TOSCCA: a framework for interpretation and testing of sparse canonical correlations

Summary

In clinical and biomedical research, multiple high-dimensional datasets are nowadays routinely collected from omics and imaging devices. Multivariate methods, such as Canonical Correlation Analysis (CCA), integrate two (or more) datasets to discover and understand underlying biological mechanisms. For an explorative method like CCA, interpretation is key. We present a sparse CCA method based on soft-thresholding that produces near-orthogonal components, allows for browsing over various sparsity levels, and permutation-based hypothesis testing. Our soft-thresholding approach avoids tuning of a penalty parameter. Such tuning is computationally burdensome and may render unintelligible results. In addition, unlike alternative approaches, our method is less dependent on the initialization. We examined the performance of our approach with simulations and illustrated its use on real cancer genomics data from drug sensitivity screens. Moreover, we compared its performance to Penalized Matrix Analysis (PMA), which is a popular alternative of sparse CCA with a focus on yielding interpretable results. Compared to PMA, our method offers improved interpretability of the results, while not compromising, or even improving, signal discovery.

Availability and implementation

The software and simulation framework are available at https://github.com/nuria-sv/toscca.

Adherence to statin treatment in patients with familial hypercholesterolemia: A dynamic prediction model

BACKGROUND

Statins are the primary therapy in patient with heterozygous familial hypercholesterolemia (HeFH). Non-adherence to statin therapy is associated with increased cardiovascular risk.

OBJECTIVE

We constructed a dynamic prediction model to predict statin adherence for an individual HeFH patient for each upcoming statin prescription.

METHODS

All patients with HeFH, identified by the Dutch Familial Hypercholesterolemia screening program between 1994 and 2014, were eligible. National pharmacy records dated between 1995 and 2015 were linked. We developed a dynamic prediction model that estimates the probability of statin adherence (defined as proportion of days covered >80%) for an upcoming prescription using a mixed effect logistic regression model. Static and dynamic patient-specific predictors, as well as data on a patient's adherence to past prescriptions were included. The model with the lowest AIC (Akaike Information Criterion) value was selected.

RESULTS

We included 1094 patients for whom 21,171 times a statin was prescribed. Based on the model with the lowest AIC, age at HeFH diagnosis, history of cardiovascular event, time since HeFH diagnosis and duration of the next statin prescription contributed to an increased adherence, while adherence decreased with higher untreated LDL-C levels and higher intensity of statin therapy. The dynamic prediction model showed an area under the curve of 0.63 at HeFH diagnosis, which increased to 0.85 after six years of treatment.

CONCLUSION

This dynamic prediction model enables clinicians to identify HeFH patients at risk for non-adherence during statin treatment. These patients can be offered timely interventions to improve adherence and further reduce cardiovascular risk.

Effect of parental and ART treatment characteristics on perinatal outcomes

STUDY QUESTION

Do parental characteristics and treatment with ART affect perinatal outcomes in singleton pregnancies?

SUMMARY ANSWER

Both parental and ART treatment characteristics affect perinatal outcomes in singleton pregnancies.

WHAT IS KNOWN ALREADY

Previous studies have shown that singleton pregnancies resulting from ART are at risk of preterm birth. ART children are lighter at birth after correction for duration of gestation and at increased risk of congenital abnormalities compared to naturally conceived children. This association is confounded by parental characteristics that are also known to affect perinatal outcomes. It is unclear to which extent parental and ART treatment characteristics independently affect perinatal outcomes.

STUDY DESIGN, SIZE, DURATION

All IVF clinics in the Netherlands (n = 13) were requested to provide data on all ART treatment cycles (IVF, ICSI and frozen-thawed embryo transfers (FET)), performed between 1 January 2000, and 1 January 2011, which resulted in a pregnancy. Using probabilistic data-linkage, these data (n = 36 683) were linked to the Dutch Perinatal Registry (Perined), which includes all children born in the Netherlands in the same time period (n = 2 548 977).

PARTICIPANTS/MATERIALS, SETTING, METHODS

Analyses were limited to singleton pregnancies that resulted from IVF, ICSI or FET cycles. Multivariable models for linear and logistic regression were fitted including parental characteristics as well as ART treatment characteristics. Analyses were performed separately for fresh cycles and for fresh and FET cycles combined. We assessed the impact on the following perinatal outcomes: birth weight, preterm birth below 37 or 32 weeks of gestation, congenital malformations and perinatal mortality.

MAIN RESULTS AND THE ROLE OF CHANCE

The perinatal outcomes of 31 184 out of the 36 683 ART treatment cycles leading to a pregnancy were retrieved through linkage with the Perined (85% linkage). Of those, 23 671 concerned singleton pregnancies resulting from IVF, ICSI or FET. Birth weight was independently associated with both parental and ART treatment characteristics. Characteristics associated with lower birth weight included maternal hypertensive disease, non-Dutch maternal ethnicity, nulliparity, increasing duration of subfertility, hCG for luteal phase support (compared to progesterone), shorter embryo culture duration, increasing number of oocytes retrieved and fresh embryo transfer. The parental characteristic with the greatest effect size on birth weight was maternal diabetes (adjusted difference 283 g, 95% CI 228–338). FET was the ART treatment characteristic with the greatest effect size on birth weight (adjusted difference 100 g, 95% CI 84–117) compared to fresh embryo transfer. Preterm birth was more common among mothers of South-Asian ethnicity. Preterm birth was less common among multiparous women and women with ‘male factor’ as treatment indication (compared to ‘tubal factor’).

LIMITATIONS, REASONS FOR CAUTION

Due to the retrospective nature of our study, we cannot prove causality. Further limitations of our study were the inability to adjust for mothers giving birth more than once in our dataset, missing values for several variables and limited information on parental lifestyle and general health.

WIDER IMPLICATIONS OF THE FINDINGS

Multiple parental and ART treatment characteristics affect perinatal outcomes, with birth weight being influenced by the widest range of factors. This highlights the importance of assessing both parental and ART treatment characteristics in studies that focus on the health of ART-offspring, with the purpose of modifying these factors where possible. Our results further support the hypothesis that the embryo is sensitive to its early environment.

STUDY FUNDING/COMPETING INTEREST(S)

This study was funded by Foreest Medical School, Alkmaar, the Netherlands (grants: FIO 1307 and FIO 1505). B.W.M. reports grants from NHMRC and consultancy for ObsEva, Merck KGaA, iGenomics and Guerbet. F.B. reports research support grants from Merck Serono and personal fees from Merck Serono. A.C. reports travel support from Ferring BV. and Theramex BV. and personal fees from UpToDate (Hyperthecosis), all outside the remit of the current work. The remaining authors report no conflict of interests.

TRIAL REGISTRATION NUMBER

N/A.

Feeding patterns and BMI trajectories during infancy: a multi-ethnic, prospective birth cohort

BACKGROUND

Milk feeding type (exclusive breastfeeding [EBF], formula feeding or mixed feeding) and timing of complementary feeding (CF) have been associated with infant growth. However, studies evaluating their combined role, and the role of ethnicity, are scarce. We examined associations of feeding patterns (milk feeding type combined with timing of CF) with infant body mass index (BMI) trajectories and potential ethnic-specific associations.

METHODS

Infant feeding and BMI data during the 1st year of life from 3524 children (Dutch n = 2880, Moroccan n = 404 and Turkish n = 240) from the Amsterdam Born Children and their Development (ABCD) cohort were used. Six feeding patterns were defined: EBF/earlyCF, EBF/lateCF (reference), formula/earlyCF, formula/lateCF, mixed/earlyCF and mixed/lateCF. A covariate adjusted latent class mixed model was applied to simultaneously model BMI trajectories and associations with feeding patterns. Potential ethnic differences in the associations were studied in a separate model where interactions between ethnicity and feeding patterns were included.

RESULTS

Four distinct BMI trajectories (low, mid-low, mid-high and high) were identified. Feeding pattern of formula/earlyCF was associated with lower odds for low (OR: 0.43; 95% CI: 0.25, 0.76) or mid-high (0.28; 0.16, 0.51) (ref: high) trajectory compared with EBF/lateCF pattern (ref). An ethnic-specific model revealed that among Dutch infants, formula/earlyCF pattern was associated with lower odds for low trajectory (0.46; 0.24, 0.87), whereas among Turkish/Moroccan infants almost all feeding patterns were associated with lower odds for the low trajectory (ref: high).

CONCLUSION

Infant feeding patterns are associated with early BMI trajectories with specific ethnic differences. Future studies should take the role of ethnicity into account in the associations between infant feeding and growth.

Improved selection of participants in genetic longevity studies: family scores revisited

BACKGROUND

Although human longevity tends to cluster within families, genetic studies on longevity have had limited success in identifying longevity loci. One of the main causes of this limited success is the selection of participants. Studies generally include sporadically long-lived individuals, i.e. individuals with the longevity phenotype but without a genetic predisposition for longevity. The inclusion of these individuals causes phenotype heterogeneity which results in power reduction and bias. A way to avoid sporadically long-lived individuals and reduce sample heterogeneity is to include family history of longevity as selection criterion using a longevity family score. A main challenge when developing family scores are the large differences in family size, because of real differences in sibship sizes or because of missing data.

METHODS

We discussed the statistical properties of two existing longevity family scores: the Family Longevity Selection Score (FLoSS) and the Longevity Relatives Count (LRC) score and we evaluated their performance dealing with differential family size. We proposed a new longevity family score, the mLRC score, an extension of the LRC based on random effects modeling, which is robust for family size and missing values. The performance of the new mLRC as selection tool was evaluated in an intensive simulation study and illustrated in a large real dataset, the Historical Sample of the Netherlands (HSN).

RESULTS

Empirical scores such as the FLOSS and LRC cannot properly deal with differential family size and missing data. Our simulation study showed that mLRC is not affected by family size and provides more accurate selections of long-lived families. The analysis of 1105 sibships of the Historical Sample of the Netherlands showed that the selection of long-lived individuals based on the mLRC score predicts excess survival in the validation set better than the selection based on the LRC score .

CONCLUSIONS

Model-based score systems such as the mLRC score help to reduce heterogeneity in the selection of long-lived families. The power of future studies into the genetics of longevity can likely be improved and their bias reduced, by selecting long-lived cases using the mLRC.

Contributions of amino acid, acylcarnitine and sphingolipid profiles to type 2 diabetes risk among South-Asian Surinamese and Dutch adults

INTRODUCTION

People of South Asian origin are at high risk of type 2 diabetes (T2D), but the underpinning mechanisms are not fully understood. We determined ethnic differences in acylcarnitine, amino acid and sphingolipid concentrations and determined the associations with T2D.

RESEARCH DESIGN AND METHODS

Associations between these metabolites and incident T2D among Dutch and South-Asian Surinamese were determined in participants from the Healthy Life in an Urban Setting (HELIUS) study (Amsterdam, the Netherlands) using Prentice-weighted Cox regression. The HELIUS study includes 95 incident T2D cases and a representative subcohort of 700 people from a cohort of 5977 participants with a mean follow-up of 4 years.

RESULTS

Concentrations of acylcarnitines were comparable between both ethnic groups. Amino acid and lactosylceramide concentrations were higher among South-Asian Surinamese than Dutch (eg, isoleucine 65.7 (SD 16.3) vs 60.7 (SD 15.6) µmol/L). Ceramide concentrations were lower among South-Asian Surinamese than Dutch (eg, Cer d18:1 8.48 (SD 2.04) vs 9.08 (SD 2.29) µmol/L). Metabolic dysregulation preceded T2D without evidence for a multiplicative interaction by ethnicity. Most amino acids and (dihydro)ceramides were associated with increased risk (eg, Cer d18:1 HR 2.38, 95% CI 1.81 to 3.12) while acylcarnitines, glycine, glutamine and lactosylceramides were associated with decreased risk for T2D (eg, LacCer d18:2 HR 0.56, 95% CI 0.42 to 0.77).

CONCLUSIONS

Overall, these data suggest that the disturbances underlying amino acid and sphingolipid metabolism may be predictive of T2D risk in populations of both South Asian and European background. These observations may be used as starting point to unravel the underlying metabolic disturbances.

Multiset sparse partial least squares path modeling for high dimensional omics data analysis

BACKGROUND

Recent technological developments have enabled the measurement of a plethora of biomolecular data from various omics domains, and research is ongoing on statistical methods to leverage these omics data to better model and understand biological pathways and genetic architectures of complex phenotypes. Current reviews report that the simultaneous analysis of multiple (i.e. three or more) high dimensional omics data sources is still challenging and suitable statistical methods are unavailable. Often mentioned challenges are the lack of accounting for the hierarchical structure between omics domains and the difficulty of interpretation of genomewide results. This study is motivated to address these challenges. We propose multiset sparse Partial Least Squares path modeling (msPLS), a generalized penalized form of Partial Least Squares path modeling, for the simultaneous modeling of biological pathways across multiple omics domains. msPLS simultaneously models the effect of multiple molecular markers, from multiple omics domains, on the variation of multiple phenotypic variables, while accounting for the relationships between data sources, and provides sparse results. The sparsity in the model helps to provide interpretable results from analyses of hundreds of thousands of biomolecular variables.

RESULTS

With simulation studies, we quantified the ability of msPLS to discover associated variables among high dimensional data sources. Furthermore, we analysed high dimensional omics datasets to explore biological pathways associated with Marfan syndrome and with Chronic Lymphocytic Leukaemia. Additionally, we compared the results of msPLS to the results of Multi-Omics Factor Analysis (MOFA), which is an alternative method to analyse this type of data.

CONCLUSIONS

msPLS is an multiset multivariate method for the integrative analysis of multiple high dimensional omics data sources. It accounts for the relationship between multiple high dimensional data sources while it provides interpretable results through its sparse solutions. The biomarkers found by msPLS in the omics datasets can be interpreted in terms of biological pathways associated with the pathophysiology of Marfan syndrome and of Chronic Lymphocytic Leukaemia. Additionally, msPLS outperforms MOFA in terms of variation explained in the chronic lymphocytic leukaemia dataset while it identifies the two most important clinical markers for Chronic Lymphocytic Leukaemia AVAILABILITY: http://uva.csala.me/mspls.https://github.com/acsala/2018_msPLS.

20-Year Follow-up of Statins in Children with Familial Hypercholesterolemia

BACKGROUND

Familial hypercholesterolemia is characterized by severely elevated low-density lipoprotein (LDL) cholesterol levels and premature cardiovascular disease. The short-term efficacy of statin therapy in children is well established, but longer follow-up studies evaluating changes in the risk of cardiovascular disease are scarce.

METHODS

We report a 20-year follow-up study of statin therapy in children. A total of 214 patients with familial hypercholesterolemia (genetically confirmed in 98% of the patients), who were previously participants in a placebo-controlled trial evaluating the 2-year efficacy and safety of pravastatin, were invited for follow-up, together with their 95 unaffected siblings. Participants completed a questionnaire, provided blood samples, and underwent measurements of carotid intima-media thickness. The incidence of cardiovascular disease among the patients with familial hypercholesterolemia was compared with that among their 156 affected parents.

RESULTS

Of the original cohort, 184 of 214 patients with familial hypercholesterolemia (86%) and 77 of 95 siblings (81%) were seen in follow-up; among the 214 patients, data on cardiovascular events and on death from cardiovascular causes were available for 203 (95%) and 214 (100%), respectively. The mean LDL cholesterol level in the patients had decreased from 237.3 to 160.7 mg per deciliter (from 6.13 to 4.16 mmol per liter) - a decrease of 32% from the baseline level; treatment goals (LDL cholesterol <100 mg per deciliter [2.59 mmol per liter]) were achieved in 37 patients (20%). Mean progression of carotid intima-media thickness over the entire follow-up period was 0.0056 mm per year in patients with familial hypercholesterolemia and 0.0057 mm per year in siblings (mean difference adjusted for sex, -0.0001 mm per year; 95% confidence interval, -0.0010 to 0.0008). The cumulative incidence of cardiovascular events and of death from cardiovascular causes at 39 years of age was lower among the patients with familial hypercholesterolemia than among their affected parents (1% vs. 26% and 0% vs. 7%, respectively).

CONCLUSIONS

In this study, initiation of statin therapy during childhood in patients with familial hypercholesterolemia slowed the progression of carotid intima-media thickness and reduced the risk of cardiovascular disease in adulthood. (Funded by the AMC Foundation.).

Multiset sparse redundancy analysis for high-dimensional omics data

Redundancy Analysis (RDA) is a well‐known method used to describe the directional relationship between related data sets. Recently, we proposed sparse Redundancy Analysis (sRDA) for high‐dimensional genomic data analysis to find explanatory variables that explain the most variance of the response variables. As more and more biomolecular data become available from different biological levels, such as genotypic and phenotypic data from different omics domains, a natural research direction is to apply an integrated analysis approach in order to explore the underlying biological mechanism of certain phenotypes of the given organism. We show that the multiset sparse Redundancy Analysis (multi‐sRDA) framework is a prominent candidate for high‐dimensional omics data analysis since it accounts for the directional information transfer between omics sets, and, through its sparse solutions, the interpretability of the result is improved. In this paper, we also describe a software implementation for multi‐sRDA, based on the Partial Least Squares Path Modeling algorithm. We test our method through simulation and real omics data analysis with data sets of 364,134 methylation markers, 18,424 gene expression markers, and 47 cytokine markers measured on 37 patients with Marfan syndrome.

TREatment of ATopic eczema (TREAT) Registry Taskforce: an international Delphi exercise to identify a core set of domains and domain items for national atopic eczema photo- and systemic therapy registries

BACKGROUND

Evidence of immunomodulatory therapies to guide clinical management of atopic eczema (AE) is scarce, despite frequent and often off-label use. Patient registries provide valuable evidence for the effects of treatments under real-world conditions that can inform treatment guidelines, give the opportunity for health economic evaluation and the evaluation of quality of care, as well as pharmacogenetic and dynamic research, which cannot be adequately addressed in clinical trials.

OBJECTIVES

The TREatment of ATopic eczema (TREAT) Registry Taskforce aims to seek international consensus on a core set of domains and items ('what to measure') for AE research registries, using a Delphi approach.

METHODS

Participants from six stakeholder groups were included: doctors, nurses, nonclinical researchers, patients, industry and regulatory body representatives. The eDelphi comprised three sequential online rounds, requesting participants to rate the importance of each proposed domain item. Participants could add domain items to the proposed list in round 1. A final consensus meeting was held to ratify the core set.

RESULTS

Participants (n = 479) from 36 countries accessed the eDelphi platform, of whom 86%, 79% and 74% completed rounds 1, 2 and 3, respectively. At the face-to-face consensus meeting attended by 42 participants the final core set was established containing 19 domains with 69 domain items (49 baseline and 20 follow-up items).

CONCLUSIONS

This core set of domains and items to be captured by national AE systemic therapy registries will standardize data collection and thereby allow direct comparability across registries and facilitate data pooling between countries. Ultimately, it will provide greater insight into the effectiveness, safety and cost-effectiveness of photo- and systemic immunomodulatory therapies.

Sparse redundancy analysis of high-dimensional genetic and genomic data

Motivation

Recent technological developments have enabled the possibility of genetic and genomic integrated data analysis approaches, where multiple omics datasets from various biological levels are combined and used to describe (disease) phenotypic variations. The main goal is to explain and ultimately predict phenotypic variations by understanding their genetic basis and the interaction of the associated genetic factors. Therefore, understanding the underlying genetic mechanisms of phenotypic variations is an ever increasing research interest in biomedical sciences. In many situations, we have a set of variables that can be considered to be the outcome variables and a set that can be considered to be explanatory variables. Redundancy analysis (RDA) is an analytic method to deal with this type of directionality. Unfortunately, current implementations of RDA cannot deal optimally with the high dimensionality of omics data (p≫n). The existing theoretical framework, based on Ridge penalization, is suboptimal, since it includes all variables in the analysis. As a solution, we propose to use Elastic Net penalization in an iterative RDA framework to obtain a sparse solution.

Results

We proposed sparse redundancy analysis (sRDA) for high dimensional omics data analysis. We conducted simulation studies with our software implementation of sRDA to assess the reliability of sRDA. Both the analysis of simulated data, and the analysis of 485 512 methylation markers and 18,424 gene-expression values measured in a set of 55 patients with Marfan syndrome show that sRDA is able to deal with the usual high dimensionality of omics data.

Availability and implementation

http://uva.csala.me/rda.

Contact

a.csala@amc.uva.nl.

Supplementary information

Supplementary data are available at Bioinformatics online.

Adaptive list sequential sampling method for population-based observational studies

Background

In population-based observational studies, non-participation and delayed response to the invitation to participate are complications that often arise during the recruitment of a sample. When both are not properly dealt with, the composition of the sample can be different from the desired composition. Inviting too many individuals or too few individuals from a particular subgroup could lead to unnecessary costs or decreased precision. Another problem is that there is frequently no or only partial information available about the willingness to participate. In this situation, we cannot adjust the recruitment procedure for non-participation before the recruitment period starts.

Methods

We have developed an adaptive list sequential sampling method that can deal with unknown participation probabilities and delayed responses to the invitation to participate in the study. In a sequential way, we evaluate whether we should invite a person from the population or not. During this evaluation, we correct for the fact that this person could decline to participate using an estimated participation probability. We use the information from all previously invited persons to estimate the participation probabilities for the non-evaluated individuals.

Results

The simulations showed that the adaptive list sequential sampling method can be used to estimate the participation probability during the recruitment period, and that it can successfully recruit a sample with a specific composition.

Conclusions

The adaptive list sequential sampling method can successfully recruit a sample with a specific desired composition when we have partial or no information about the willingness to participate before we start the recruitment period and when individuals may have a delayed response to the invitation.