PubMed Health⌕ Search

SEARCH · PubMed Health

Results for “Model interpretability”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 631 records · Page 35Linked to original sources
PubMed2004

Dielectric spectroscopy and electrophoretic mobility measurements interpreted with the standard electrokinetic model.

We report results from complementary electrokinetic measurements-dielectric relaxation and electrophoretic mobility-undertaken to test the applicability of the standard electrokinetic theory with a model system. Dielectric spectra were obtained at frequencies between 1 kHz and 40 MHz with a new, two-electrode cell design [Hollingsworth and Saville, J. Colloid Interface Sci. 257 (2003) 65-76]; mobility data were acquired with an electrophoretic light scattering instrument. Data from the two-electrode cell were collected at different electrode separations and interpreted with newly developed procedures to remove the influence of electrode polarization. Methodology A employs extrapolation to infinite electrode separation to compute the dielectric constant and conductivity as functions of frequency. The contributions from suspended particles are reported in terms of dielectric constant and conductivity increments. Methodology B uses a theoretical model of electrode polarization and the standard electrokinetic model in a nonlinear regression scheme. Results are presented in several forms: frequency-dependent dielectric constant and conductivity increments, frequency-dependent dielectric constants and conductivities, and the complex dipole coefficient. It is shown that the standard model provides a consistent methodology for interpreting particle behavior; zeta-potentials inferred from mobility and dielectric relaxation agree to within experimental error. Moreover, the cell design and interpretation are straightforward and provide relatively simple ways to obtain complementary measurements over a wide frequency range. The results unambiguously show that electrokinetic character of this dispersion follows the standard model.

Journal Article↗
PubMed2001

Cumulative logit models for ordinal data: a case study involving allergic rhinitis severity scores.

Ordered categorical data arise in numerous settings, a common example being pain scores in analgesic trials. The modelling of such data is intrinsically more difficult than the modelling of continuous data due to the constraints on the underlying probabilities and the reduced amount of information that discrete outcomes contain. In this paper we discuss the class of cumulative logit models, which provide a natural framework for ordinal data analysis. We show how viewing the categorical outcome as the discretization of an underlying continuous response allows a natural interpretation of model parameters. We also show how covariates are incorporated into the model and how various types of correlation among repeated measures on the same individual may be accounted for. The models are illustrated using longitudinal allergy data consisting of sneezing scores measured on a four-point scale. Our approach throughout is Bayesian and we present a range of simple diagnostics to aid model building.

Antibodies, Monoclonal↗
PubMed1999

Matching theory and experiment in protein folding.

There has been considerable progress made over the past year in linking experimental and theoretical approaches to protein folding. Recent results from several independent lines of investigation suggest that protein folding mechanisms and landscapes are largely determined by the topology of the native state and are relatively insensitive to details of the interatomic interactions. This dependence on low-resolution structural features, rather than high-resolution detail, suggests that it should be possible to describe the fundamental physics of the folding process using relatively low-resolution models. Recent experiments have set benchmarks for testing new models and progress has been made in developing theoretical models for interpreting and predicting experimental results.

Diffusion↗
PubMed2002

Development of reference ranges in elite athletes for markers of altered erythropoiesis.

BACKGROUND AND OBJECTIVES: Our previous research developed two statistical models that are useful indicators of current (ON-model) or recently discontinued (OFF-model) recombinant human erythropoietin (rHuEPO) use by athletes. The component variables of the ON-model are hematocrit (Hct), reticulocyte hematocrit (RetHct), serum erythropoietin (EPO), percent macrocytes (%Macro), and soluble transferrin receptor (sTfr), whilst the OFF-model uses only the first three variables. Genetics and training modalities of elite athletes may conceivably produce unusual values for blood parameters related to erythropoiesis. The aims of this study were to develop reference ranges in elite athletes for key hematologic parameters as well as ON- and OFF-models scores, and to evaluate the effect of ethnicity, gender, residence at moderate altitude (approximately 2000 m) and within-individual variation on the variables and model scores. DESIGN AND METHODS: Over a period of three weeks, 413 female and 739 male elite athletes from 12 countries visited laboratories to provide three blood samples for analysis of blood parameters sensitive to erythropoiesis. For each parameter and for the ON- and OFF-model scores, we used mixed modeling to establish the range within which we could be 95% certain that the value for a randomly chosen athlete would fall, taking into account various random effects (variation within and between subjects and laboratories) and fixed effects (means for different levels of ethnicity, age, sport, altitude of residency). We performed similar analyses for changes in the ON- and OFF-model scores between the three visits. RESULTS: Most fixed effects were accompanied by clear-cut, small to moderate differences in several parameters. However, residency at moderate altitude was accompanied by a much higher hematocrit than residency nearer sea level, with the mean (and 95% confidence limits) for the difference being 2.3 (0.9 to 3.7) and 1.8 (0.1 to 3.5) units for males and females, respectively. Males at altitude also demonstrated a moderately higher ON-model score. Otherwise the influence of these effects was small for ON-, OFF- and changes in model scores. INTERPRETATION AND CONCLUSIONS: Assessment of an athlete's blood parameters and ON- and OFF-model scores may need adjustment for training modalities and other characteristics of the subject. Changes in model scores (together with monitoring of urine samples for the presence of rHuEPO) provide a promising approach to detection of rHuEPO abuse, because they are less sensitive to subject characteristics and less variable than raw model scores.

Adolescent↗
PubMed2005

Limitations of ASM1 and ASM3: a comparison based on batch oxygen uptake rate profiles from different full-scale wastewater treatment plants.

The two most popular models for the description of the biological COD removal are ASM1 and ASM3. However, some numerical inconsistencies arise when using these models to interpret the data obtained in short-term respirometric batch experiments. In this study, both models are fitted to four different respirometric batch profiles obtained with biomass from different WWTP. The parameter estimation results and the practical (local) identifiability are analysed, and the limitations of both models are discussed. The growth yield obtained by fitting ASM1 to the short-term respirometric batch profiles is higher than the default one, as well as the storage yield obtained by fitting ASM3 is lower than the default one. Based on these values, possible improvements to the modelling of the biological COD removal, such as the inclusion of simultaneous growth and storage on external substrate, are proposed.

Biodegradation, Environmental↗
PubMed2026

Predicting food taste with bound-driven optimization.

The prediction of sensory attributes from ingredient-level formulations is an emerging challenge at the intersection of food science and artificial intelligence. We address the fundamental question of whether the taste of a food can be predicted from its ingredients by treating recipes as composite materials. We apply Hashin-Shtrikman (HS) and Reuss-Voigt (RV) bounds, techniques originally developed for elastic moduli, as a null-hypothesis additive baseline for five taste dimensions (sweetness, sourness, bitterness, umami, saltiness) on a curated dataset of 70 recipes decomposed into 115 distinct ingredients scored against a library of 209 ingredient-level taste references with trained-panel ground truth. This baseline systematically under-predicts perceived taste: 77% of actual taste values exceeded the HS upper bound, with the exceedance rate ranging from 26% (bitterness) to 97% (saltiness). We traced this gap to specific processing chemistry (Maillard reactions, caramelization, evaporative concentration, protein hydrolysis, and nucleotide synergy) and introduced a hybrid model that augments the HS baseline with eight chemistry-proxy features encoding these mechanisms. Our results show that our interpretable hybrid model eliminates the systematic bias and reduces mean absolute error by 27%-62% for sweetness, sourness, umami, and saltiness while using only 10 interpretable features, achieving performance comparable to a black-box Lasso regression on 115 per-ingredient features. We further demonstrate constrained inverse design via Differential Evolution, recovering ingredient formulations that match target taste profiles subject to compositional bounds. Our work demonstrates how key chemical processes during food preparation can inform and augment physics-based and machine learning models, providing a quantitative fingerprint of processing chemistry's contribution to taste perception and paving the way for model-driven food formulation with targeted sensory characteristics.

Composite material bounds↗
PubMed2003

Conformational flexibility of the group B meningococcal polysaccharide in solution.

To elucidate the role of secondary structure in the immune response against alpha(2-->8)-linked polysialic acid, the capsular polysaccharide of Group B meningococci, we have investigated its solution dynamics by using specific models of molecular motion and hydrodynamic modeling to interpret experimental NMR data. (13)C-[(1)H] NMR relaxation times and steady-state NOE enhancements were measured for two aqueous solutions of alpha(2-->8)-linked sialic acid polysaccharides. Each contained a unique distribution of polysaccharide chain lengths, with average lengths estimated at 40 or 400 residues. Models for rigid molecule tumbling, including two based on helical conformations proposed for the polysaccharide,(31) could not explain the NMR measurements. In general for these helices, the correlation times for their overall tumbling that best account for the NMR data correspond to polysaccharide chains between 9 and 18 residues in length, far short of the average lengths estimated for either solution. The effects of internal motions incorporated into these helices was modeled with an effective correlation time representing helix tumbling as well as internal motion. This modeling demonstrated that even with extreme amounts of internal motion, "flexible helices" of 25 residues or more still could not produce the NMR measurements. All data are consistent with internal and segmental motions dominating the nuclear magnetic relaxation of the polysaccharide and not molecular tumbling. Statistical distributions of correlation times have been found specifically for the pyranose rings, linkage groups, and methoxy groups that can account for the measured relaxation times and NOE enhancements. The distributions suggest that considerable flexibility attends the polysaccharide in solution, and the ranges of motional frequencies for the linkage groups and pyranose rings are comparable. We conclude that the Group B meningococcal polysaccharide is a random coil chain in solution, and therefore, does not have antigenic epitopes dependent upon a rigid, ordered conformation.

Bacterial Capsules↗
PubMed2005

Patient-doctor decision-making about treatment within the consultation--a critical analysis of models.

This paper highlights some of the limitations of models of patient involvement in decision-making and explores the reasons for, and implications of, these limitations. Taking the three models of interpretative, shared and informed decision-making as examples, we focus on two limitations of the models: (1) neglect of which decisions the patient should be involved in (the framing problem) and (2) how the patient should be involved in decision-making (the nature of reasoning problem). Although there will inevitably be a gap between models and practice--this much is in the nature of the models--we suggest that these two issues are substantially neglected by the models and yet are fundamental to understanding patient-doctor decision-making. We also suggest that the fundamental problem that lies behind these limitations is insufficient attention to, and explicitness about, the dilemmas of professional ethics, which are played out in the professional-patient relationships that the models are supposed to represent, particularly with respect to the issue of expert and lay accountability.

Attitude of Health Personnel↗
PubMed2004

Sources and dechlorination of polychlorinated biphenyl congeners in the sediments of Fox River, Wisconsin.

Polychlorinated biphenyl (PCB) congeners were analyzed in eight deep, dated sediment cores collected from the immediate upstream and downstream of DePere dam in Fox River, Wisconsin. The average time span of the cores is about 100 yr, except for one core (FR-9) which is influenced by mixing or covers a short time period (2 yr). The total PCB concentrations have a range of 0.2-6.8 ppm for the upstream and 0.3-17.6 ppm for the downstream cores. The PCB data obtained from the sampling were analyzed as upstream and downstream data, using a factor analysis (FA) model with nonnegative constraints to identify PCB sources and congener patterns. The factor loadings obtained from the FA model were interpreted in terms of the presence of possible environmental degradation mechanisms. In addition, a recently developed model, which is used to identify and quantify possible pathways of anaerobic dechlorination of PCBs in the sediments, was validated on the basis of in situ data from the literature, and then applied to the congener patterns obtained from the FA model. The major PCB source to the Fox River sediments is identified as Aroclor 1242, for both the upstream and the downstream sediments. Loss of di- and trichlorobiphenyls (e.g., 2-4, 25-2) from the sediments suggests desorption from the sediments. On the other hand, observation of elevated amounts of certain congeners such as 24-3 and 25-3 indicates the presence of anaerobic dechlorination activity. The anaerobic dechlorination model demonstrates significant similarities between the Aroclor 1242 profile altered according to dechlorination processes Q (upstream) and H' (downstream) and the dominant congener patterns obtained from the FA model.

Bacteria, Anaerobic↗
PubMed2005

Structure-based design of novel Chk1 inhibitors: insights into hydrogen bonding and protein-ligand affinity.

We report the discovery, synthesis, and crystallographic binding mode of novel furanopyrimidine and pyrrolopyrimidine inhibitors of the Chk1 kinase, an oncology target. These inhibitors are synthetically tractable and inhibit Chk1 by competing for its ATP site. A chronological account allows an objective comparison of modeled compound docking modes to the subsequently obtained crystal structures. The comparison provides insights regarding the interpretation of modeling results, in relationship to the multiple reasonable docking modes which may be obtained in a kinase-ATP site. The crystal structures were used to guide medicinal chemistry efforts. This led to a thorough characterization of a pair of ligand-protein complexes which differ by a single hydrogen bond. An analysis indicates that this hydrogen bond is expected to contribute a fraction of the 10-fold change in binding affinity, adding a valuable observation to the debate about the energetic role of hydrogen bonding in molecular recognition.

Adenosine Triphosphate↗
PubMed1994

Novel Methods of Measuring Hydraulic Conductivity of Tree Root Systems and Interpretation Using AMAIZED (A Maize-Root Dynamic Model for Water and Solute Transport).

Steady-state and dynamic methods were used to measure the conductivity to water flow in large woody root systems. The methods were destructive in that the root must be excised from the shoot but do not require removal of the root from the soil. The methods involve pushing water from the excised base of the root to the apex, causing flow in a direction opposite to that during normal transpiration. Sample data are given for two tropical (Cecropia obtusifolia and Lacistema aggregatum) and two temperate species (Acer saccharum and Juglans regia cv Lara). A hysteresis was observed in the relationship between applied pressure and resulting flow during dynamic measurements. A mathematical model (AMAIZED) was derived for the dynamics of solute and water flow in roots. The model was used to interpret results obtained from steady-state and dynamic measurements. AMAIZED is mathematically identical with the equations that describe Munch pressure flow of solute and water in the phloem of leaves. Results are discussed in terms of the predictions of AMAIZED, and suggestions for the improvement of methods are made.

Journal Article↗
PubMed1995

Physiological relevance of T-tube model parameters with emphasis on arterial compliances.

The T-tube model of systemic arterial circulation was examined with respect to the physiological relevance of model parameters. root aortic pressure [Pao(t)] and flow [Qao(t)] and descending aortic flow [Qb(t)] were measured in anesthetized, open-chest dogs under control conditions, during inflation of a balloon positioned in the left external iliac artery (n = 5), and during infusion of vasoactive drugs nitroprusside (NTP, n = 4) and phenylephrine (PHL, n = 5). With Pao(t) as the input, the model accurately predicted both Qao(t) and Qb(t) under all conditions (r2 > 0.96). The balloon inflation data established the ability of the model to discriminate between proximal and distal arterial mechanical properties. Furthermore, proximal properties (i.e., tube characteristic impedances and transit times) were independent of distal properties such as terminal compliances and resistances (or equivalently, wave reflections). The effects of NTP and PHL were pharmacologically consistent and served to further validate this model. NTP primarily affected distal (load) properties, whereas PHL altered both load and tube parameters. Physiological interpretation of model parameters, particularly compliance, is also discussed. The ability of the model to correctly discriminate between proximal and distal arterial properties is relevant because these properties may affect cardiovascular function differently.

Animals↗
PubMed2005

Interpreting spatial and temporal neural activity through a recurrent neural network brain-machine interface.

We propose the use of optimized brain-machine interface (BMI) models for interpreting the spatial and temporal neural activity generated in motor tasks. In this study, a nonlinear dynamical neural network is trained to predict the hand position of primates from neural recordings in a reaching task paradigm. We first develop a method to reveal the role attributed by the model to the sampled motor, premotor, and parietal cortices in generating hand movements. Next, using the trained model weights, we derive a temporal sensitivity measure to asses how the model utilized the sampled cortices and neurons in real-time during BMI testing.

Algorithms↗
PubMed2000

Fundamental theorem of natural selection and frequency-dependent selection: analysis of the matrix game diploid model.

Following Ewens' interpretation about Fisher's fundamental theorem of natural selection, the matrix game model for diploid populations undergoing non-overlapping, discrete generations is investigated. The total genetic variance is decomposed and it is shown that the partial change in the mean fitness, which is equal to the additive genetic variance over the mean fitness, can be thought of as a change due only to the partial changes in the phenotypic frequencies.

Animals↗
PubMed1997

Anatomical variability and functional ability of vascular trees modeled by constrained constructive optimization.

The aim of this study was to investigate the extent that functional capability of vascular trees is related to anatomical variability. To these ends we used the method of Constrained Constructive Optimization (CCO) to generate optimized computer models of coronary arterial trees. All these model trees were optimized according to the same principle under equal boundary conditions of pressures and flows. However, by stochastically casting the locations of the terminal segments, different tree structures were generated, each of which conformed to the same boundary conditions. The structural variability of these models was interpreted as the correlate of the anatomical variability found in real arterial trees. The advantage that computer model trees are known in numerical detail was exploited to perform comprehensive and exact classifications of all segments into bifurcation levels, STRAHLER orders and composite vessels, and to compute the area expansion ratio. The unexpected result was that, despite striking visual differences in anatomical structure, the model trees were almost identical with regard to functional performance. We conclude that models optimized on the computer for a given perfusion task show little differences in their morphometric parameters even if they differ considerably regarding the course of the large vessels.

Computer Simulation↗
PubMed1994

Famine, revolt, and the dynastic cycle: population dynamics in historic China.

"Historians have long noticed that population declines in ancient China often coincided with dynasty changes, and that most of these declines were the result of internecine wars which, in turn, were often initiated by famine or density pressure. Since the interactions between density pressure, internecine wars, and dynasty changes cannot be explained by the traditional age-specific density-dependent population structure, we propose to use a bandit/peasant/ruler occupation-specific population model to interpret the dynamic socio-economic transitions of ancient Chinese population, and provide econometric support to our model. We also highlight the rich dynamics of the composition of human population, a factor which was often neglected in previous research on general populations."

Asia↗
PubMed1998

Crystal structure of Escherichia coli cystathionine gamma-synthase at 1.5 A resolution.

The transsulfuration enzyme cystathionine gamma-synthase (CGS) catalyses the pyridoxal 5'-phosphate (PLP)-dependent gamma-replacement of O-succinyl-L-homoserine and L-cysteine, yielding L-cystathionine. The crystal structure of the Escherichia coli enzyme has been solved by molecular replacement with the known structure of cystathionine beta-lyase (CBL), and refined at 1.5 A resolution to a crystallographic R-factor of 20.0%. The enzyme crystallizes as an alpha4 tetramer with the subunits related by non-crystallographic 222 symmetry. The spatial fold of the subunits, with three functionally distinct domains and their quaternary arrangement, is similar to that of CBL. Previously proposed reaction mechanisms for CGS can be checked against the structural model, allowing interpretation of the catalytic and substrate-binding functions of individual active site residues. Enzyme-substrate models pinpoint specific residues responsible for the substrate specificity, in agreement with structural comparisons with CBL. Both steric and electrostatic designs of the active site seem to achieve proper substrate selection and productive orientation. Amino acid sequence and structural alignments of CGS and CBL suggest that differences in the substrate-binding characteristics are responsible for the different reaction chemistries. Because CGS catalyses the only known PLP-dependent replacement reaction at Cgamma of certain amino acids, the results will help in our understanding of the chemical versatility of PLP.

Binding Sites↗