Discussion of "the evolution of boosting algorithms" and "extending statistical boosting".

This article is part of a For-Discussion-Section of Methods of Information in Medicine about the papers "The Evolution of Boosting Algorithms - From Machine Learning to Statistical Modelling" and "Extending Statistical Boosting - An Overview of Recent Methodological Developments", written by Andreas Mayr and co-authors. It is introduced by an editorial. This article contains the combined commentaries invited to independently comment on the Mayr et al. papers. In subsequent issues the discussion can continue through letters to the editor.

Influence of calcium and phosphorus, lactose, and salt-to-moisture ratio on Cheddar cheese quality: pH buffering properties of cheese.

The pH buffering capacity of cheese is an important determinant of cheese pH. However, the effects of different constituents of cheese on its pH buffering capacity have not been fully clarified. The objective of this study was to characterize the chemical species and chemical equilibria that are responsible for the pH buffering properties of cheese. Eight cheeses with 2 levels of Ca and P (0.67 and 0.47% vs. 0.53 and 0.39%, respectively), residual lactose (2.4 vs. 0.78%), and salt-to-moisture ratio (6.4 vs. 4.8%) were manufactured. The pH-titration curves for these cheeses were obtained by titrating cheese:water (1:39 wt/wt) dispersions with 1 N HCl, and backtitrating with 1 N NaOH. To understand the role of different chemical equilibria and the respective chemical species in controlling the pH of cheese, pH buffering was modeled mathematically. The 36 chemical species that were found to be relevant for modeling can be classified as cations (Na+, Ca2+, Mg2+), anions (phosphate, citrate, lactate), protein-bound amino acids with a side-chain pKa in the range of 3 to 9 (glutamate, histidine, serine phosphate, aspartate), metal ion complexes (phosphate, citrate, and lactate complexes of Na+, Ca2+, and Mg2+), and calcium phosphate precipitates. A set of 36 corresponding equations was solved to give the concentrations of all chemical species as a function of pH, allowing the prediction of buffering curves. Changes in the calculated species concentrations allowed the identification of the chemical species and chemical equilibria that dominate the pH buffering properties of cheese in different pH ranges. The model indicates that pH buffering in the pH range from 4.5 to 5.5 is predominantly due to a precipitate of Ca and phosphate, and the protonation equilibrium involving the side chains of protein-bound glutamate. In the literature, the precipitate is often referred to as amorphous colloidal calcium phosphate. A comparison of experimental data and model predictions shows that the buffering properties of the precipitate can be explained, assuming that it consists of hydroxyapatite [Ca5(OH)(PO4)3] or Ca3(PO4)2. The pH buffering in the region from pH 3.5 to 4.5 is due to protonation of side-chain carboxylates of protein-bound glutamate, aspartate, and lactate, in order of decreasing significance. In addition, pH buffering between pH 5 to 8 in the backtitration results from the reprecipitation of calcium and phosphate either as CaHPO4 or Ca4H(PO4)3.

Influence of natural, electrically neutral lipids on the potentiometric responses of cation-selective polymeric membrane electrodes.

Ionophore-free ion exchanger electrodes were found to exhibit quite a high selectivity for the creatininium ion; however, measurements in diluted urine samples revealed large emf drifts. Potentiometric, chromatographic, NMR, and mass spectrometric evidence did not reveal any major cationic interfering agents, and anionic interfering agents cannot trivially explain the consistently positive emf drifts. Ultrafiltration of urine samples showed that the interfering agents have molecular weights below 1000 u. The drifts are apparently caused by electrically neutral lipophilic compounds of low molecular weight that are easily extracted into organic phases. Follow-up experiments showed that p-cresol and cholesterol cause no significant emf responses but that coproporphyrin, phosphatidylserine, taurocholic acid, cholic acid, phosphatidylethanolamine, and octanoic acid cause positive emf drifts of the type that was observed with the urine samples. The extent of the responses and the response time depend not only on the specific compound but also on the cation in the sample solution. These results suggest that the emf drifts are due to extraction of such natural lipids into the organic membrane phase where they interact in an ionophore-like fashion with the analyte and interfering ions. Changes in the potentiometric selectivities after contact with natural lipids support this interpretation. The same effect of natural lipids is also expected for ionophore-based electrodes. Indeed, exposure of a valinomycin-based electrode to a methylene chloride extract of urine resulted in a significant reduction of the Na+ discrimination, increasing log Kpot(K,Na) from -3.9 to -3.1.

Scanning tunneling microscopy with chemically modified tips: discrimination of porphyrin centers based on metal coordination and hydrogen bond interactions.

STM gold tips chemically modified with 4-mercaptopyridine (4MP) were found capable of discriminating zinc(II) 5,15-bis(4-octadecyloxyphenyl)porphyrin (Por-Zn) from its metal-free porphyrin (Por-2H) and nickel(II) complexes (Por-Ni) in the mixed monolayers of these compounds, spontaneously formed at the solution/graphite interface. The porphyrin centers in STM images observed with 4MP-modified tips exhibited bright spots, while those measured with unmodified tips exhibited the porphyrin centers as dark depressions. The centers of Por-Zn were brighter than those of Por-2H and Por-Ni, thereby allowing the discrimination of Por-Zn from Por-2H or Por-Ni in mixed monolayers. The changes in the contrasts of porphyrin centers of Por-2H and Por-Zn/ Por-Ni were explained by facilitated electron tunneling due to hydrogen bond and metal coordination interactions, respectively, between porphyrin centers and the pyridyl group of 4MP on the tip.

Origin of non-Nernstian anion response slopes of metalloporphyrin-based liquid/polymer membrane electrodes.

The origin of the non-Nernstian potentiometric anion response behavior exhibited by several metalloporphyrin-based liquid/polymeric membrane electrodes is examined. UV-visible spectrophotometry of organic-phase solutions and thin plasticized PVC films containing In(III) and Ga(III) octaethylporphyrins suggests that, in the absence of preferred axial coordination anions, the metalloporphyrins form hydroxide ion bridged dimers within the organic phases, as indicated by a significant blue shift of the Soret band in the visible spectrum. As increasing levels of the preferred anions are added, the degree of dimerization decreases and the intensity of the Soret band corresponding to the monomer species increases. Observation of Nernstian responses with membranes doped with picket fence-type In(III) and Ga(III) porphyrins not capable of forming hydroxide bridged structures (as determined by UV-visible spectroscopy) confirms that dimerization is likely responsible for the super-Nernstian slopes of membrane electrodes formulated with the non-picket fence species. A phase boundary model based on simultaneous binding equilibria of hydroxide ions with two metalloporphyrins to form the dimeric species, while the target anions bind with metalloporphyrins to form neutral 1:1 complexes, is shown to fully predict the observed non-Nernstian behavior. The prospect of utilizing this anion-dependent dimer-monomer metalloporphyrin equilibrium to fabricate anion-selective optical sensors using thin films of metalloporphyrin-doped polymers is also discussed.

Lifetime of ion-selective electrodes based on charged ionophores

The lifetime of solvent polymeric ion-selective electrodes (ISEs) is limited by leaching of the membrane components into the sample solutions. In this article, leaching of charged ionophores is discussed. Because of the electroneutrality principle, the loss of the charged ionophore into the sample must be accompanied by parallel transport of an ion of the opposite charge sign into the sample or by ion exchange with a sample ion of the same charge sign. Because ionic sites of high lipophilicity are available, the loss of ionic sites is, in general, not a concern. Therefore, it is assumed here that the cotransported or ion-exchanging ions are primary or interfering ions forming complexes with the ionophore. A general theory that allows quantification of ionophore lipophilicities and a discussion of changes in the membrane composition and selectivity with time is presented. A high complex stability and high analyte concentrations diminish the rate of ionophore loss into the sample if a charged ionophore is coextracted from the membrane into the sample together with an analyte ion of opposite charge. On the other hand, if the charged ionophore has the same charge sign as the ion that it binds, a large binding constant and high analyte concentrations enhance ionophore leaching into the sample. The model is applied to interpret results for an electrically charged ionophore, for which selectivity changes as a function of the leaching time were observed and the lipophilicity was determined with potentiometric measurements. Using the lipophilicities of neutral ionophores, as described previously, and the lipophilicities of charged ionophores, as described here, a direct comparison of the expected leaching rates of charged and neutral ionophores has become possible.

Cationic or anionic sites? Selectivity optimization of ion-selective electrodes based on charged ionophores

The influence of ionic sites on the selectivities of ionophore-based ion-selective electrodes (ISEs) is described on the basis of a phase boundary potential model. The discussion presented here is significantly more general than previous ones. It is formulated for primary and interfering ions of any charges and it is valid for ISEs based on electrically charged or neutral ionophores. Furthermore, it also applies to membranes that contain more than one type of complex of the primary or interfering ion. It has been believed thus far that only ionic sites of the same charge sign as the primary ion improve the selectivities of ISEs based on charged ionophores. However, it is shown here that the charge sign of the ionic sites that give the highest potentiometric selectivities depends on the charge number of the primary and interfering ions and on the stoichiometry of their complexes with the ionophore. The validity of our model was confirmed experimentally with three ISEs based on different charged ionophores. ISEs based on lasalocid or 1-(N,N-dicyclohexylcarbamoyl)-2- (N,N-dioctadecylcarbamoyl)ethylphosphonic acid monomethyl ester (ETH 5639) as the ionophore responded selectively to Sr2+ or Mg2+, respectively, and discriminated well against other alkaline earth cations when their membranes contained anionic sites. These two electrodes are the first examples of ISEs based on charged ionophores for which maximum selectivities are obtained with membranes containing ionic sites with a charge sign opposite to that of the primary ion. On the other hand, the experimental F- selectivities of membranes based on oxo(5,10,15,20-tetraphenylporphyrinato)molybdenum-(V) improved gradually when the concentration of anionic sites was increased from 0 to 75 mol%. The selectivity-modifying influence of ionic sites for these three types of ISEs can be explained by considering the different stabilities of the 1:2 ion-ionophore complexes of the primary and of the interfering ions.

Selectivity of potentiometric ion sensors.

Selectivities of solvent polymeric membrane ion-selective electrodes (ISEs) are quantitatively related to equilibria at the interface between the sample and the electrode membrane. However, only correctly determined selectivity coefficients allow accurate predictions of ISE responses to real-world samples. Moreover, they are also required for the optimization of ionophore structures and membrane compositions. Best suited for such purposes are potentiometric selectivity coefficients as defined already in the 1960s. This paper briefly reviews the basic relationships and focuses on possible biases in the determination of selectivity coefficients. The traditional methods to determine selectivity coefficients (separate solution method, fixed interference method) are still the same as those originally proposed by IUPAC in 1976. However, several precautions are needed to obtain meaningful data. For example, errors arise when the response to a weakly interfering ion is also influenced by the primary ion leaching from the membrane. Wrong selectivity coefficients may be also obtained when the interfering agent is highly preferred and the electrode shows counterion interference. Recent advances show how such pitfalls can be avoided. A detailed recipe to determine correct potentiometric selectivity coefficients unaffected by such biases is presented.

An ion-selective electrode for acetate based on a urea-functionalized porphyrin as a hydrogen-bonding ionophore.

An ion-selective electrode for acetate based on (α,α,α,α)-5,10,15,20-tetrakis[2-(4-fluorophenylureylene)phenyl]porphyrin as an ionophore that has no metal center and forms hydrogen bonds to the analyte is described. At pH 7.0 (0.1 M HEPES-NaOH buffer), the electrode based on this ionophore and cationic sites (50 mol % relative to the ionophore) responds to acetate in a linear range from 1.58 × 10(-)(4) to 1.58 × 10(-)(2) M with a slope of -54.8 ± 0.8 mV/decade and a detection limit of (3.06 ± 1.15) × 10(-)(5) M. Selectivity coefficients determined with the separate solution method (SSM) indicate that interferences of hydrophobic inorganic anions are relatively small (log[Formula: see text] (SSM): NO(3)(-), +0.68; SCN(-), +0.60; NO(2)(-), +0.22; I(-), +0.20; ClO(4)(-), +0.12; Br(-), -0.13). Responses to anions that are good hydrogen bond acceptors, i.e., Cl(-), HSO(3)(-), and HCO(3)(-), were Nernstian and were weaker than the response to acetate (log[Formula: see text] (SSM): -0.54, -0.56, and -1.34, respectively). Negligibly small responses were observed for very hydrophilic anions, i.e., F(-), SO(4)(2)(-), and H(2)PO(4)(-)/HPO(4)(2)(-). While aliphatic carboxylates such as formate, propanoate, pyruvate, and lactate gave Nernstian responses similar to acetate, interferences of salicylate and benzoate were considerably decreased in comparison with electrodes based on cationic sites only. Concentrations of acetic acid in vinegar samples were determined by direct potentiometry and agreed with values determined by a standard enzymatic method.

Polypyrrole-modified tips for functional group recognition in scanning tunneling microscopy.

Tailored chemical modification of scanning tunneling microscopy (STM) tips is a promising method for the recognition of specific chemical species and functional groups in STM images. The present study shows for the first time that tips modified with polypyrrole can be used to measure STM images with molecular resolution. A high conductivity of the polypyrrole film was found to be important for the observation of STM images, while the thickness of the polymer film did not affect the images significantly. Furthermore, it was shown that recognition of functional groups in STM images is possible with tips coated with conductive polypyrroles. 1-Octadecanol and 1-octadecanoic acid monolayers with polypyrrole-modified tips gave high-resolution STM images in which aligned OH and COOH residues were represented by easily recognizable elevated bands. These selective contrast enhancements resemble those observed by us previously with gold tips modified with self-assembled monolayers (SAMs) and seem to be due to hydrogen bond interactions between functional groups of the tip-modifying molecules and the sample. The reproducibility of contrast enhancements in this study was significantly higher than for SAM-modified tips, suggesting that polymer modification of STM tips is particularly promising for specific functional group recognition with chemically modified STM tips.

A phase boundary potential model for apparently "twice-nernstian" responses of liquid membrane ion-selective electrodes.

A model that describes divalent cation responses of liquid membrane ion-selective electrodes based on acidic ionophores and ionic sites is presented. Response slopes for membranes with ionophore and anionic sites are predicted to change from Nernstian to apparently "twice-Nernstian" and then back to Nernstian again as the pH of the sample solution decreases. A maximum measuring range for apparently "twice-Nernstian" responses is expected for membranes with 50 mol % anionic sites relative to the ionophore. On the other hand, membranes with ionophore and cationic sites are expected to give only Nernstian responses, either to divalent cations at high pH or to H(+) at low pH. The validity of the present model has been confirmed experimentally with the two Ba(2+)-selective carboxylate ionophores monensin and lasalocid and the Ca(2+)-selective organophosphate ionophore bis(2-heptylundecyl) phosphate. Addition of anionic sites gave apparently "twice-Nernstian" slopes for monensin at pH 7.0 (56.6 mV/decade), for lasalocid at pH 4.0 (53.3 mV/decade), and for bis(2-heptylundecyl) phosphate at pH 3.5 (53.6 mV/decade). Membranes with cationic sites showed only pH responses at the respective pH. The apparently "twice-Nernstian" responses as discussed here are the first examples of super-Nernstian responses that can be explained with a quantitative model based on thermodynamic equilibria.

What is a linear process?

We argue that given even an infinitely long data sequence, it is impossible (with any test statistic) to distinguish perfectly between linear and nonlinear processes (including slightly noisy chaotic processes). Our approach is to consider the set of moving-average (linear) processes and study its closure under a suitable metric. We give the precise characterization of this closure, which is unexpectedly large, containing nonergodic processes, which are Poisson sums of independent and identically distributed copies of a stationary process. Proofs of these results will appear elsewhere.