A Data-Analytic Strategy for Protein-Biomarker Discovery: Profiling of High-Dimensional Proteomic Data for Cancer Detection

With recent advances in mass spectrometry techniques, it is now possible to investigate proteins over a wide range of molecular weights in small biological specimens. This advance has generated data-analytic challenges in proteomics, similar to those created by microarray technologies in genetics, namely, discovery of "signature" protein profiles specific to each pathologic state (e.g., normal vs. cancer) or differential profiles between experimental conditions (e.g., treated by a drug of interest vs. untreated) from high-dimensional data. We propose a data analytic strategy for discovering protein biomarkers based on such high-dimensional mass-spectrometry data. A real biomarker-discovery project on prostate cancer is taken as a concrete example throughout the paper: the project aims to identify proteins in serum that distinguish cancer, benign hyperplasia, and normal states of prostate using the Surface Enhanced Laser Desorption/Ionization (SELDI) technology, a recently developed mass spectrometry technique. Our data analytic strategy takes properties of the SELDI mass-spectrometer into account: the SELDI output of a specimen contains about 48,000 (x, y) points where x is the protein mass divided by the number of charges introduced by ionization and y is the protein intensity of the corresponding mass per charge value, x, in that specimen. Given high coefficients of variation and other characteristics of protein intensity measures (y values), we reduce the measures of protein intensities to a set of binary variables that indicate peaks in the y-axis direction in the nearest neighborhoods of each mass per charge point in the x-axis direction. We then account for a shifting (measurement error) problem of the x-axis in SELDI output. After these pre-analysis processing of data, we combine the binary predictors to generate classification rules for cancer, benign hyperplasia, and normal states of prostate. Our approach is to apply the boosting algorithm to select binary predictors and construct a summary classifier. We empirically evaluate sensitivity and specificity of the resulting summary classifiers with a test dataset that is independent from the training dataset used to construct the summary classifiers. The proposed method performed nearly perfectly in distinguishing cancer and benign hyperplasia from normal. In the classification of cancer vs. benign hyperplasia, however, an appreciable proportion of the benign specimens were classified incorrectly as cancer. We discuss practical issues associated with our proposed approach to the analysis of SELDI output and its application in cancer biomarker discovery.

Initiation of high-frequency oscillatory ventilation and its effects upon cerebral circulation in pigs: an experimental study

BACKGROUND: Current practice at high-frequency oscillatory ventilation (HFOV) initiation is a stepwise increase of the constant applied airway pressure to achieve lung recruitment. We hypothesized that HFOV would lead to more adverse cerebral haemodynamics than does pressure controlled ventilation (PCV) in the presence of experimental intracranial hypertension (IH) and acute lung injury (ALI) in pigs with similar mean airway pressure settings. METHODS: In 12 anesthetized pigs (24-27 kg) with IH and ALI, mean airway pressure (P(mean)) was increased (to 20, 25, 30 cm H(2)O every 30 min), either with HFOV or with PCV. The order of the two ventilatory modes (cross-over) was randomized. Mean arterial pressure (MAP), intracranial pressure (ICP), cerebral perfusion pressure (CPP), cerebral blood flow (CBF) (fluorescent microspheres), cerebral metabolism, transpulmonary pressures (P(T)), and blood gases were determined at each P(mean) setting. Our end-points of interest related to the cerebral circulation were ICP, CPP and CBF. RESULTS: CBF and cerebral metabolism were unaffected but there were no differences between the values for HFOV and PCV. ICP increased slightly (HFOV median +1 mm Hg, P<0.05; PCV median +2 mm Hg, P<0.05). At P(mean) setting of 30 cm H(2)O, CPP decreased during HFOV (median -13 mm Hg, P<0.05) and PCV (median -17 mm Hg, P<0.05) paralleled by a decrease of MAP (HFOV median -11 mm Hg, P<0.05; PCV median -13 mm Hg, P<0.05). P(T) increased (HFOV median +8 cm H(2)O, P<0.05; PCV median +8 cm H(2)O, P<0.05). Oxygenation improved and normocapnia maintained by HFOV and PCV. There were no differences between both ventilatory modes. CONCLUSIONS: In animals with elevated ICP and ALI, both ventilatory modes had effects upon cerebral haemodynamics. The effects upon cerebral haemodynamics were dependent of the P(T) level without differences between both ventilatory modes at similar P(mean) settings. HFOV seems to be a possible alternative ventilatory strategy when MAP deterioration can be avoided.

Effects of high-dose heroin versus morphine in intravenous drug users: a randomised double-blind crossover study

The purpose of this study is to evaluate the effects of high doses of injected opiates as prescribed maintenance in intravenous drugs users. This was accomplished via a randomised double-blind study with crossover at an outpatient clinic in Bern, Switzerland. The subjects were 39 patients with a long history of intravenous opioid use and persistent abuse despite treatment; they were randomly allocated to two groups. Group A was started on controlled injection of graduated doses of morphine up to a satisfying individual dose and was then switched as a double blind to heroin at a randomly determined day between week three and four. Subsequently this group was given heroin for the remaining two to three weeks of the study. Group B was started on heroin and was then switched to morphine in the same manner. Equipotent solutions of 3% morphine and 2% heroin were administered. The main outcome measures were clinical observations, structural interviews and self report of subjective experiences to assess the effects of the drugs. In 16 cases, the study had to be discontinued owing to severe morphine-induced histamine reactions. Thirteen participants in Group B presented these adverse reactions on the day of the switch-over. Full data were thus only obtainable for 17 participants. Average daily doses were 491 mg for heroin and 597 mg for morphine. The findings indicate that heroin significantly produced a lower grade of itching, flushing, urticaria and pain/nausea. A negative correlation between dose and euphoria was observed for both heroin and morphine. The authors concluded that as heroin produces fewer side effects it is the preferred high-dose maintenance prescription to morphine. The perceived euphoric effects are limited in both substances.

EXPLORATION, NORMALIZATION, AND GENOTYPE CALLS OF HIGH DENSITY OLIGONUCLEOTIDE SNP ARRAY DATA

In most microarray technologies, a number of critical steps are required to convert raw intensity measurements into the data relied upon by data analysts, biologists and clinicians. These data manipulations, referred to as preprocessing, can influence the quality of the ultimate measurements. In the last few years, the high-throughput measurement of gene expression is the most popular application of microarray technology. For this application, various groups have demonstrated that the use of modern statistical methodology can substantially improve accuracy and precision of gene expression measurements, relative to ad-hoc procedures introduced by designers and manufacturers of the technology. Currently, other applications of microarrays are becoming more and more popular. In this paper we describe a preprocessing methodology for a technology designed for the identification of DNA sequence variants in specific genes or regions of the human genome that are associated with phenotypes of interest such as disease. In particular we describe methodology useful for preprocessing Affymetrix SNP chips and obtaining genotype calls with the preprocessed data. We demonstrate how our procedure improves existing approaches using data from three relatively large studies including one in which large number independent calls are available. Software implementing these ideas are avialble from the Bioconductor oligo package.

Exposure modeling of high-frequency electromagnetic fields

We developed a geospatial model that calculates ambient high-frequency electromagnetic field (HF-EMF) strengths of stationary transmission installations such as mobile phone base stations and broadcast transmitters with high spatial resolution in the order of 1 m. The model considers the location and transmission patterns of the transmitters, the three-dimensional topography, and shielding effects by buildings. The aim of the present study was to assess the suitability of the model for exposure monitoring and for epidemiological research. We modeled time-averaged HF-EMF strengths for an urban area in the city of Basel as well as for a rural area (Bubendorf). To compare modeling with measurements, we selected 20 outdoor measurement sites in Basel and 18 sites in Bubendorf. We calculated Pearson's correlation coefficients between modeling and measurements. Chance-corrected agreement was evaluated by weighted Cohen's kappa statistics for three exposure categories. Correlation between measurements and modeling of the total HF-EMF strength was 0.67 (95% confidence interval (CI): 0.33-0.86) in the city of Basel and 0.77 (95% CI: 0.46-0.91) in the rural area. In both regions, kappa coefficients between measurements and modeling were 0.63 and 0.77 for the total HF-EMF strengths and for all mobile phone frequency bands. First evaluation of our geospatial model yielded substantial agreement between modeling and measurements. However, before the model can be applied for future epidemiologic research, additional validation studies focusing on indoor values are needed to improve model validity.Journal of Exposure Science and Environmental Epidemiology (2008) 18, 183-191; doi:10.1038/sj.jes.7500575; published online 4 April 2007.

Effects of high-yield thrombocytapheresis on the quality of platelet products

BACKGROUND: The steadily increasing demands for single-donor apheresis platelet (PLT) concentrates (APCs) are a challenge to the PLT supply system. Therefore, efforts to improve plateletpheresis yield, allowing apheresis products to be split into 2 or more units, are valuable strategies. No data to demonstrate in vivo transfusion efficacy of these high-yield split-APCs are currently available, however. STUDY DESIGN AND METHODS: The transfusion efficacy of APCs produced by two apheresis methods involving different harvest and storing procedures and varying PLT yields was investigated. Efficacy measures were the 1-hour percent PLT recovery (PPR(1h)) and the 1-hour corrected count increment (CCI(1h)). In total, 400 APCs, produced with either an Amicus device (Baxter) and stored in PLT additive solution (T-Sol; Amicus method [AM], n = 107) or a Trima device (Gambro) and stored in plasma (Trima method [TM], n = 293), were transfused to 55 children (31 girls; median age, 9.5 years; range, 0.2-18.5 years) with thrombocytopenia due to chemotherapy or aplastic anemia (median, 4 APCs per child; range, 1-68). RESULTS: Transfusion efficacy was significantly lower for AM-APCs than for TM-APCs (median PPR(1h), 17 and 33%; median CCI(1h), 7.9 and 15.6, respectively; p < 0.001). Reduced transfusion efficacy correlated in a yield-dependent manner with high apheresis PLT yields (>/=6 x 10(11)) for AM-APCs (p < 0.001). CONCLUSION: Although in vitro validation of AM- and TM-APCs has been performed, only by evaluating transfusion efficacy in vivo did the AM turn out to be not suitable for high-yield thrombocytapheresis. This study recommends the implementation of in vivo transfusion efficacy studies for high-yield APC apheresis donations.

Volume dependence of high-frequency respiratory mechanics in healthy adults

Respiratory system input impedance (Zrs) at low to medium frequencies below 100 Hz, and study of its volume dependence, have been used extensively to quantify airway and tissue mechanics. Zrs at high oscillation frequencies including the first antiresonant frequency (far,1) may contain important information about airway mechanics. Changes in high-frequency Zrs with lung volume have not been studied. The volume-dependent behavior of high-frequency Zrs, specifically far,1 and respiratory system resistance at first antiresonance (Rrs(far,1)), was characterized in 16 healthy adults. Zrs was measured with a forced oscillation signal (5-302.5 Hz) through a wavetube setup. To track Zrs, subjects performed slow deep inspiratory and expiratory maneuvers over 30-s measurements, during which average impedance was calculated over 0.4-s intervals, with successive overlapping estimates every 0.156 s. Flow was measured using a pneumotachometer and integrated to obtain volume. Transpulmonary pressure dependence (Ptp) of Zrs was separately determined in five subjects. Both far,1 and Rrs(far,1) decreased with increasing lung volume and Ptp, consistent with an increase in airway caliber and decreased airway wall compliance as volume increased. These characterizations provide insight into airway mechanics, and are furthermore a necessary first step toward determining whether volume dependence of the first antiresonance is altered in disease.