Balanced Sample Generation for Predicting the Mean of a Finite Population

Two algorithms for balancing samples are proposed. They permit to disminish the errors of a predictor. A simple linear regression superpopulation model generates the responses. The behavior of the predictor under two balancing algorithms is evaluated in a Montecarlo experiment. The data arose from the determination of the content of minerals and vitamins in feed.     

Kynurenine metabolism predicts cognitive function in patients following cardiac bypass and thoracic surgery

Cardiac surgery involving extra-corporeal circulation can lead to cognitive dysfunction. As such surgery is associated with signs of inflammation and pro-inflammatory mediators activate tryptophan oxidation to neuroactive kynurenines which modulate NMDA receptor function and oxidative stress, we have measured blood concentrations of kynurenines and inflammatory markers in 28 patients undergoing coronary arterial graft surgery and, for comparison, 28 patients undergoing non-bypass thoracic surgery. A battery of cognitive tests was completed before and after the operations. The results show increased levels of tryptophan with decreased levels of kynurenine, anthranilic acid and 3-hydroxyanthranilic acid associated with bypass, and a later increase in kynurenic acid. Levels of neopterin and lipid peroxidation products rose after surgery in non-bypass patients whereas tumour necrosis factor-α and S100B levels increased after bypass. Changes of neopterin levels were greater after non-bypass surgery. Cognitive testing showed that the levels of tryptophan, kynurenine, kynurenic acid and the kynurenine/tryptophan ratio, correlated with aspects of post-surgery cognitive function, and were significant predictors of cognitive performance in tasks sensitive to frontal executive function and memory. Thus, anaesthesia and major surgery are associated with inflammatory changes and alterations in tryptophan oxidative metabolism which predict, and may play a role in, post-surgical cognitive function.     

Dataset size and composition impact the reliability of performance benchmarks for peptide-MHC binding predictions

Background

It is important to accurately determine the performance of peptide:MHC binding predictions, as this enables users to compare and choose between different prediction methods and provides estimates of the expected error rate. Two common approaches to determine prediction performance are cross-validation, in which all available data are iteratively split into training and testing data, and the use of blind sets generated separately from the data used to construct the predictive method. In the present study, we have compared cross-validated prediction performances generated on our last benchmark dataset from 2009 with prediction performances generated on data subsequently added to the Immune Epitope Database (IEDB) which served as a blind set.

Results

We found that cross-validated performances systematically overestimated performance on the blind set. This was found not to be due to the presence of similar peptides in the cross-validation dataset. Rather, we found that small size and low sequence/affinity diversity of either training or blind datasets were associated with large differences in cross-validated vs. blind prediction performances. We use these findings to derive quantitative rules of how large and diverse datasets need to be to provide generalizable performance estimates.

Conclusion

It has long been known that cross-validated prediction performance estimates often overestimate performance on independently generated blind set data. We here identify and quantify the specific factors contributing to this effect for MHC-I binding predictions. An increasing number of peptides for which MHC binding affinities are measured experimentally have been selected based on binding predictions and thus are less diverse than historic datasets sampling the entire sequence and affinity space, making them more difficult benchmark data sets. This has to be taken into account when comparing performance metrics between different benchmarks, and when deriving error estimates for predictions based on benchmark performance.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2105-15-241) contains supplementary material, which is available to authorized users.     

Plant families as predictors of plant biodiversity in Mexico

Three speciose plant families (Asteraceae, Fabaceae s.l. , and Poaceae) are evaluated for their appropriateness as predictors of overall regional patterns of plant diversity in Mexico. These families fulfil criteria for usefulness as indicators advocated by several authors, such as adequate taxonomic knowledge, wide geographical distribution, and occurrence in different plant communities. Known richness was determined from a database including 17,903 species from floristic inventories, review of taxonomic literature, and the study of herbarium specimens. Mexico was divided into 253 cells, each 1° latitude by 1° longitude; for each cell, total plant diversity and number of species in each predictor's family were determined. In addition, each species was categorized by 'preferred' vegetation type (temperate forest, dry tropical forest, humid tropical forest, xerophytic scrubland). Both multiple and simple regression models show that the three families are good predictors of total vascular floristic richness. Fabaceae s.l. has the highest semipartial correlations for all multiple models except with the xerophytic scrubland data set, for which it shares the highest values with Poaceae. Univariate models also predict satisfactorily the vascular floristic richness, especially when Fabaceae s.l. is used as the predictor, and the effect of vegetation type is included. Our results indicate that these three families can be used as potential predictors for total vascular plant species richness in Mexico although Fabaceae s.l. is the best predictor.     

Ovarian reserve tests and their utility in predicting response to controlled ovarian stimulation in rhesus monkeys

Controlled ovarian stimulation (COS) is an alternative to natural breeding in nonhuman primates; however, these protocols are costly with no guarantee of success. Toward the objective of predicting COS outcome in rhesus monkeys, this study evaluated three clinically used ovarian reserve tests (ORTs): day 3 (d3) follicle‐stimulating hormone (FSH) with d3 inhibin B (INHB), the clomiphene citrate challenge test (CCCT), and the exogenous FSH Ovarian Reserve Test. A COS was also performed and response was classified as either successful (COS+) or unsuccessful (COS?) and retrospectively compared with ORT predictions. FSH and INHB were assessed for best hormonal index in conjunction with the aforementioned tests. INHB was consistently more accurate than FSH in all the ORTs used. Overall, a modified version of the CCCT using INHB values yielded the best percentage of correct predictions. This is the first report of ORT evaluation in rhesus monkeys and may provide a useful diagnostic test before costly follicle stimulations, as well as predicting the onset of menopause. Am. J. Primatol. 72:672–680, 2010. © 2010 Wiley‐Liss, Inc.     

Comparing the effects of continuous and discrete covariate mismeasurement,with emphasis on the dichotomization of mismeasured predictors

It is well known that imprecision in the measurement of predictor variables typically leads to bias in estimated regression coefficients. We compare the bias induced by measurement error in a continuous predictor with that induced by misclassification of a binary predictor in the contexts of linear and logistic regression. To make the comparison fair, we consider misclassification probabilities for a binary predictor that correspond to dichotomizing an imprecise continuous predictor in lieu of its precise counterpart. On this basis, nondifferential binary misclassification is seen to yield more bias than nondifferential continuous measurement error. However, it is known that differential misclassification results if a binary predictor is actually formed by dichotomizing a continuous predictor subject to nondifferential measurement error. When the postulated model linking the response and precise continuous predictor is correct, this differential misclassification is found to yield less bias than continuous measurement error, in contrast with nondifferential misclassification, i.e., dichotomization reduces the bias due to mismeasurement. This finding, however, is sensitive to the form of the underlying relationship between the response and the continuous predictor. In particular, we give a scenario where dichotomization involves a trade-off between model fit and misclassification bias. We also examine how the bias depends on the choice of threshold in the dichotomization process and on the correlation between the imprecise predictor and a second precise predictor.