Genotype Imputation Research Articles

Introducing field-programmable gate arrays in genotype phasing and imputation.

We recently developed EagleImp, a free software that combines genotype phasing and imputation in a single tool. By introducing algorithmic and technical improvements we accelerated the classical two-step approach using Eagle2 and PBWT. Here, we demonstrate how to use field-programmable gate arrays (FPGAs) to accelerate EagleImp even further by a factor of up to 93% without loss of phasing and imputation quality. Due to the speed advantage over a not accelerated processor-based implementation, the FPGA extension of EagleImp allows the user to choose a more resource-intensive parameter setting in exchange for computation time to further improve phasing and imputation quality. EagleImp and its FPGA extension are freely available at https://github.com/ikmb/eagleimp and https://github.com/ikmb/eagleimp-fpga.

Genotype imputation in F2 crosses of inbred lines.

Crosses among inbred lines are a fundamental tool for the discovery of genetic loci associated with phenotypes of interest. In organisms for which large reference panels or SNP chips are not available, imputation from low-pass whole-genome sequencing is an effective method for obtaining genotype data from a large number of individuals. To date, a structured analysis of the conditions required for optimal genotype imputation has not been performed. We report a systematic exploration of the effect of several design variables on imputation performance in F2 crosses of inbred medaka lines using the imputation software STITCH. We determined that, depending on the number of samples, imputation performance reaches a plateau when increasing the per-sample sequencing coverage. We also systematically explored the trade-offs between cost, imputation accuracy, and sample numbers. We developed a computational pipeline to streamline the process, enabling other researchers to perform a similar cost-benefit analysis on their population of interest. The source code for the pipeline is available at https://github.com/birneylab/stitchimpute. While our pipeline has been developed and tested for an F2 population, the software can also be used to analyse populations with a different structure.

Genetic susceptibility and late bone outcomes in childhood acute lymphoblastic leukemia survivors.

Childhood acute lymphoblastic leukemia (cALL) survivors are at increased risk for bone comorbidities, but accurate screening tools for such comorbidities are limited. Polygenic scores (PGS) could stratify cALL survivors for risk of long-term adverse bone outcomes. We evaluated 214 (51% female) cALL survivors from the Prévenir les Effets TArdifs de la LEucémie study (median age 21yr). Bone mineral density (BMD) measurements were obtained using dual X-ray absorptiometry at the lumbar spine (LS-BMD), femoral neck (FN-BMD), and total body (TB-BMD), and vertebral fractures (VF) were documented using the vertebral deformity criterion. We computed a PGS for adult heel quantitative ultrasound speed of sound (gSOS), known to be associated with the risk of osteoporotic fracture, using imputed genotype data of the participants, and tested it for association with BMD Z-scores and VF risk, adjusting for clinical risk factors, and in sex and prognostic risk-stratified analyses. We found that a gSOS below the mean was associated with lower BMD in all three sites in univariate and multivariate models. In univariate analyses, 1 SD increase in gSOS conferred a 0.16 SD increase in LS-BMD (95% CI 0.005-0.31), whereas a gSOS above the mean was associated with a 0.31 SD higher LS-BMD (95% CI 0.008-0.61), a 0.36 SD higher TB-BMD (95% CI 0.06-0.67), and a 0.43 SD higher FN-BMD (95% CI 0.13-0.72). Models combining gSOS with clinical risk factors explained up to 16% of the variance of BMD phenotypes and obtained an area under the receiver operating characteristic curve for VF of 0.77 in subgroup analyses. Cranial radiation, high cumulative glucocorticoid doses, high risk group, and male sex were significant risk factors for lower BMD Z-scores. In conclusion, a PGS, in combination with clinical risk factors, could be used as a tool to risk stratify cALL survivors for treatment-related bone morbidity.

The hazards of genotype imputation when mapping disease susceptibility variants

BackgroundThe cost-free increase in statistical power of using imputation to infer missing genotypes is undoubtedly appealing, but is it hazard-free? This case study of three type-2 diabetes (T2D) loci demonstrates that it is not; it sheds light on why this is so and raises concerns as to the shortcomings of imputation at disease loci, where haplotypes differ between cases and reference panel.ResultsT2D-associated variants were previously identified using targeted sequencing. We removed these significantly associated SNPs and used neighbouring SNPs to infer them by imputation. We compared imputed with observed genotypes, examined the altered pattern of T2D-SNP association, and investigated the cause of imputation errors by studying haplotype structure. Most T2D variants were incorrectly imputed with a low density of scaffold SNPs, but the majority failed to impute even at high density, despite obtaining high certainty scores. Missing and discordant imputation errors, which were observed disproportionately for the risk alleles, produced monomorphic genotype calls or false-negative associations. We show that haplotypes carrying risk alleles are considerably more common in the T2D cases than the reference panel, for all loci.ConclusionsImputation is not a panacea for fine mapping, nor for meta-analysing multiple GWAS based on different arrays and different populations. A total of 80% of the SNPs we have tested are not included in array platforms, explaining why these and other such associated variants may previously have been missed. Regardless of the choice of software and reference haplotypes, imputation drives genotype inference towards the reference panel, introducing errors at disease loci.

Imputation-Based HLA Typing with GWAS SNPs.

SNP-based imputation approaches for human leukocyte antigen (HLA) typing take advantage of the haplotype structure within the major histocompatibility complex (MHC) region. These methods predict HLA classical alleles using dense SNP genotypes, commonly found on array-based platforms used in genome-wide association studies (GWAS). The analysis of HLA classical alleles can be conducted on current SNP datasets at no additional cost. Here, we describe the workflow of HIBAG, an imputation method with attribute bagging, to infer a sample's HLA classical alleles using SNP data. Two examples are offered to demonstrate the functionality using public HLA and SNP data from the latest release of the 1000 Genomes project: genotype imputation using pre-built classifiers in a GWAS, and model training to create a new prediction model. The GPU implementation facilitates model building, making it hundreds of times faster compared to the single-threaded implementation.

Graph-Based Imputation Methods and Their Applications to Single Donors and Families.

The outcome of Hematopoietic Stem Cell (HSCT) and organ transplant is strongly affected by the matching of the HLA alleles of the donor and the recipient. However, donors and sometimes recipients are often typed at low resolution, with some alleles either missing or ambiguous. Thus, imputation methods are required to detect the most probably high-resolution HLA haplotypes consistent with a typing. Such imputation algorithms require predefined haplotype frequencies. As such, the phasing of the typing is required for both imputation and frequency generation.We have developed a new approach to HLA haplotype and genotype imputation, where first all candidate phases of a typing are explicated, and then the ambiguity within each phase is solved. This ambiguity is solved through a graph structure of all partial haplotypes and the haplotypes consistent with them.This phasing approach was used to produce an imputation algorithm (GRIMM-Graph Imputation and Matching). GRIMM was then combined with the possibility of combining information from multiple races to produce MR-GRIMM (Multi-Race GRIMM). When family information is available, the phasing of each family member can be restricted by the others. We propose GRAMM (GRaph-bAsed faMily iMputation) to phase alleles in family pedigree HLA typing data and in mother-cord blood unit pairs. Finally, we combined MR-GRIMM with an expectation-maximization (EM) algorithm to estimate haplotype frequencies sharing information between races to produce MR-GRIMME (MR-GRIMM EM).We have shown that these algorithms naturally combine information between races and family members. The accuracy of each of these algorithms is significantly better than its current parallel methods. MR-GRIMM leads to high accuracy in matching predictions. GRAMM better imputes family members than either MR-GRIMM or any existing algorithm and has practically no phasing errors. MR-GRIMME obtains a higher likelihood than existing algorithms.MR-GRIMM, MR-GRIMME, and GRAMM are available as servers or through stand-alone versions in GITHUB and PyPi, as detailed in the appropriate sections.

Genotyping of SNPs in bread wheat at reduced cost from pooled experiments and imputation

Key messagePooling and imputation are computational methods that can be combined for achieving cost-effective and accurate high-density genotyping of both common and rare variants, as demonstrated in a MAGIC wheat population.The plant breeding industry has shown growing interest in using the genotype data of relevant markers for performing selection of new competitive varieties. The selection usually benefits from large amounts of marker data, and it is therefore crucial to dispose of data collection methods that are both cost-effective and reliable. Computational methods such as genotype imputation have been proposed earlier in several plant science studies for addressing the cost challenge. Genotype imputation methods have though been used more frequently and investigated more extensively in human genetics research. The various algorithms that exist have shown lower accuracy at inferring the genotype of genetic variants occurring at low frequency, while these rare variants can have great significance and impact in the genetic studies that underlie selection. In contrast, pooling is a technique that can efficiently identify low-frequency items in a population, and it has been successfully used for detecting the samples that carry rare variants in a population. In this study, we propose to combine pooling and imputation and demonstrate this by simulating a hypothetical microarray for genotyping a population of recombinant inbred lines in a cost-effective and accurate manner, even for rare variants. We show that with an adequate imputation model, it is feasible to accurately predict the individual genotypes at lower cost than sample-wise genotyping and time-effectively. Moreover, we provide code resources for reproducing the results presented in this study in the form of a containerized workflow.

Polygenic scores of subcortical brain volumes as possible modulators of treatment response in depression

A significant proportion of patients with major depressive disorder (MDD) do not experience remission after one or more pharmacological treatments. Research has explored brain structural measures, particularly hippocampal volume, as potential predictors of treatment response, as well as genetic factors.This study investigated the association of polygenic scores (PGSs) for seven subcortical brain volumes (including the hippocampus, nucleus accumbens, amygdala, and caudate nucleus) with treatment non-response and non-remission in MDD.Patients with MDD were recruited in the context of five clinical studies, including a total of 3637 individuals. PGSs were estimated using a Bayesian framework and continuous shrinkage priors (PRS-CS-auto) after standard genotype quality control and imputation. Logistic regressions were performed between PGSs and non-response or non-remission in each sample, adjusting for age, sex, baseline symptom severity, recruitment sites, and population stratification. Results were meta-analysed across samples, using a random-effect model.No association was significant in the meta-analysis after Bonferroni correction. The top finding was found for the caudate volume PGS and non-remission (OR = 1.09, 95% CI = 1.01–1.19, p = 0.036), with no evidence of heterogeneity. Leave-one-out sensitivity analyses showed that this result was influenced by the two largest samples in the meta-analysis.This result should be considered as preliminary as it did not reach the Bonferroni-adjusted significance threshold. Future studies with greater statistical power may enhance the predictive performance of PGSs and contribute to the identification of polygenic predictors of treatment outcomes in MDD, contributing to precision psychiatry.

A compendium of genetic regulatory effects across pig tissues

The Farm Animal Genotype-Tissue Expression (FarmGTEx) project has been established to develop a public resource of genetic regulatory variants in livestock, which is essential for linking genetic polymorphisms to variation in phenotypes, helping fundamental biological discovery and exploitation in animal breeding and human biomedicine. Here we show results from the pilot phase of PigGTEx by processing 5,457 RNA-sequencing and 1,602 whole-genome sequencing samples passing quality control from pigs. We build a pig genotype imputation panel and associate millions of genetic variants with five types of transcriptomic phenotypes in 34 tissues. We evaluate tissue specificity of regulatory effects and elucidate molecular mechanisms of their action using multi-omics data. Leveraging this resource, we decipher regulatory mechanisms underlying 207 pig complex phenotypes and demonstrate the similarity of pigs to humans in gene expression and the genetic regulation behind complex phenotypes, supporting the importance of pigs as a human biomedical model.

Accuracy of genotype imputation of a low-density SNP array for the Amazon fish Colossoma macropomum.

In South America, Tambaqui (Colossoma macropomum) stands as the primary target for aquaculture, yet breeding programs for this Amazon native species are in their early stages. While high-density single nucleotide polymorphism (SNP) arrays are pivotal for aquaculture breeding, their costs can be prohibitive for non- or semi-industrial species. To overcome this, a cost-effective approach involves developing low-density SNP arrays followed by genotype imputation to higher densities. In this study, a 1K SNP array for tambaqui was created and validated, offering a balance between SNP quantity and genome representativity. The imputation accuracy from various SNP densities to a medium-density array was evaluated, with the 1K density demonstrating the best trade-off (accuracy of 0.93). This subset was further utilized to construct a commercial array through Agriseq™ targeted genotyping-by-sequencing, validated in 192 DNA samples, affirming its high quality for genotyping tambaqui. The low-density SNP array, with genome-wide coverage and high polymorphism, emerges as an effective tool for exploring genetic variation within diverse populations. Population analyses using the 1K panel proved to be an efficient tool for genetic characterization of sampled broodstocks, making it a valuable resource for genetic improvement programs targeting this Amazon native species.

A rapid and reference-free imputation method for low-cost genotyping platforms

Most current genotype imputation methods are reference-based, which posed several challenges to users, such as high computational costs and reference panel inaccessibility. Thus, deep learning models are expected to create reference-free imputation methods performing with higher accuracy and shortening the running time. We proposed a imputation method using recurrent neural networks integrating with an additional discriminator network, namely GRUD. This method was applied to datasets from genotyping chips and Low-Pass Whole Genome Sequencing (LP-WGS) with the reference panels from The 1000 Genomes Project (1KGP) phase 3, the dataset of 4810 Singaporeans (SG10K), and The 1000 Vietnamese Genome Project (VN1K). Our model performed more accurately than other existing methods on multiple datasets, especially with common variants with large minor allele frequency, and shrank running time and memory usage. In summary, these results indicated that GRUD can be implemented in genomic analyses to improve the accuracy and running-time of genotype imputation.

A Genome-Wide Association Study of Metabolic Syndrome in the Taiwanese Population.

The purpose of this study was to investigate genetic factors associated with metabolic syndrome (MetS) by conducting a large-scale genome-wide association study (GWAS) in Taiwan, addressing the limited data on Asian populations compared to Western populations. Using data from the Taiwan Biobank, comprehensive clinical and genetic information from 107,230 Taiwanese individuals was analyzed. Genotyping data from the TWB1.0 and TWB2.0 chips, including over 650,000 single nucleotide polymorphisms (SNPs), were utilized. Genotype imputation using the 1000 Genomes Project was performed, resulting in more than 9 million SNPs. MetS was defined based on a modified version of the Adult Treatment Panel III criteria. Among all participants (mean age: 50 years), 23% met the MetS definition. GWAS analysis identified 549 SNPs significantly associated with MetS, collectively mapping to 10 genomic risk loci. Notable risk loci included rs1004558, rs3812316, rs326, rs4486200, rs2954038, rs10830963, rs662799, rs62033400, rs183130, and rs34342646. Gene-set analysis revealed 22 associated genes: CETP, LPL, APOA5, SIK3, ZPR1, APOC1, BUD13, MLXIPL, TOMM40, GCK, YKT6, RPS6KB1, FTO, VMP1, TUBD1, BCL7B, C19orf80 (ANGPTL8), SIDT2, SENP7, PAFAH1B2, DOCK6, and FOXA2. This study identified genomic risk loci for MetS in a large Taiwanese population through a comprehensive GWAS approach. These associations provide novel insights into the genetic basis of MetS and hold promise for the potential discovery of clinical biomarkers.

Accurate detection of identity-by-descent segments in human ancient DNA

Long DNA segments shared between two individuals, known as identity-by-descent (IBD), reveal recent genealogical connections. Here we introduce ancIBD, a method for identifying IBD segments in ancient human DNA (aDNA) using a hidden Markov model and imputed genotype probabilities. We demonstrate that ancIBD accurately identifies IBD segments >8 cM for aDNA data with an average depth of >0.25× for whole-genome sequencing or >1× for 1240k single nucleotide polymorphism capture data. Applying ancIBD to 4,248 ancient Eurasian individuals, we identify relatives up to the sixth degree and genealogical connections between archaeological groups. Notably, we reveal long IBD sharing between Corded Ware and Yamnaya groups, indicating that the Yamnaya herders of the Pontic-Caspian Steppe and the Steppe-related ancestry in various European Corded Ware groups share substantial co-ancestry within only a few hundred years. These results show that detecting IBD segments can generate powerful insights into the growing aDNA record, both on a small scale relevant to life stories and on a large scale relevant to major cultural-historical events.

Development and evaluation of a haplotype reference panel of Zhikong scallop (Chlamys farreri) for genotype imputation

Single nucleotide polymorphisms (SNPs) across the genome are remarkably useful resources for investigating the genome architecture, studying the genetic basis of complex traits and implementation of genomic selection. However, information on high-density SNPs is unavailable for most aquaculture species owing to the high costs of whole genome sequencing. Genotype imputation is an attractive approach for obtaining large number of genomic variants at a low cost. In this study, the first haplotype reference panel was constructed for Chlamys farreri, an economically important mollusk species and a major contributor to aquaculture in China. A total of 174 scallops from six populations in the main scallop culture area were selected for whole-genome resequencing. More than 87.47 million SNPs were identified and 3,968,417 SNPs genome-wide were selected for constructing the haplotype reference panel for C. farreri. The number of SNPs per chromosome ranged from 125,000 to 279,981 with an average spacing of 242 bases. Several factors were evaluated based on the imputation performance of the reference panel. The imputation accuracies of 19 chromosomes ranged from 0.91 to 0.95, and there were no significant differences among them (p > 0.05). The imputation accuracies of populations could reach 0.95, 0.93, and 0.90 for ZY, PR, and CS respectively, which confirmed a positive relationship between the kinship within the population and the imputation accuracy. The imputation accuracies increased with sequencing depth and plateaued at 0.5×, reaching accuracies of 0.91–0.93 for artificially breeding populations. Taken together, we constructed a haplotype reference panel for C. farreri, which can provide a cost-effective way to extend the limited number of genotypes for boosting the power of downstream analyses such as genome-wide association studies and genomic selection.

Imputation strategies for genomic prediction using nanopore sequencing

BackgroundGenomic prediction describes the use of SNP genotypes to predict complex traits and has been widely applied in humans and agricultural species. Genotyping-by-sequencing, a method which uses low-coverage sequence data paired with genotype imputation, is becoming an increasingly popular SNP genotyping method for genomic prediction. The development of Oxford Nanopore Technologies’ (ONT) MinION sequencer has now made genotyping-by-sequencing portable and rapid. Here we evaluate the speed and accuracy of genomic predictions using low-coverage ONT sequence data in a population of cattle using four imputation approaches. We also investigate the effect of SNP reference panel size on imputation performance.ResultsSNP array genotypes and ONT sequence data for 62 beef heifers were used to calculate genomic estimated breeding values (GEBVs) from 641 k SNP for four traits. GEBV accuracy was much higher when genome-wide flanking SNP from sequence data were used to help impute the 641 k panel used for genomic predictions. Using the imputation package QUILT, correlations between ONT and low-density SNP array genomic breeding values were greater than 0.91 and up to 0.97 for sequencing coverages as low as 0.1 × using a reference panel of 48 million SNP. Imputation time was significantly reduced by decreasing the number of flanking sequence SNP used in imputation for all methods. When compared to high-density SNP arrays, genotyping accuracy and genomic breeding value correlations at 0.5 × coverage were also found to be higher than those imputed from low-density arrays.ConclusionsHere we demonstrated accurate genomic prediction is possible with ONT sequence data from sequencing coverages as low as 0.1 × , and imputation time can be as short as 10 min per sample. We also demonstrate that in this population, genotyping-by-sequencing at 0.1 × coverage can be more accurate than imputation from low-density SNP arrays.

Reconstruction of private genomes through reference-based genotype imputation

BackgroundGenotype imputation is an essential step in genetic studies to improve data quality and statistical power. Public imputation servers are widely used by researchers to impute their data using otherwise access-controlled reference panels of high-fidelity genomes held by these servers.ResultsWe report evidence against the prevailing assumption that providing access to panels only indirectly via imputation servers poses a negligible privacy risk to individuals in the panels. To this end, we present algorithmic strategies for adaptively constructing artificial input samples and interpreting their imputation results that lead to the accurate reconstruction of reference panel haplotypes. We illustrate this possibility on three reference panels of real genomes for a range of imputation tools and output settings. Moreover, we demonstrate that reconstructed haplotypes from the same individual could be linked via their genetic relatives using our Bayesian linking algorithm, which allows a substantial portion of the individual’s diploid genome to be reassembled. We also provide population genetic estimates of the proportion of a panel that could be linked when an adversary holds a varying number of genomes from the same population.ConclusionsOur results show that genomes in imputation server reference panels can be vulnerable to reconstruction, implying that additional safeguards may need to be considered. We suggest possible mitigation measures based on our findings. Our work illustrates the value of adversarial algorithms in uncovering new privacy risks to help inform the genomics community towards secure data sharing practices.

Low-pass sequencing plus imputation using avidity sequencing displays comparable imputation accuracy to sequencing by synthesis while reducing duplicates.

Low-pass sequencing with genotype imputation has been adopted as a cost-effective method for genotyping. The most widely used method of short-read sequencing uses sequencing by synthesis (SBS). Here we perform a study of a novel sequencing technology-avidity sequencing. In this short note, we compare the performance of imputation from low-pass libraries sequenced on an Element AVITI system (which utilizes avidity sequencing) to those sequenced on an Illumina NovaSeq 6000 (which utilizes SBS) with an SP flow cell for the same set of biological samples across a range of genetic ancestries. We observed dramatically lower optical duplication rates in the data deriving from the AVITI system compared to the NovaSeq 6000, resulting in higher effective coverage given a fixed number of sequenced bases, and comparable imputation accuracy performance between sequencing chemistries across ancestries. This study demonstrates that avidity sequencing is a viable alternative to the standard SBS chemistries for applications involving low-pass sequencing plus imputation.

Prioritized imputed sequence variants from multi-population GWAS improve prediction accuracy for sea lice count in Atlantic salmon (Salmo salar)

Sea lice infestation is one of the major fish health problems that occurs during the grow-out phase in Atlantic salmon (Salmo salar) aquaculture. In this study, we integrated different genomic approaches, including whole-genome sequencing (WGS), genotype imputation and meta-analysis of genome-wide association studies (GWAS), to identify single-nucleotide polymorphisms (SNPs) associated with sea lice count in Atlantic salmon. Different sets of trait-associated SNPs were prioritized and compared against randomly chosen markers, based on the accuracy of genomic predictions for the trait. Lice count phenotypes and dense genotypes of five breeding populations challenged with sea lice were used. Genotype imputation was applied to increase SNP density of challenged animals to WGS level. The summary statistics from GWAS of each population were then combined in a meta-analysis to increase the sample size and improve the statistical power of associations. Eight different genotyping scenarios were considered for genomic prediction: 70K_array: 70 K standard genotyping panel; 70K_priori: 70 K SNPs with the highest p-values identified in the meta-analysis; 30K_priori: 30 K SNPs with the highest p-values identified in the meta-analysis; WGS: SNPs imputed to whole-genome sequencing level. The remaining four scenarios were the same SNP sets with a linkage disequilibrium (LD) pruning filter: 70K_array_LD; 70K_priori_LD; 30K_priori_LD and WGS_LD. Genomic prediction accuracy was evaluated using a five-fold cross-validation scheme in two different populations which were excluded from the meta-analysis to remove possible validation-reference bias. Results showed significant genetic variation for the number of sea lice in Atlantic salmon across populations, with heritabilities ranging from 0.06 to 0.24. The meta-analysis identified several SNPs associated with sea lice resistance, mainly in Ssa03 and Ssa09 chromosomes. Genomic prediction using the GWAS-based prioritized SNPs showed higher accuracy compared to the standard SNP array in most of scenarios; achieving up to 57% increase in accuracy. Accuracy of prioritized scenarios was higher for the 70K_priori than the 30K_priori. The use of WGS data in genomic prediction presented marginal or negative accuracy gain compared to the standard SNP array. The LD-pruning filter presented no benefits, reducing accuracy in most scenarios. Overall, our study demonstrated the potential of prioritized imputed sequence variants from multi-population GWAS meta-analysis to improve prediction accuracy for sea lice counts in Atlantic salmon. The findings suggest that incorporating WGS data and prioritized SNPs from GWAS meta-analysis can accelerate the genetic progress of selection for polygenic traits in salmon aquaculture.

GWAS Identifies Variants Associated with Minimal Residual Disease after Induction I in Pediatric Patients with Newly Diagnosed Acute Myeloid Leukemia

Acute myeloid leukemia (AML) patients' clinical outcomes are associated with genetic abnormalities present in leukemic blasts. However, the significance of common genetic variants with regards to variation in AML treatment outcome remains elusive. To that end, we conducted a genome wide association study (GWAS) on 400 pediatric patients diagnosed with de novo AML to identify single nucleotide polymorphisms (SNPs) that may be associated with minimal residual disease (MRD) after the initial induction chemotherapy. Integrating our top SNPs from GWAS results with matched AML-derived transcriptomic data, we performed an expression quantitative trait loci (eQTL) analysis to determine which variants may affect the expression of one or more genes. Array-based whole-genome genotyping with imputation was used to investigate SNPs for association with MRD outcome in 2 pediatric AML cohorts: the multi-site AML02 [NCT00136084, n=167] and AML08 [NCT00703820, n=233] trials. Genotype imputation was performed via TOPMED Imputation Server and yielded 6,741,200 SNPs. MRD positivity was defined as 1 or more leukemic cells per 1000 mononuclear bone-marrow cells. Association analysis with adjustment for ancestry was conducted using logistic regression to model MRD positivity with SNP as a predictor. Genome-wide significance level was set at Bonferroni-corrected P<5x10 -8 and a suggestive significance level of P<1x10 -4. For gene expression profiling, the mRNA expression levels for 137 patients were obtained at diagnosis using GeneChip Human Genome U133A array from patients in the AML02 cohort. For eQTL analysis, we evaluated the association of gene expression using a 500kb window surrounding the top SNPs identified in the GWAS using a linear regression model. A total of 92 SNPs reached the suggestive significance level but none passed the genome-wide threshold. Of the 92 SNPs, 10 SNPs were mapped to the Major Histocompatibility Complex, Class II, DP Alpha 1 gene (HLA-DPA1), with 8 of the 10 SNPs associated with an increased risk of MRD positivity. The top SNP in HLA-DPA1 is located in intron 1 with presence of the variant A allele conferring increased MRD positivity (OR=2.17, 95% CI (1.55-3.04), P=6.64x10 -6, Figure 1A). Of note, HLA-DPA1 SNP was not significantly associated with event free survival (EFS; HR: 1.21, 95% CI (0.95-1.56), P=0.13) or overall survival (OS; HR: 1.28, 95% CI (0.96-1.71), P=0.09), however the direction of association was consistent with the GWAS results. HLA-DPA1 is a protein coding gene and plays an integral role in the immune response by presenting peptides derived from extracellular proteins. HLA-DPA1 and several other MHC class II genes have shown to be downregulated in patients diagnosed with AML who experienced a relapse (PMID: 30380364). The eQTL analysis revealed a significant association with presence of the variant A associated with reduced mRNA expression of HLA-DPA1 (P=0.01, Figure 1B). In addition to the cis-eQTL association of our top SNP with HLA-DPA1 expression, significant associations were also observed with neighboring genes: HLA-DOA, HLA-DQA1, HLA-DQB1, and DAXX. Furthermore, when examining transcriptomic data for association with survival outcomes, we observed high HLA-DPA1 gene expression was significantly associated with favorable OS (HR: 0.46, 95%CI (0.23-0.91), P=0.025), though this association had similar direction it was not significant for association with EFS (HR: 0.68, 95% CI (0.39-1.18), P=0.169). Although limited by numbers, this study is one of the largest GWAS conducted in pediatric AML to identify common SNPs associated with MRD status after initial chemotherapy induction using an integrated genomic and transcriptomic validation approach. None of the SNPs reached genome-wide significant levels, however our top results show that genetic variation in several genes of biological relevance to have an impact on an early outcome endpoint. Moreover, we also have shown that our top SNP located in HLA-DPA1 gene is an eQTL and its mRNA expression is significantly associated with OS. Future studies are aimed at integrating other clinical endpoints such as EFS and OS with hope of creating a polygenic risk score to personalize treatment approaches. In addition, validation of these findings in larger cohorts of AML patients are warranted with future analyses to include known prognostic factors such cytogenetically defined risk.

Sequenced-based GWAS for linear classification traits in Belgian Blue beef cattle reveals new coding variants in genes regulating body size in mammals.

Cohorts of individuals that have been genotyped and phenotyped for genomic selection programs offer the opportunity to better understand genetic variation associated with complex traits. Here, we performed an association study for traits related to body size and muscular development in intensively selected beef cattle. We leveraged multiple trait information to refine and interpret the significant associations. After a multiple-step genotype imputation to the sequence-level for 14,762 Belgian Blue beef (BBB) cows, we performed a genome-wide association study (GWAS) for 11 traits related to muscular development and body size. The 37 identified genome-wide significant quantitative trait loci (QTL) could be condensed in 11 unique QTL regions based on their position. Evidence for pleiotropic effects was found in most of these regions (e.g., correlated association signals, overlap between credible sets (CS) of candidate variants). Thus, we applied a multiple-trait approach to combine information from different traits to refine the CS. In several QTL regions, we identified strong candidate genes known to be related to growth and height in other species such as LCORL-NCAPG or CCND2. For some of these genes, relevant candidate variants were identified in the CS, including three new missense variants in EZH2, PAPPA2 and ADAM12, possibly two additional coding variants in LCORL, and candidate regulatory variants linked to CCND2 and ARMC12. Strikingly, four other QTL regions associated with dimension or muscular development traits were related to five (recessive) deleterious coding variants previously identified. Our study further supports that a set of common genes controls body size across mammalian species. In particular, we added new genes to the list of those associated with height in both humans and cattle. We also identified new strong candidate causal variants in some of these genes, strengthening the evidence of their causality. Several breed-specific recessive deleterious variants were identified in our QTL regions, probably as a result of the extreme selection for muscular development in BBB cattle.