Comparison of statistical methods and the use of quality control samples for batch effect correction in human transcriptome data.

Batch effects are technical sources of variation introduced by the necessity of conducting gene expression analyses on different dates due to the large number of biological samples in population-based studies. The aim of this study is to evaluate the performances of linear mixed models (LMM) and Combat in batch effect removal. We also assessed the utility of adding quality control samples in the study design as technical replicates. In order to do so, we simulated gene expression data by adding "treatment" and batch effects to a real gene expression dataset. The performances of LMM and Combat, with and without quality control samples, are assessed in terms of sensitivity and specificity while correcting for the batch effect using a wide range of effect sizes, statistical noise, sample sizes and level of balanced/unbalanced designs. The simulations showed small differences among LMM and Combat. LMM identifies stronger relationships between big effect sizes and gene expression than Combat, while Combat identifies in general more true and false positives than LMM. However, these small differences can still be relevant depending on the research goal. When any of these methods are applied, quality control samples did not reduce the batch effect, showing no added value for including them in the study design.

Meeting report: moving upstream-evaluating adverse upstream end points for improved risk assessment and decision-making.

BACKGROUND: Assessing adverse effects from environmental chemical exposure is integral to public health policies. Toxicology assays identifying early biological changes from chemical exposure are increasing our ability to evaluate links between early biological disturbances and subsequent overt downstream effects. A workshop was held to consider how the resulting data inform consideration of an "adverse effect" in the context of hazard identification and risk assessment. OBJECTIVES: Our objective here is to review what is known about the relationships between chemical exposure, early biological effects (upstream events), and later overt effects (downstream events) through three case studies (thyroid hormone disruption, antiandrogen effects, immune system disruption) and to consider how to evaluate hazard and risk when early biological effect data are available. DISCUSSION: Each case study presents data on the toxicity pathways linking early biological perturbations with downstream overt effects. Case studies also emphasize several factors that can influence risk of overt disease as a result from early biological perturbations, including background chemical exposures, underlying individual biological processes, and disease susceptibility. Certain effects resulting from exposure during periods of sensitivity may be irreversible. A chemical can act through multiple modes of action, resulting in similar or different overt effects. CONCLUSIONS: For certain classes of early perturbations, sufficient information on the disease process is known, so hazard and quantitative risk assessment can proceed using information on upstream biological perturbations. Upstream data will support improved approaches for considering developmental stage, background exposures, disease status, and other factors important to assessing hazard and risk for the whole population.

Absolute estimation of initial concentrations of amplicon in a real-time RT-PCR process.

BACKGROUND: Since real time PCR was first developed, several approaches to estimating the initial quantity of template in an RT-PCR reaction have been tried. While initially only the early thermal cycles corresponding to exponential duplication were used, lately there has been an effort to use all of the cycles in a PCR. The efforts have included both fitting empirical sigmoid curves and more elaborate mechanistic models that explore the chemical reactions taking place during each cycle. The more elaborate mechanistic models require many more parameters than can be fit from a single amplification, while the empirical models provide little insight and are difficult to tailor to specific reactants. RESULTS: We directly estimate the initial amount of amplicon using a simplified mechanistic model based on chemical reactions in the annealing step of the PCR. The basic model includes the duplication of DNA with the digestion of Taqman probe and the re-annealing between previously synthesized DNA strands of opposite orientation. By modelling the amount of Taqman probe digested and matching that with the observed fluorescence, the conversion factor between the number of fluorescing dye molecules and observed fluorescent emission can be estimated, along with the absolute initial amount of amplicon and the rate parameter for re-annealing. The model is applied to several PCR reactions with known amounts of amplicon and is shown to work reasonably well. An expanded version of the model allows duplication of amplicon without release of fluorescent dye, by adding 1 more parameter to the model. The additional process is helpful in most cases where the initial primer concentration exceeds the initial probe concentration. Software for applying the algorithm to data may be downloaded at http://www.niehs.nih.gov/research/resources/software/pcranalyzer/ CONCLUSION: We present proof of the principle that a mechanistically based model can be fit to observations from a single PCR amplification. Initial amounts of amplicon are well estimated without using a standard solution. Using the ratio of the predicted initial amounts of amplicon from 2 PCRs is shown to work well even when the absolute amounts of amplicon are underestimated in the individual PCRs.

Application of a statistical dynamic model investigating the short-term cellular kinetics induced by riddelliine, a hepatic endothelial carcinogen.

In recent studies, riddelliine, a pyrrolizidine alkaloid, was found to increase rates of replication and apoptosis and induce hemangiosarcoma in the liver of rats and mice. To analyze DNA replication and apoptosis data taken from the same animals, we have developed a predictive mathematical model for describing BrdU labeling and apoptotic processes. The model allows the incorporation of simple diurnal patterns in cellular kinetics and is applied to data on hepatocytes and endothelial cells taken from riddelliine exposed rats. Predictions from the model were used with multivariable nonlinear regression techniques to estimate replication and apoptotic rate constants for both cell types and all treatment groups. Hypothesis tests were used with the predicted rates to separate the competing effects of riddelliine on replication and apoptosis of hepatocytes and endothelial cells as well as compare replication rates between cell types. That estimated replication rates were found to be significantly higher for endothelial cells supports the supposition of induction of hemangiosarcoma by riddelliine in the liver.