Missing value imputation in high-dimensional phenomic data: imputable or not,and how?

Background

In modern biomedical research of complex diseases, a large number of demographic and clinical variables, herein called phenomic data, are often collected and missing values (MVs) are inevitable in the data collection process. Since many downstream statistical and bioinformatics methods require complete data matrix, imputation is a common and practical solution. In high-throughput experiments such as microarray experiments, continuous intensities are measured and many mature missing value imputation methods have been developed and widely applied. Numerous methods for missing data imputation of microarray data have been developed. Large phenomic data, however, contain continuous, nominal, binary and ordinal data types, which void application of most methods. Though several methods have been developed in the past few years, not a single complete guideline is proposed with respect to phenomic missing data imputation.

Results

In this paper, we investigated existing imputation methods for phenomic data, proposed a self-training selection (STS) scheme to select the best imputation method and provide a practical guideline for general applications. We introduced a novel concept of “imputability measure” (IM) to identify missing values that are fundamentally inadequate to impute. In addition, we also developed four variations of K-nearest-neighbor (KNN) methods and compared with two existing methods, multivariate imputation by chained equations (MICE) and missForest. The four variations are imputation by variables (KNN-V), by subjects (KNN-S), their weighted hybrid (KNN-H) and an adaptively weighted hybrid (KNN-A). We performed simulations and applied different imputation methods and the STS scheme to three lung disease phenomic datasets to evaluate the methods. An R package “phenomeImpute” is made publicly available.

Conclusions

Simulations and applications to real datasets showed that MICE often did not perform well; KNN-A, KNN-H and random forest were among the top performers although no method universally performed the best. Imputation of missing values with low imputability measures increased imputation errors greatly and could potentially deteriorate downstream analyses. The STS scheme was accurate in selecting the optimal method by evaluating methods in a second layer of missingness simulation. All source files for the simulation and the real data analyses are available on the author’s publication website.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-014-0346-6) contains supplementary material, which is available to authorized users.     

Chromosome cohesion: ring around the sisters?

Sister chromatid cohesion is a key aspect of accurate chromosome transmission during mitosis, yet little is known about the structure of cohesin, the protein complex that links the two sister chromatids. Recent studies shed light on the structure of the cohesin complex, leading to intriguing models that could explain how sister chromatids are held together.     

A bell pepper cultivar tolerant to chilling enhanced nitrogen allocation and stress‐related metabolite accumulation in the roots in response to low root‐zone temperature

Two bell pepper (Capsicum annuum) cultivars, differing in their response to chilling, were exposed to three levels of root‐zone temperatures. Gas exchange, shoot and root phenology, and the pattern of change of the central metabolites and secondary metabolites caffeate and benzoate in the leaves and roots were profiled. Low root‐zone temperature significantly inhibited gaseous exchange, with a greater effect on the sensitive commercial pepper hybrid (Canon) than on the new hybrid bred to enhance abiotic stress tolerance (S103). The latter was less affected by the treatment with respect to plant height, shoot dry mass, root maximum length, root projected area, number of root tips and root dry mass. More carbon was allocated to the leaves of S103 than nitrogen at 17°C, while in the roots at 17°C, more nitrogen was allocated and the ratio between C/N decreased. Metabolite profiling showed greater increase in the root than in the leaves. Leaf response between the two cultivars differed significantly. The roots accumulated stress‐related metabolites including γ‐aminobutyric acid (GABA), proline, galactinol and raffinose and at chilling (7°C) resulted in an increase of sugars in both cultivars. Our results suggest that the enhanced tolerance of S103 to root cold stress, reflected in the relative maintenance of shoot and root growth, is likely linked to a more effective regulation of photosynthesis facilitated by the induction of stress‐related metabolism.     

A Missense Mutation in a Highly Conserved Region of CASQ2 Is Associated with Autosomal Recessive Catecholamine-Induced Polymorphic Ventricular Tachycardia in Bedouin Families from Israel

Catecholamine-induced polymorphic ventricular tachycardia (PVT) is characterized by episodes of syncope, seizures, or sudden death, in response to physical activity or emotional stress. Two modes of inheritance have been described: autosomal dominant and autosomal recessive. Mutations in the ryanodine receptor 2 gene (RYR2), which encodes a cardiac sarcoplasmic reticulum (SR) Ca(2+)-release channel, were recently shown to cause the autosomal dominant form of the disease. In the present report, we describe a missense mutation in a highly conserved region of the calsequestrin 2 gene (CASQ2) as the potential cause of the autosomal recessive form. The CASQ2 protein serves as the major Ca(2+) reservoir within the SR of cardiac myocytes and is part of a protein complex that contains the ryanodine receptor. The mutation, which is in full segregation in seven Bedouin families affected by the disorder, converts a negatively charged aspartic acid into a positively charged histidine, in a highly negatively charged domain, and is likely to exert its deleterious effect by disrupting Ca(2+) binding.     

Differentiation-Induced Internal Translation of c-sis mRNA: Analysis of the cis Elements and Their Differentiation-Linked Binding to the hnRNP C Protein

In previous reports we showed that the long 5' untranslated region (5' UTR) of c-sis, the gene encoding the B chain of platelet-derived growth factor, has translational modulating activity due to its differentiation-activated internal ribosomal entry site (D-IRES). Here we show that the 5' UTR contains three regions with a computer-predicted Y-shaped structure upstream of an AUG codon, each of which can confer some degree of internal translation by itself. In nondifferentiated cells, the entire 5' UTR is required for maximal basal IRES activity. The elements required for the differentiation-sensing ability (i.e., D-IRES) were mapped to a 630-nucleotide fragment within the central portion of the 5' UTR. Even though the region responsible for IRES activation is smaller, the full-length 5' UTR is capable of mediating the maximal translation efficiency in differentiated cells, since only the entire 5' UTR is able to confer the maximal basal IRES activity. Interestingly, a 43-kDa protein, identified as hnRNP C, binds in a differentiation-induced manner to the differentiation-sensing region. Using UV cross-linking experiments, we show that while hnRNP C is mainly a nuclear protein, its binding activity to the D-IRES is mostly nuclear in nondifferentiated cells, whereas in differentiated cells such binding activity is associated with the ribosomal fraction. Since the c-sis 5' UTR is a translational modulator in response to cellular changes, it seems that the large number of cross-talking structural entities and the interactions with regulated trans-acting factors are important for the strength of modulation in response to cellular changes. These characteristics may constitute the major difference between strong IRESs, such as those seen in some viruses, and IRESs that serve as translational modulators in response to developmental signals, such as that of c-sis.     

Reduced hepatic injury in Toll-like receptor 4-deficient mice following D-galactosamine/lipopolysaccharide-induced fulminant hepatic failure

Liver transplantation is the only therapy of proven benefit in fulminant hepatic failure (FHF). Lipopolysaccharide (LPS), D-galactosamine (GalN)-induced FHF is a well established model of liver injury in mice. Toll-Like Receptor 4 (TLR4) has been identified as a receptor for LPS. The aim of this study was to investigate the role of TLR4 in FHF induced by D-GalN/LPS administration in mice. Wild type (WT) and TLR4 deficient (TLR4ko) mice were studied in vivo in a fulminant model induced by GalN/LPS. Hepatic TLR4 expression, serum liver enzymes, hepatic and serum TNF-α and interleukin-1β levels were determined. Apoptotic cells were identified by immunohistochemistry for caspase-3. Nuclear factor-kappaβ (NF-κ β) and phosphorylated c-Jun hepatic expression were studied using Western blot analysis. All WT mice died within 24 hours after administration of GalN/LPS while all TLR4ko mice survived. Serum liver enzymes, interleukin-1β, TNF-α level, TLR4 mRNA expression, hepatic injury and hepatocyte apoptosis all significantly decreased in TLR4ko mice compared with WT mice. A significant decrease in hepatic c-Jun and IκB signaling pathway was noted in TLR4ko mice compared with WT mice. In conclusion, following induction of FHF, the inflammatory response and the liver injury in TLR4ko mice was significantly attenuated through decreased hepatic c-Jun and NF-κB expression and thus decreased TNF-α level. Down-regulation of TLR4 expression plays a pivotal role in GalN/LPS induced FHF. These findings might have important implications for the use of the anti TLR4 protein signaling as a potential target for therapeutic intervention in FHF.     

Human and porcine early kidney precursors as a new source for transplantation

Kidney transplantation has been one of the major medical advances of the past 30 years. However, tissue availability remains a major obstacle. This can potentially be overcome by the use of undifferentiated or partially developed kidney precursor cells derived from early embryos and fetal tissue. Here, transplantation in mice reveals the earliest gestational time point at which kidney precursor cells, of both human and pig origin, differentiate into functional nephrons and not into other, non-renal professional cell types. Moreover, successful organogenesis is achieved when using the early kidney precursors, but not later-gestation kidneys. The formed, miniature kidneys are functional as evidenced by the dilute urine they produce. In addition, decreased immunogenicity of the transplants of early human and pig kidney precursors compared with adult kidney transplants is demonstrated in vivo. Our data pinpoint a window of human and pig kidney organogenesis that may be optimal for transplantation in humans.