Evaluation of a two-step iterative resampling procedure for internal validation of genome-wide association studies

The association between prescription drugs and colorectal cancer prognosis: a nationwide cohort study using a medication-wide association study

BACKGROUND

With the availability of health insurance claim data, pharmacovigilance for various drugs has been suggested; however, it is necessary to establish an appropriate analysis method. To detect unintended drug effects and to generate new hypotheses, we conducted a hypothesis-free study to systematically examine the relationship between all prescription nonanticancer drugs and the mortality of colorectal cancer patients.

METHODS

We used the Korean National Health Insurance Service-National Sample Cohort database. A total of 2,618 colorectal cancer patients diagnosed between 2004 and 2015 were divided into drug discovery and drug validation sets (1:1) through random sampling. Drugs were classified using the Anatomical Therapeutic Chemical (ATC) classification system: 76 drugs classified as ATC level 2 and 332 drugs classified as ATC level 4 were included in the analysis. We used a Cox proportional hazard model adjusted for sex, age, colorectal cancer treatment, and comorbidities. The relationship between all prescription nonanticancer drugs and the mortality of colorectal cancer patients was analyzed, controlling for multiple comparisons with the false discovery rate.

RESULTS

We found that one ATC level-2 drug (drugs that act on the nervous system, including parasympathomimetics, addictive disorder drugs, and antivertigo drugs) showed a protective effect related to colorectal cancer prognosis. At the ATC level 4 classification, 4 drugs were significant: two had a protective effect (anticholinesterases and opioid anesthetics), and the other two had a detrimental effect (magnesium compounds and Pregnen [4] derivatives).

CONCLUSIONS

In this hypothesis-free study, we identified four drugs linked to colorectal cancer prognosis. The MWAS method can be useful in real-world data analysis.

Proteomic Profiles of Body Mass Index and Waist-to-Hip Ratio and Their Role in Incidence of Diabetes

CONTEXT

It is unclear to what extent the plasma proteome of abdominal fat distribution differs from that of body mass index, and whether the differences have clinical implications.

OBJECTIVE

To evaluate the difference between the plasma proteomic profiles of body mass index (BMI) and waist-to-hip ratio (WHR), and then examine the identified BMI- or WHR-specific proteins in relation to incidence of diabetes.

METHODS

Data were obtained from the Malmö Diet and Cancer-Cardiovascular Cohort study in the general community. Participants (n = 4203) with no previous diabetes (aged 57.2 ± 6.0 years, 37.8% men) were included. Plasma proteins (n = 136) were measured by the Proseek proximity extension method. BMI- and WHR-specific proteins were identified at baseline using a 2-step iterative resampling approach to optimize internal replicability followed by β coefficient comparisons. The identified proteins were considered internally replicated and were then studied in relation to incident diabetes by Cox proportional hazards regression analysis. The main outcome measure was incident diabetes over a mean follow-up of 20.3 ± 5.9 years.

RESULTS

After excluding 21 overlapping proteins and proteins that did not show significantly different associations with BMI vs WHR, 10 internally replicated proteins were found to be specific to BMI, and 22 were found to be specific to WHR (false discovery rate-adjusted P < .05). Of the WHR-specific proteins, 18 remained associated with diabetes risk after multivariate adjustments, whereas none of the BMI-specific proteins showed associations with diabetes risk.

CONCLUSION

Abdominal fat distribution was associated with some unique characteristics of the plasma proteome that potentially could be related to its additional risk of diabetes beyond general obesity.

A multivariant recall-by-genotype study of the metabolomic signature of BMI

OBJECTIVE

This study estimated the effect of BMI on circulating metabolites in young adults using a recall-by-genotype study design.

METHODS

A recall-by-genotype study was implemented in the Avon Longitudinal Study of Parents and Children. Samples from 756 participants were selected for untargeted metabolomics analysis based on low versus high genetic liability for higher BMI defined by a genetic risk score (GRS). Regression analyses were performed to investigate associations between BMI GRS group and relative abundance of 973 metabolites.

RESULTS

After correction for multiple testing, 29 metabolites were associated with BMI GRS group. Bilirubin was among the most strongly associated metabolites, with reduced levels measured in individuals in the high-BMI GRS group (β = -0.32, 95% CI: -0.46 to -0.18, Benjamini-Hochberg adjusted p = 0.005). This study observed associations between BMI GRS group and the levels of several potentially diet-related metabolites, including hippurate, which had lower mean abundance in individuals in the high-BMI GRS group (β = -0.29, 95% CI: -0.44 to -0.15, Benjamini-Hochberg adjusted p = 0.008).

CONCLUSIONS

Together with existing literature, these results suggest that a genetic predisposition to higher BMI captures differences in metabolism leading to adiposity gain. In the absence of prospective data, separating these effects from the downstream consequences of weight gain is challenging.

A polygenic score for acute vaso-occlusive pain in pediatric sickle cell disease

Individuals with monogenic disorders can experience variable phenotypes that are influenced by genetic variation. To investigate this in sickle cell disease (SCD), we performed whole-genome sequencing (WGS) of 722 individuals with hemoglobin HbSS or HbSβ0-thalassemia from Baylor College of Medicine and from the St. Jude Children's Research Hospital Sickle Cell Clinical Research and Intervention Program (SCCRIP) longitudinal cohort study. We developed pipelines to identify genetic variants that modulate sickle hemoglobin polymerization in red blood cells and combined these with pain-associated variants to build a polygenic score (PGS) for acute vaso-occlusive pain (VOP). Overall, we interrogated the α-thalassemia deletion -α3.7 and 133 candidate single-nucleotide polymorphisms (SNPs) across 66 genes for associations with VOP in 327 SCCRIP participants followed longitudinally over 6 years. Twenty-one SNPs in 9 loci were associated with VOP, including 3 (BCL11A, MYB, and the β-like globin gene cluster) that regulate erythrocyte fetal hemoglobin (HbF) levels and 6 (COMT, TBC1D1, KCNJ6, FAAH, NR3C1, and IL1A) that were associated previously with various pain syndromes. An unweighted PGS integrating all 21 SNPs was associated with the VOP event rate (estimate, 0.35; standard error, 0.04; P = 5.9 × 10-14) and VOP event occurrence (estimate, 0.42; standard error, 0.06; P = 4.1 × 10-13). These associations were stronger than those of any single locus. Our findings provide insights into the genetic modulation of VOP in children with SCD. More generally, we demonstrate the utility of WGS for investigating genetic contributions to the variable expression of SCD-associated morbidities.

A genome-wide association study of facial morphology identifies novel genetic loci in Han Chinese

The human face is a heritable surface with many complex sensory organs. In recent years, many genetic loci associated with facial features have been reported in different populations, yet there is a lack of studies on the Han Chinese population. Here, we report a genome-wide association study of 3D normal human faces of 2,659 Han Chinese with autosegment phenotypes of facial morphology. We identify single-nucleotide polymorphisms (SNPs) encompassing four genomic regions showing significant associations with different facial regions, including SNPs in DENND1B associated with the chin, SNPs among PISRT1 associated with eyes, SNPs between DCHS2 and SFRP2 associated with the nose, and SNPs in VPS13B associated with the nose. We replicate 24 SNPs from previously reported genetic loci in different populations, whose candidate genes are DCHS2, SUPT3H, HOXD1, SOX9, PAX3, and EDAR. These results provide a more comprehensive understanding of the genetic basis of variation in human facial morphology.

Variable selection in social-environmental data: sparse regression and tree ensemble machine learning approaches

Background

Social-environmental data obtained from the US Census is an important resource for understanding health disparities, but rarely is the full dataset utilized for analysis. A barrier to incorporating the full data is a lack of solid recommendations for variable selection, with researchers often hand-selecting a few variables. Thus, we evaluated the ability of empirical machine learning approaches to identify social-environmental factors having a true association with a health outcome.

Methods

We compared several popular machine learning methods, including penalized regressions (e.g. lasso, elastic net), and tree ensemble methods. Via simulation, we assessed the methods’ ability to identify census variables truly associated with binary and continuous outcomes while minimizing false positive results (10 true associations, 1000 total variables). We applied the most promising method to the full census data (p = 14,663 variables) linked to prostate cancer registry data (n = 76,186 cases) to identify social-environmental factors associated with advanced prostate cancer.

Results

In simulations, we found that elastic net identified many true-positive variables, while lasso provided good control of false positives. Using a combined measure of accuracy, hierarchical clustering based on Spearman’s correlation with sparse group lasso regression performed the best overall. Bayesian Adaptive Regression Trees outperformed other tree ensemble methods, but not the sparse group lasso. In the full dataset, the sparse group lasso successfully identified a subset of variables, three of which replicated earlier findings.

Conclusions

This analysis demonstrated the potential of empirical machine learning approaches to identify a small subset of census variables having a true association with the outcome, and that replicate across empiric methods. Sparse clustered regression models performed best, as they identified many true positive variables while controlling false positive discoveries.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-020-01183-9.

Identification of Inflammatory and Disease-Associated Plasma Proteins that Associate with Intake of Added Sugar and Sugar-Sweetened Beverages and Their Role in Type 2 Diabetes Risk

It has been suggested that high intake of added sugar and sugar-sweetened beverages (SSBs) increase the level of circulating inflammatory proteins and that chronic inflammation plays a role in type 2 diabetes (T2D) development. We aim to examine how added sugar and SSB intake associate with 136 measured plasma proteins and C-reactive protein (CRP) in the Malmö Diet and Cancer-Cardiovascular Cohort (n = 4382), and examine if the identified added sugar- and SSB-associated proteins associate with T2D incidence. A two-step iterative resampling approach was used to internally replicate proteins that associated with added sugar and SSB intake. Nine proteins were identified to associate with added sugar intake, of which only two associated with T2D incidence (p < 0.00045). Seven proteins were identified to associate with SSB intake, of which six associated strongly with T2D incidence (p < 6.9 × 10-8). No significant associations were observed between added sugar and SSB intake and CRP concentrations. In summary, our elucidation of the relationship between plasma proteome and added sugar and SSB intake, in relation to future T2D risk, demonstrated that SSB intake, rather than the total intake of added sugar, was related to a T2D-pathological proteomic signature. However, external replication is needed to verify the findings.

Find the Needle in the Haystack, Then Find It Again: Replication and Validation in the 'Omics Era

Advancements in high-throughput technologies have made it feasible to study thousands of biological pathways simultaneously for a holistic assessment of health and disease risk via ‘omics platforms. A major challenge in ‘omics research revolves around the reproducibility of findings—a feat that hinges upon balancing false-positive associations with generalizability. Given the foundational role of reproducibility in scientific inference, replication and validation of ‘omics findings are cornerstones of this effort. In this narrative review, we define key terms relevant to replication and validation, present issues surrounding each concept with historical and contemporary examples from genomics (the most well-established and upstream ‘omics), discuss special issues and unique considerations for replication and validation in metabolomics (an emerging field and most downstream ‘omics for which best practices remain yet to be established), and make suggestions for future research leveraging multiple ‘omics datasets.

Views on GWAS statistical analysis

Genome-wide association study (GWAS) is a popular approach to investigate relationships between genetic information and diseases. A number of associations are tested in a study and the results are often corrected using multiple adjustment methods. It is observed that GWAS studies suffer adequate statistical power for reliability. Hence, we document known models for reliability assessment using improved statistical power in GWAS analysis.

A common polymorphism in the retinoic acid pathway modifies adrenocortical carcinoma age-dependent incidence

BACKGROUND

Genome-wide association studies (GWASs) have enriched the fields of genomics and drug development. Adrenocortical carcinoma (ACC) is a rare cancer with a bimodal age distribution and inadequate treatment options. Paediatric ACC is frequently associated with TP53 mutations, with particularly high incidence in Southern Brazil due to the TP53 p.R337H (R337H) germline mutation. The heterogeneous risk among carriers suggests other genetic modifiers could exist.

METHODS

We analysed clinical, genotype and gene expression data derived from paediatric ACC, R337H carriers, and adult ACC patients. We restricted our analyses to single nucleotide polymorphisms (SNPs) previously identified in GWASs to associate with disease or human traits.

RESULTS

A SNP, rs971074, in the alcohol dehydrogenase 7 gene significantly and reproducibly associated with allelic differences in ACC age-of-onset in both cohorts. Patients homozygous for the minor allele were diagnosed up to 16 years earlier. This SNP resides in a gene involved in the retinoic acid (RA) pathway and patients with differing levels of RA pathway gene expression in their tumours associate with differential ACC progression.

CONCLUSIONS

These results identify a novel genetic component to ACC development that resides in the retinoic acid pathway, thereby informing strategies to develop management, preventive and therapeutic treatments for ACC.

Effect Declines Are Systematic, Strong, and Ubiquitous: A Meta-Meta-Analysis of the Decline Effect in Intelligence Research

Empirical sciences in general and psychological science in particular are plagued by replicability problems and biased published effect sizes. Although dissemination bias-related phenomena such as publication bias, time-lag bias, or visibility bias are well-known and have been intensively studied, another variant of effect distorting mechanisms, so-called decline effects, have not. Conceptually, decline effects are rooted in low initial (exploratory) study power due to strategic researcher behavior and can be expected to yield overproportional effect declines. Although decline effects have been documented in individual meta-analytic investigations, systematic evidence for decline effects in the psychological literature remains to date unavailable. Therefore, we present in this meta-meta-analysis a systematic investigation of the decline effect in intelligence research. In all, data from 22 meta-analyses comprising 36 meta-analytical and 1,391 primary effect sizes (N = 697,000+) that have been published in the journal Intelligence were included in our analyses. Two different analytic approaches showed consistent evidence for a higher prevalence of cross-temporal effect declines compared to effect increases, yielding a ratio of about 2:1. Moreover, effect declines were considerably stronger when referenced to the initial primary study within a meta-analysis, yielding about twice the magnitude of effect increases. Effect misestimations were more substantial when initial studies had smaller sample sizes and reported larger effects, thus indicating suboptimal initial study power as the main driver of effect misestimations in initial studies. Post hoc study power comparisons of initial versus subsequent studies were consistent with this interpretation, showing substantially lower initial study power of declining, than of increasing effects. Our findings add another facet to the ever accumulating evidence about non-trivial effect misestimations in the scientific literature. We therefore stress the necessity for more rigorous protocols when it comes to designing and conducting primary research as well as reporting findings in exploratory and replication studies. Increasing transparency in scientific processes such as data sharing, (exploratory) study preregistration, but also self- (or independent) replication preceding the publication of exploratory findings may be suitable approaches to strengthen the credibility of empirical research in general and psychological science in particular.