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 1,315 records · Page 73Linked to original sources

PubMed1992

Overoxidized polypyrrole films: a model for the design of permselective electrodes.

Studies have been conducted to further characterize overoxidized polypyrrole films grown on carbon electrodes. Differences in apparent diffusion coefficients for various probes are interpreted via a model based on both hydrophobic and electrostatic interactions between the probe and polymer film. These electrode coatings are also compared to Naflon films, and the results of this study are discussed in terms of important considerations for the design of permselective electrodes.

Carbon↗

PubMed2006

Modelling strategies for the industrial exploitation of lactic acid bacteria.

Lactic acid bacteria (LAB) have a long tradition of use in the food industry, and the number and diversity of their applications has increased considerably over the years. Traditionally, process optimization for these applications involved both strain selection and trial and error. More recently, metabolic engineering has emerged as a discipline that focuses on the rational improvement of industrially useful strains. In the post-genomic era, metabolic engineering increasingly benefits from systems biology, an approach that combines mathematical modelling techniques with functional-genomics data to build models for biological interpretation and--ultimately--prediction. In this review, the industrial applications of LAB are mapped onto available global, genome-scale metabolic modelling techniques to evaluate the extent to which functional genomics and systems biology can live up to their industrial promise.

Biomass↗

PubMed2025

HallmarkGraph: a cancer hallmark informed graph neural network for classifying hierarchical tumor subtypes.

MOTIVATION: Accurate tumor subtype diagnosis is crucial for precision oncology, yet current methodologies face significant challenges. These include balancing model accuracy with interpretability and the high costs of generating multi-omics data in clinical settings. Moreover, there is a lack of validated models capable of classifying hierarchical tumor subtypes across a comprehensive pan-cancer cohort. RESULTS: We present a graph neural network, HallmarkGraph, the first biologically informed model developed to classify hierarchical tumor subtypes in human cancer. Inspired by cancer hallmarks, the model's architecture integrates transcriptome profiles and gene regulatory interactions to perform multi-label classification. We evaluate the model on a comprehensive pan-cancer cohort comprising 11 476 samples from 26 primary cancers with 405 subtypes up to eight levels. The model demonstrates exceptional performance, achieving 5-fold cross-validation accuracy between 85% and 99% for tumor subtypes labeled with increasing details of genomic information. It also shows good generalizability on a validation dataset of 887 samples, assessed using three metrics that consider tumor subtypes at individual, combined, and sample levels. Benchmarking and ablation experiments show that hallmark-based embeddings slightly influence model performance, while the integrated multilayer perceptron plays a significant role in determining classifier accuracy. Additionally, we use the SHAP method to link cancer hallmarks with genes, identifying key features that influence model decisions. Our findings present a biologically informed machine learning framework capable of tracking tumor transcriptomic trajectories and distinguishing inter- and intra-tumor heterogeneity in pan-cancer. This approach holds promise for enhancing cancer diagnostics. AVAILABILITY AND IMPLEMENTATION: HallmarkGraph is accessible at https://github.com/laixn/HallmarkGraph.

Humans↗

PubMed1995

Analysis of cancer rates using excess risk age-period-cohort models.

BACKGROUND: Recently the age-period-cohort (APC) model has become a popular epidemiological tool. However, it is well known that the model suffers from the identifiability problem. The simple multiplicative formulation of the model in terms of the age, period, and cohort variables without resorting to the underlying biology also casts doubt on the interpretability of the model parameters. METHODS: Excess risk APC models for cancers are developed based on carcinogenesis processes in human populations. These models have the beneficial feature of biological plausibility and do not suffer from the identifiability problem. Apart from the age, period, and cohort effects, a new kind of effect, the impact effect, is also introduced into the models. A computer program has been developed to fit the models which contain non-linear as well as restricted parameters. RESULTS: Two published mortality datasets are used to demonstrate the methodology. The proposed models fit better than the conventional APC model in both examples. CONCLUSIONS: Despite all the merits of the proposed models, several statistical issues should be investigated further before accepting this methodology as a general data-analytical tool.

Adult↗

PubMed1997

Model with distributed vectorial premotor bursters accounts for the component stretching of oblique saccades.

During oblique saccades, the durations of the horizontal and vertical components are stretched until they are approximately equal. Models of the saccadic system have been proposed that provide a mechanism for that stretching. However, they fail to simulate the pattern of activity recorded from premotor medium lead burst neurons (MLBNs) in the brain stem. A new model of the saccadic system is proposed that accounts for both the component stretching of oblique movements and the pattern of activity recorded in MLBNs. MLBNs that project to horizontal (or vertical) motoneurons actually have a wide span of on-directions (the direction associated with the largest discharge) around the cardinal direction. We infer from the wide span of their on-directions that, at the level of individual MLBNs, the vectorial signal present in spatially organized structures (e.g., the superior colliculus) is not decomposed into the separate horizontal and vertical components represented by the motoneurons. Nonetheless, all prior models of the saccadic system have decomposed the vectorial premotor command into horizontal and vertical commands at the level of the MLBNs. That decomposition was explicit, because individual MLBNs, with a sine- or cosine-shaped directional tuning curve, were used. We propose here that the decomposition into horizontal and vertical commands is carried out only at the level of the motoneurons. This decomposition is implicit, because no single MLBN encodes the horizontal or vertical command; the command only exists implicitly in the activity of the population of MLBNs. The new vectorial burster model correctly simulates the pattern of activity recorded in primate MLBNs, and the components of its oblique saccades are stretched. Two mechanisms contribute to this stretching: the distribution of MLBN tuning curves and the inhibition exerted by the contralateral population of MLBNs. In contrast, feedback control of the saccade contributes negligibly to the stretching. Even though the vectorial burster model predicts a component stretching, it is not constrained to produce perfectly straight oblique saccades because no trajectory control is implemented. The amount of curvature depends on the similarity of the horizontal and vertical systems (both neural and mechanical). In this model, stretching is interpreted simply as a side effect of the properties of the MLBNs' tuning curves. The distributed MLBNs of the vectorial burster model forces the general organization of the saccadic system to be reconsidered. We propose that a distributed architecture in which several different neural systems cooperate is needed.

Computer Simulation↗

PubMedDate not supplied

Cost-effectiveness models of clinical trials of new pharmaceuticals for AIDS-related Pneumocystis carinii pneumonia: are they helpful to policy makers?

BACKGROUND: Pneumocystis carinii pneumonia (PCP) has been the most common and most costly complication of acquired immunodeficiency syndrome (AIDS). Because of concern over the high costs of care for persons with AIDS, policy makers have instituted a number of new measures to encourage cost-effectiveness. Although the clinical efficacy of new pharmaceutical agents is evaluated extensively before approval, comparative trials of a new agent with its major competitor are sometimes not carried out, and estimates of cost-effectiveness are therefore difficult to obtain. METHODS: We describe methodologic issues associated with the development of economic models of new pharmaceutical agents and illustrate these issues with an analysis of second-line therapeutic options for PCP. RESULTS: A new drug, trimetrexate, and the standard second-line therapy, pentamidine, are both inferior to the standard first-line therapy, trimethoprim/sulfamethoxazole, as initial therapy for moderate to severe PCP. However, as many as half of the patients with PCP are either intolerant or refractory to trimethoprim/sulfamethoxazole therapy for PCP. Economic models suggest that, if the two drugs have identical survival rates, then trimetrexate, despite a higher acquisition cost, is both less expensive and more effective in achieving toxicity-free survival than pentamidine in patients who require second-line therapy for PCP. Sensitivity analyses indicate that trimetrexate is less expensive than pentamidine over a wide range of estimates of costs and effectiveness. CONCLUSIONS: To make well-informed therapeutic decisions, policy makers and physicians require head-to-head studies of a new pharmaceutical agent with its major competitor. However, economic models can be used to derive estimates of cost-effectiveness of new pharmaceutical agents when such data are lacking. The interpretation of these models raises general issues related to the perspective of the investigators, study design, estimation of costs of care, rates of response, toxicity, and survival, and generalizability of the results to other settings as well as methodologic issues that are unique to human immunodeficiency virus (HIV) disease. If a comparative trial found better survival rates, then cost-effectiveness models would be of limited usefulness; almost all physicians would use the drug with the higher survival rate.

AIDS-Related Opportunistic Infections↗

PubMed2005

Feature selection in quantitative structure-activity relationships.

A key component of building quantitative structure-activity relationship (QSAR) models is the selection of an appropriate set of molecular features or descriptors. Feature selection can affect the accuracy, stability and interpretability of a model. There are thousands of molecular descriptors currently available, and the selection of an appropriate descriptor set for a particular model can be a daunting task. While there are no absolute rules for selecting appropriate sets of descriptors, a number of recent publications describe automated methods for identifying optimal feature sets. This review provides an overview of a number of the methods described in these publications.

Animals↗

PubMed1999

Cost-effectiveness model of a phase II clinical trial of a new pharmaceutical for essential thrombocythemia: is it helpful to policy makers?

Essential thrombocythemia (ET) is a rare, chronic myeloproliferative disorder of unknown origin characterized by thrombocytosis, excessive megakaryocytes, hemorrhages, and thrombotic complications. Because of the high costs of care for persons with rare diseases, policymakers are concerned with both clinical effectiveness and cost-effectiveness of new treatments. Although the clinical efficacy of all new pharmaceutical agents for rare diseases is evaluated extensively in clinical trial settings before approval by the Food and Drug Administration (FDA), comparative phase III trials of a new agent with its major competitor are sometimes not possible to carry out, and estimates of cost-effectiveness are therefore difficult to obtain. We describe methodologic issues associated with the development of economic models of new pharmaceutical agents for rare diseases and illustrate these issues with an analysis of a new therapy for ET. Anagrelide is a newly approved platelet aggregation inhibitor that can be used as primary therapy for ET. The agent reduces platelet counts by 50% in more than 70% of ET patients. Economic models suggest that, over the first year of anagrelide therapy, monthly costs for therapy and complications decreased from $775 to $490, the effectiveness improved to 98%, and the cost-effectiveness improved to $1,505 per major complication (gastrointestinal bleed, transient ischemic attack or stroke, or preinfarction angina or myocardial infarction) prevented. Sensitivity analyses indicate that, after the first 3 months of treatment, total costs of anagrelide treatment were in the range of $1,505 to $1,615 per major complication prevented. To make well-informed therapeutic decisions, policymakers and physicians require head-to-head studies of a new pharmaceutical agent with its major competitor. However, economic models can be used to derive estimates of cost-effectiveness of new pharmaceutical agents when such data are lacking. The interpretation of these models raises general issues related to the perspective of the investigator, study design, estimation of costs of care, rates of response, toxicity, survival, and the ability to generalize the results to other settings, as well as methodologic issues that are unique to rare diseases. If a comparative study found better therapeutic outcomes, then cost-effectiveness models would be of limited usefulness. Almost all physicians would use the drug with the better therapeutic profile.

Clinical Trials, Phase II as Topic↗

PubMed2001

A model-based analysis of professional practices in suicide prevention.

AIMS: The purpose of this study was to investigate what suicide prevention means in terms of practical interventions. Another purpose was to assess the feasibility of a previously developed theoretical model for analysing suicide prevention (1-3). METHODS: The data consisted of plans for action provided by professionals from five fields (n = 173) (psychologists [n = 41], clergy [n=37], nurses [n=34], social workers [n=32], and physicians [n=29]) as responses to an inquiry within the National Suicide Prevention Project in Finland. The plans were operationally described and theoretically interpreted using the model. RESULTS: The analysis indicated that practice patterns were more or less similar irrespective of the focus of suicide prevention. Neither did clear sectoral or professional profiles of practice appear. Clinical topics, individually focused interventions and curative strategies constituted the main approach. Interventions focused mainly on risk factors, the priorities being life crises, acute risk of suicide, coping of survivors, and a suicide attempt. The bulk of the strategies were aimed at developing professional interventions and skills. CONCLUSIONS: The model proved to be a feasible tool for differentiating and theoretically interpreting suicide-prevention approaches. The paradigm adopted by the sectors was versatile and comprehensive: the activities were multifocused, all aims of prevention and locations of intervention were included, and a common set of interventions--a "multimethod approach"--was applied. The adoption of the idea of risk as well as of protective factors can be interpreted as reflecting a process theory of suicide development.

Attitude of Health Personnel↗

PubMed2006

FITSK: online local learning with generic fuzzy input Takagi-Sugeno-Kang fuzzy framework for nonlinear system estimation.

Existing Takagi-Sugeno-Kang (TSK) fuzzy models proposed in the literature attempt to optimize the global learning accuracy as well as to maintain the interpretability of the local models. Most of the proposed methods suffer from the use of offline learning algorithms to globally optimize this multi-criteria problem. Despite the ability to reach an optimal solution in terms of accuracy and interpretability, these offline methods are not suitably applicable to learning in adaptive or incremental systems. Furthermore, most of the learning methods in TSK-model are susceptible to the limitation of the curse-of-dimensionality. This paper attempts to study the criteria in the design of TSK-models. They are: 1) the interpretability of the local model; 2) the global accuracy; and 3) the system dimensionality issues. A generic framework is proposed to handle the different scenarios in this design problem. The framework is termed the generic fuzzy input Takagi-Sugeno-Kang fuzzy framework (FITSK). The FITSK framework is extensible to both the zero-order and the first-order FITSK models. A zero-order FITSK model is suitable for the learning of adaptive system, and the bias-variance of the system can be easily controlled through the degree of localization. On the other hand, a first-order FITSK model is able to achieve higher learning accuracy for nonlinear system estimation. A localized version of recursive least-squares algorithm is proposed for the parameter tuning of the first-order FITSK model. The local recursive least-squares is able to achieve a balance between interpretability and learning accuracy of a system, and possesses greater immunity to the curse-of-dimensionality. The learning algorithms for the FITSK models are online, and are readily applicable to adaptive system with fast convergence speed. Finally, a proposed guideline is discussed to handle the model selection of different FITSK models to tackle the multi-criteria design problem of applying the TSK-model. Extensive simulations were conducted using the proposed FITSK models and their learning algorithms; their performances are encouraging when benchmarked against other popular fuzzy systems.

Algorithms↗

PubMed2000

The use of long-term care services by the Dutch elderly.

The main focus of this paper is the development of an appropriate framework to characterize the process of long-term care utilization by the Dutch elderly. Three broad categories of care services are considered, namely, informal care, formal care at home, and institutional care. The use of these care alternatives is modelled jointly, and stochastic dependence is allowed between the various care options. Special attention is given to the concept of health status and to the potential endogeneity of this variable in the model. We apply a flexible non-parametric method to summarize the multidimensional concept of health status into a limited set of interpretable indices. The model is applied on the Longitudinal Ageing Study Amsterdam (LASA). We find strong effects of health status, gender, socio-economic variables, and prices on the utilization of long-term care services.

Aged↗

PubMed1979

Mutational mosaicism and genetic counseling in retinoblastoma.

Mutations may arise throughout an organism's life cycle. Typically, sporadic meiotic mutations give rise to individuals with all their germinal and somatic cells bearing the mutant gene. These mutations may be amorphs (with full penetrance and expressivity) or hypomorphs (with reduced penetrance and expressivity). Mutational mosaicism, however, involves the origin of mutations occurring during mitosis, whether in the parent at some stage prior to reproductive maturity or in the offspring at some time following fertilization. The phenotypic expression and transmission of these new mutations are dependent on the proportion of cells bearing the mutant gene as well as the location of these cells in somatic and/or germinal tissues. Mutational mosaicism was used as a developmental model to analyze 1,500 sporadic and 179 familial cases of retinoblastoma from the world literature. This model provided an interpretation for the origin, onset, and transmissibility of the sporadic unilateral retinoblastoma cases, which represent over 60% of all retinoblastoma patients. The model also permits a reclassification of all transmissible types of retinoblastoma; based on this classification, more accurate risk figures for genetic counseling can be offered. In addition, mutational mosaicism can be extended as a model to other autosomal dominant and X-linked mutations.

Animals↗

PubMed1999

Modelling asthma epidemics on the relationship between air pollution and asthma emergency visits in Barcelona, Spain.

We have investigated different methods of controlling for asthma epidemics in the time series regression of the relationship between air pollution and asthma emergency visits in Barcelona, Spain. The relationship between air pollution and asthma emergency room visits was modelled using autoregressive Poisson models. We examined the effect of using no control by epidemics, and modelling asthma epidemics with a single dummy variable, six dummy variables, and a dummy variable for each epidemic day. Air pollution coefficients increased when controlling asthma epidemics with six dummy variables instead of a single variable. They further increased when autocorrelation was allowed for. Standard errors were relatively unaffected when either the epidemics or the autocorrelation were included in the model. Black smoke, nitrogen dioxide and ozone were statistically significant associated to asthma emergency visits after using six dummy variables to control for asthma epidemics. We have shown that different models, including different confounding variables, give markedly different estimates of the effect of a pollutant on health. Care is needed in the interpretation of such models, and careful reporting so it is clear how the confounding variables have been modelled.

Air Pollutants↗

PubMed2006

A better lemon squeezer? Maximum-likelihood regression with beta-distributed dependent variables.

Uncorrectable skew and heteroscedasticity are among the "lemons" of psychological data, yet many important variables naturally exhibit these properties. For scales with a lower and upper bound, a suitable candidate for models is the beta distribution, which is very flexible and models skew quite well. The authors present maximum-likelihood regression models assuming that the dependent variable is conditionally beta distributed rather than Gaussian. The approach models both means (location) and variances (dispersion) with their own distinct sets of predictors (continuous and/or categorical), thereby modeling heteroscedasticity. The location sub-model link function is the logit and thereby analogous to logistic regression, whereas the dispersion sub-model is log linear. Real examples show that these models handle the independent observations case readily. The article discusses comparisons between beta regression and alternative techniques, model selection and interpretation, practical estimation, and software.

Analysis of Variance↗

PubMedDate not supplied

A psychological model of mental disorder.

A coherent conceptualization of the role of psychological factors is of great importance in understanding mental disorder. Academic articles and professional reports alluding to psychological models of the etiology of mental disorder are becoming increasingly common, and there is evidence of a marked policy shift toward the provision of psychological therapies and interventions. This article discusses the relationship between biological, social, and psychological factors in the causation and treatment of mental disorder. It argues that simple biological reductionism is not scientifically justified, and also that the specific role of psychological processes within the biopsychosocial model requires further elaboration. The biopsychosocial model is usually interpreted as implying that biological, psychological, and social factors are co-equal partners in the etiology of mental disorder. The psychological model of mental disorder presented here suggests that disruption or dysfunction in psychological processes is a final common pathway in the development of mental disorder. These processes include, but are not limited to, cognitive processes. The model proposes that biological and social factors, together with a person's individual experiences, lead to mental disorder through their conjoint effects on those psychological processes. Implications for research, interventions, and policy are discussed.

Brain↗

PubMed2004

Regulatory motif finding by logic regression.

MOTIVATION: Multiple transcription factors coordinately control transcriptional regulation of genes in eukaryotes. Although many computational methods consider the identification of individual transcription factor binding sites (TFBSs), very few focus on the interactions between these sites. We consider finding TFBSs and their context specific interactions using microarray gene expression data. We devise a hybrid approach called LogicMotif composed of a TFBS identification method combined with the new regression methodology logic regression. LogicMotif has two steps: First, potential binding sites are identified from transcription control regions of genes of interest. Various available methods can be used in this step when the genes of interest can be divided into groups such as up-and downregulated. For this step, we also develop a simple univariate regression and extension method MFURE to extract candidate TFBSs from a large number of genes in the availability of microarray gene expression data. MFURE provides an alternative method for this step when partitioning of the genes into disjoint groups is not preferred. This first step aims to identify individual sites within gene groups of interest or sites that are correlated with the gene expression outcome. In the second step, logic regression is used to build a predictive model of outcome of interest (either gene expression or up- and down-regulation) using these potential sites. This 2-fold approach creates a rich diverse set of potential binding sites in the first step and builds regression or classification models in the second step using logic regression that is particularly good at identifying complex interactions. RESULTS: LogicMotif is applied to two publicly available datasets. A genome-wide gene expression data set of Saccharomyces cerevisiae is used for validation. The regression models obtained are interpretable and the biological implications are in agreement with the known resuts. This analysis suggests that LogicMotif provides biologically more reasonable regression models than previous analysis of this dataset with standard linear regression methods. Another dataset of S.cerevisiae illustrates the use of LogicMotif in classification questions by building a model that discriminates between up- and down-regulated genes in iron copper deficiency. LogicMotif identifies an inductive and two repressor motifs in this dataset. The inductive motif matches the binding site of the transcription factor Aft1p that has a key role in regulation of the uptake process. One of the novel repressor sites is highly present in transcription control regions of FeS genes. This site could represent a TFBS for an unknown transcription factor involved in repression of genes encoding FeS proteins in iron deficiency. We establish the robustness of the method to the type of outcome variable used by considering both continuous and binary outcome variables for this dataset. Our results indicate that logic regression used in combination with cluster/group operating binding site identification methods or with our proposed method MFURE is a powerful and flexible alternative to linear regression based motif finding methods. AVAILABILITY: Source code for logic regression is freely available as a package of the R programming language by Ruczinski et al. (2003) and can be downloaded at http://bear.fhcrc.org/~ingor/logic/download/download.html an R package for MFURE is available at http://www.stat.berkeley.edu/~sunduz/software.html

Algorithms↗

PubMed2004

Coupling of functional gene diversity and geochemical data from environmental samples.

Genomic techniques commonly used for assessing distributions of microorganisms in the environment often produce small sample sizes. We investigated artificial neural networks for analyzing the distributions of nitrite reductase genes (nirS and nirK) and two sets of dissimilatory sulfite reductase genes (dsrAB1 and dsrAB2) in small sample sets. Data reduction (to reduce the number of input parameters), cross-validation (to measure the generalization error), weight decay (to adjust model parameters to reduce generalization error), and importance analysis (to determine which variables had the most influence) were useful in developing and interpreting neural network models that could be used to infer relationships between geochemistry and gene distributions. A robust relationship was observed between geochemistry and the frequencies of genes that were not closely related to known dissimilatory sulfite reductase genes (dsrAB2). Uranium and sulfate appeared to be the most related to distribution of two groups of these unusual dsrAB-related genes. For the other three groups, the distributions appeared to be related to pH, nickel, nonpurgeable organic carbon, and total organic carbon. The models relating the geochemical parameters to the distributions of the nirS, nirK, and dsrAB1 genes did not generalize as well as the models for dsrAB2. The data also illustrate the danger (generating a model that has a high generalization error) of not using a validation approach in evaluating the meaningfulness of the fit of linear or nonlinear models to such small sample sizes.

Bacteria↗

PubMed1998

Pulmonary disposition of lipophilic amine compounds in the isolated perfused rabbit lung.

We measured the pulmonary venous concentration vs. time curves for [3H]alfentanil, [14C]lidocaine, and [3H]codeine after the bolus injection of each of these lipophilic amine compounds (LAC) and a vascular-reference indicator (fluorescein isothiocyanate-dextran) into the pulmonary artery of isolated perfused rabbit lungs. A range of flows and perfusate albumin concentrations was studied. To evaluate the information content of the data, we developed a kinetic model describing the pulmonary disposition of these LAC that was based on indicator dilution theory, and we sought a robust approach for interpreting the estimated model parameters. We found that the distribution of the kinetic model rate constants of the lipophilic amine-tissue interactions can be described by alpha, H, and psi, where alpha is a measure of the capacity of the rapidly equilibrating interactions between the lipophilic amine and the tissue; H is a measure of the equilibrium capacity of the slowly equilibrating interactions between the lipophilic amine and the tissue; and psi is the mean sojourn time. The values of alpha, H, and psi were 0.8 +/- 0.1 (SE), 0.6 +/- 0.1, and 1.6 +/- 0.5 s; 1.9 +/- 0.1, 5.3 +/- 0.4, and 5.6 +/- 0.5 s; and 1.1 +/- 0.1, 9.8 +/- 0.4, and 4.7 +/- 0.2 s for alfentanil, lidocaine, and codeine, respectively. These values for alpha, H, and psi reveal the relative dominance of the slowly equilibrating interactions for lidocaine and codeine in comparison with alfentanil. This approach to data analysis may have utility for the potential use of LAC to reveal and to quantify changes in lung tissue composition associated with lung disease.

Alfentanil↗