obj":"2_test"},{"subj":"25557782-25420514-2052275","pred":"source","obj":"2_test"},{"subj":"25557782-19648217-2052276","pred":"source","obj":"2_test"},{"subj":"25557782-23166502-2052277","pred":"source","obj":"2_test"},{"subj":"25557782-22960212-2052278","pred":"source","obj":"2_test"},{"subj":"25557782-23910464-2052279","pred":"source","obj":"2_test"},{"subj":"25557782-19087958-2052280","pred":"source","obj":"2_test"},{"subj":"25557782-23656981-2052281","pred":"source","obj":"2_test"},{"subj":"25557782-14527126-2052282","pred":"source","obj":"2_test"},{"subj":"25557782-25043022-2052283","pred":"source","obj":"2_test"}]}],"config":{"attribute types":[{"pred":"source","value type":"selection","values":[{"id":"2_test","color":"#93ecd0","default":true}]}]}}