Variable selection and pattern recognition with gene expression data generated by the microarray technology.

Lack of adequate statistical methods for the analysis of microarray data remains the most critical deterrent to uncovering the true potential of these promising techniques in basic and translational biological studies. The popular practice of drawing important biological conclusions from just one replicate (slide) should be discouraged. In this paper, we discuss some modern trends in statistical analysis of microarray data with a special focus on statistical classification (pattern recognition) and variable selection. In addressing these issues we consider the utility of some distances between random vectors and their nonparametric estimates obtained from gene expression data. Performance of the proposed distances is tested by computer simulations and analysis of gene expression data on two different types of human leukemia. In experimental settings, the error rate is estimated by cross-validation, while a control sample is generated in computer simulation experiments aimed at testing the proposed gene selection procedures and associated classification rules.

Direct calorimetry for the measurement of heat release in preterm infants: methods and applications.

Direct calorimetry is a sensitive and accurate method for the measurement of biologic heat release in humans. At the Children's Medical Center of Brooklyn, State University of New York, we have established direct calorimetry for the measurement of heat release by low birth weight premature infants. We have tested the method and find it to be simple, safe, and accurate. We studied heat release in 10 low birth weight infants on 22 occasions. The smallest infant in the study group weighed 1.43 kg. All the infant underwent direct calorimetry between 1 week and 18 weeks of age. Heat release in the infants ranged from 1.31 kcal/kg/hr. This method of direct calorimetry offers a tool for measuring total metabolic heat release from the first weeks of life in very low birth weight infants to estimate the insensible water losses and to examine the effect of various feeding regimens and disease states on total heat release.

A stochastic model of radiation carcinogenesis: latent time distributions and their properties.

A stochastic model of radiation carcinogenesis is proposed that has much in common with the ideas suggested by M. Pike as early as 1966. The model allows us to obtain a parametric family of substochastic-type distributions for the time of tumor latency that provides a description of the rate of tumor development and the number of affected individuals. With this model it is possible to interpret data on tumor incidence in terms of promotion and progression processes. The basic model is developed for a prolonged irradiation at a constant dose rate and includes short-term irradiation as a special case. A limiting form of the latent time distribution for short-term irradiation at high doses is obtained. This distribution arises in the extreme value theory within the random minima framework. An estimate for the rate of convergence to a limiting distribution is given. Based on the proposed latent time distributions, long-term predictions of carcinogenic risk do not call for information about irradiation dose. As shown by computer simulation studies and real data analysis, the parametric estimation of carcinogenic risk appears to be robust to the loss of statistical information caused by the right-hand censoring of time-to-tumor observations. It seems likely that this property, although revealed by means of a purely empirical procedure, may be useful in selecting a model for the practical purpose of risk prediction.