Assessment of Imputation Quality: Comparison of Phasing and Imputation Algorithms in Real Data

Stahl, Katharina; Gola, Damian; König, Inke R.

doi:10.3389/fgene.2021.724037

Cited by 15 publications

(12 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We evaluated our AE model in terms of the evaluation metrics CR, Hellinger score, SEN score, and IQS for all the experiments as well as the Pearson correlation coefficient (PCC) in the LOS genotype imputation experiment (Stahl et al, 2021 ). The CR is the ratio of correctly imputed SNPs out of all SNPs.…”

Section: Methodsmentioning

confidence: 99%

“…The details for the definition of the equations for these metrics are shown in Supplementary material. The above four evaluation metrics were based on the comparison between imputed genotypes and the ground truth of the sequenced genotypes (Stahl et al, 2021). For the calculation of the CR, we first calculated the values across SNPs for each sample, and then determined the mean value for all samples.…”

Section: Evaluation Metricsmentioning

confidence: 99%

See 1 more Smart Citation

An autoencoder-based deep learning method for genotype imputation

Song

Greenbaum²,

Luttrell³

et al. 2022

Front. Artif. Intell.

View full text Add to dashboard Cite

Genotype imputation has a wide range of applications in genome-wide association study (GWAS), including increasing the statistical power of association tests, discovering trait-associated loci in meta-analyses, and prioritizing causal variants with fine-mapping. In recent years, deep learning (DL) based methods, such as sparse convolutional denoising autoencoder (SCDA), have been developed for genotype imputation. However, it remains a challenging task to optimize the learning process in DL-based methods to achieve high imputation accuracy. To address this challenge, we have developed a convolutional autoencoder (AE) model for genotype imputation and implemented a customized training loop by modifying the training process with a single batch loss rather than the average loss over batches. This modified AE imputation model was evaluated using a yeast dataset, the human leukocyte antigen (HLA) data from the 1,000 Genomes Project (1KGP), and our in-house genotype data from the Louisiana Osteoporosis Study (LOS). Our modified AE imputation model has achieved comparable or better performance than the existing SCDA model in terms of evaluation metrics such as the concordance rate (CR), the Hellinger score, the scaled Euclidean norm (SEN) score, and the imputation quality score (IQS) in all three datasets. Taking the imputation results from the HLA data as an example, the AE model achieved an average CR of 0.9468 and 0.9459, Hellinger score of 0.9765 and 0.9518, SEN score of 0.9977 and 0.9953, and IQS of 0.9515 and 0.9044 at missing ratios of 10% and 20%, respectively. As for the results of LOS data, it achieved an average CR of 0.9005, Hellinger score of 0.9384, SEN score of 0.9940, and IQS of 0.8681 at the missing ratio of 20%. In summary, our proposed method for genotype imputation has a great potential to increase the statistical power of GWAS and improve downstream post-GWAS analyses.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Evaluation Metricsmentioning

confidence: 99%

An autoencoder-based deep learning method for genotype imputation

Song

Greenbaum²,

Luttrell³

et al. 2022

Front. Artif. Intell.

View full text Add to dashboard Cite

show abstract

“…Therefore, we pursued a meta-analysis of DNA sequencing-based studies that studied the effect of ANGPTL3 , ANGPTL4 , and APOC3 deleterious variants on CAD. The rationale for excluding DNA microarray and exome bead chip-based studies was the potential risk of introducing measurement error for rare variants (50, 51), leading to bias towards the null hypothesis. DNA-sequencing-based substudies from previous papers (3, 46, 52, 53), were extracted and analyzed together with genetic association analyses conducted in the UK Biobank.…”

Section: Resultsmentioning

confidence: 99%

“…Therefore, we pursued a meta-analysis of DNA sequencing-based studies that studied the effect of ANGPTL3 , ANGPTL4 , and APOC3 LoF variants on CAD. The rationale for excluding DNA microarray and exome bead chip-based studies was that they could potentially introduce measurement error for rare variants (48, 49). This would lead to bias towards the null hypothesis.…”

Section: Resultsmentioning

confidence: 99%

Drug-target Mendelian randomization analysis supports lowering plasma ANGPTL3, ANGPTL4, and APOC3 levels as strategies for reducing cardiovascular disease risk

Landfors,

Henneman,

Chorell

et al. 2024

Preprint

View full text Add to dashboard Cite

Background and Aims: APOC3, ANGPTL3, and ANGPTL4 are circulating proteins that are actively pursued as pharmacological targets to treat dyslipidemia and reduce the risk of atherosclerotic cardiovascular disease. Here, we used human genetic data to compare the predicted therapeutic and adverse effects of APOC3, ANGPTL3, and ANGPTL4 inactivation. Methods: We conducted drug-target Mendelian randomization analyses using variants in proximity to the genes associated with circulating protein levels to compare APOC3, ANGPTL3, and ANGPTL4 as drug targets. We obtained exposure and outcome data from large-scale genome-wide association studies and used generalized least squares to correct for linkage disequilibrium-related correlation. We evaluated six primary cardiometabolic endpoints and screened for potential side effects across 694 disease-related endpoints, 43 clinical laboratory tests, and 11 internal organ MRI measurements. Results: Genetically lowering circulating ANGPTL4 levels reduced the odds of coronary artery disease (CAD) (odds ratio, 0.57 per s.d. protein [95%CI,0.47-0.70]) and type 2 diabetes (T2D) (odds ratio, 0.73 per s.d. protein [95%CI,0.57-0.94]). Genetically lowering circulating APOC3 levels also reduced the odds of CAD (odds ratio, 0.90 per s.d. protein [95%CI,0.82-0.99]). Genetically lowered ANGPTL3 levels via common variants were not associated with CAD. However, meta-analysis of loss-of-function variants revealed that ANGPTL3 inactivation protected against CAD (odds ratio, 0.71 per allele [95%CI,0.62-0.82]). Analysis of lowered ANGPTL3, ANGPTL4, and APOC3 levels did not identify important safety concerns. Conclusion: Human genetic evidence suggests that therapies aimed at reducing circulating levels of ANGPTL3, ANGPTL4, and APOC3 reduce the risk of CAD. ANGPTL4 lowering may also reduce the risk of T2D.

show abstract

“…The imputation was performed by IMPUTE 2 with the same reference panel. After imputation, we used SNPs with an imputation quality of info ≥0.8 and MAF ≥0.03 throughout the study (25,26). Genome-wide association study.…”

Section: Genotyping and Imputationmentioning

confidence: 99%

Genotype-Phenotype Models Predicting V̇O2max Response to High-Intensity Interval Training in Physically Inactive Chinese

Yang

Bao

et al. 2023

Medicine &Amp; Science in Sports &Amp; Exercise

View full text Add to dashboard Cite

Purpose This study aimed to analyze the interindividual differences of the maximal oxygen uptake (V̇O2max) response to 12 wk of high-intensity interval training (HIIT), and the genotype–phenotype models were constructed to predict the effect of HIIT on V̇O2max. Methods A total of 228 physically inactive adults who completed a 12-wk HIIT were analyzed. A genome-wide association study (GWAS) was conducted to identify genetic variants associated with the V̇O2max response. Nonresponders, responders, and the highest training responders were defined as the effect sizes (ES) <0.2, ≥0.2, and ≥0.8, respectively. We generated polygenic predictor score (PPS) using lead variants and constructed a predictive model for V̇O2max response based on a linear stepwise regression analysis. Results The V̇O2max increased significantly after HIIT (~14%, P < 0.001), but with interindividual differences (−7.8 to 17.9 mL·kg−1·min−1). In 27% of participants, the V̇O2max showed no improvement. We identified one genetic locus near the γ-aminobutyric acid type A receptor subunit beta 3 gene (GABRB3, rs17116985) associated with V̇O2max response at the genome-wide significance level (P < 5 × 10−8), and an additional nine single nucleotide polymorphisms (SNPs) at the suggestive significance level (P < 1 × 10−5). The SNPs rs474377, rs9365605, and rs17116985, respectively, explained 11%, 9%, and 6.2% of variance in V̇O2max response. The 13 SNPs (P < 1 × 10−5) were found on chromosome 6 (position: 148209316–148223568). Individuals with a PPS greater than 1.757 had the highest response, and those with a PPS lower than −3.712 were nonresponders. The PPS, baseline V̇O2max, sex, and body mass explained 56.4% of the variance in the V̇O2max response; the major predictor was the PPS, which explained 39.4% of the variance. Conclusions The PPS, baseline V̇O2max, sex, and body mass could explain the variance in V̇O2max response. Individuals who had a PPS greater than 1.757 had the highest training response after 12 wk of HIIT. Genetic variants in a region on chromosome 6, especially the sterile alpha motif domain containing 5 gene (SAMD5), which had been explored influencing angiogenesis, might have a potential role in the V̇O2max response.

show abstract

Assessment of Imputation Quality: Comparison of Phasing and Imputation Algorithms in Real Data

Cited by 15 publications

References 33 publications

An autoencoder-based deep learning method for genotype imputation

An autoencoder-based deep learning method for genotype imputation

Drug-target Mendelian randomization analysis supports lowering plasma ANGPTL3, ANGPTL4, and APOC3 levels as strategies for reducing cardiovascular disease risk

Genotype-Phenotype Models Predicting V̇O2max Response to High-Intensity Interval Training in Physically Inactive Chinese

Contact Info

Product

Resources

About