Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “model selection”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

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

At least 595 records · Page 33Linked to original sources
PubMed2000

Gradient-based optimization of hyperparameters.

Many machine learning algorithms can be formulated as the minimization of a training criterion that involves a hyperparameter. This hyperparameter is usually chosen by trial and error with a model selection criterion. In this article we present a methodology to optimize several hyperparameters, based on the computation of the gradient of a model selection criterion with respect to the hyperparameters. In the case of a quadratic training criterion, the gradient of the selection criterion with respect to the hyperparameters is efficiently computed by backpropagating through a Cholesky decomposition. In the more general case, we show that the implicit function theorem can be used to derive a formula for the hyperparameter gradient involving second derivatives of the training criterion.

Algorithms↗
PubMed1988

Differentiation of alpha-adrenergic receptors using pharmacological evaluation and molecular modeling of selective adrenergic agents.

Subtypes of alpha adrenergic receptors were studied using selective adrenergic agonists. A-53693, A-54741, and related compounds were evaluated for their affinity for alpha receptor subtypes using radioligand binding techniques. Efficacy and potency were also evaluated using in vitro bioassays of alpha-1 receptors in rabbit aorta smooth muscle and alpha-2 receptors in the phenoxybenzamine-pretreated canine saphenous vein. Active and inactive compounds were then submitted for computer-assisted molecular modeling evaluation to ascertain the structural requirements for optimal potency and selectivity. Rigid catecholamines such as A-53693 display a high degree of selectivity for alpha-2 compared to alpha-1 receptors, probably because of the unique regions of space at the ligand binding site occupied by active compounds. Imidazolines such as A-54741 also interact with extremely high affinity and potency for alpha-2 receptors, and to a lesser extent at alpha-1 receptors. The spatial domains occupied by phenethylamines and imidazolines differ, each having unique regions of permissable space at alpha receptors. Compounds such as A-53693 and A-54741 are extremely useful probes of the molecular interactions of alpha agonistic compounds which will help in the design of even more selective drugs for alpha adrenergic receptors.

Adrenergic alpha-Agonists↗
PubMed1996

Synthesis, biological activity, and molecular modeling of selective 5-HT(2C/2B) receptor antagonists.

The synthesis and biological activity are reported for a series of analogues of the previously published indole urea 2 (SB-206553), designed to probe the 5-HT(2C) receptor binding site. Small molecule modeling studies have been used to define a region in space which is allowed at the 5-HT(2C) receptor but disallowed at the 5-HT(2A) receptor. In a complementary approach, docking of 2 into our model of the 5-HT(2C) receptor has allowed us to propose a novel primary binding interaction for this series of diaryl ureas, involving a potential double hydrogen-bonding interaction between the urea carbonyl oxygen of the ligand and two serine residues in the receptor. The difference of two valine residues in the 5-HT(2C) receptor for leucine residues in the 5-HT(2A) receptor is believed to account for the observed 5-HT(2C)/5-HT(2A) selectivity with 2.

Animals↗
PubMed2006

Computational approaches to model ligand selectivity in drug design.

To be effective, a designed drug must discriminate successfully the macromolecular target from alternative structures present in the organism. The last few years have witnessed the emergence of different computational tools aimed to the understanding and modeling of this process at molecular level. Although still rudimentary, these methods are shaping a coherent approach to help in the design of molecules with high affinity and specificity, both in lead discovery and in lead optimization. It is the purpose of this review to illustrate the array of computational tools available to consider selectivity in the design process, to summarize the most relevant applications, and to sketch the challenges ahead.

Amino Acid Sequence↗
PubMed2026

On the origin of animals and placental mammals: a critique of literalist readings of the fossil record.

The fossil record is incomplete, as evidenced by the pervasive presence of ghost lineages throughout the Tree of Life. For example, across placental mammals, at least 720 Myr of basal lineages are ghost lineages, that is, lineages that have left no fossil evidence of their past history. In contrast, some studies have suggested that the fossil record is a faithful temporal archive of evolutionary history and thus the times of diversification of clades must be close to the ages of their oldest fossils. Such literalist interpretations have been contradicted by analysis of molecular datasets which, in many cases, indicate that groups including placental mammals and animals may have originated at times substantially older than their fossil records. Some of those studies have further argued that, in the case of animals and placental mammals, molecular clocks are uninformative, suffer from characteristic pathologies, and thus cannot distinguish between recent and ancient hypotheses of diversification. Here, we reexamine these two cases and show, using Bayesian model selection theory, that the explosive diversification models previously proposed for animals and placental mammals have a posterior probability of ∼0. We show the characteristic pathologies purportedly discovered do not exist, highlight errors in previous analyses, and provide advice on best practice for molecular-clock dating analysis.

Animals↗
PubMed2006

Ramsay-curve item response theory (RC-IRT) to detect and correct for nonnormal latent variables.

Popular methods for fitting unidimensional item response theory (IRT) models to data assume that the latent variable is normally distributed in the population of respondents, but this can be unreasonable for some variables. Ramsay-curve IRT (RC-IRT) was developed to detect and correct for this nonnormality. The primary aims of this article are to introduce RC-IRT less technically than it has been described elsewhere; to evaluate RC-IRT for ordinal data via simulation, including new approaches for model selection; and to illustrate RC-IRT with empirical examples. The empirical examples demonstrate the utility of RC-IRT for real data, and the simulation study indicates that when the latent distribution is skewed, RC-IRT results can be more accurate than those based on the normal model. Along with a plot of candidate curves, the Hannan-Quinn criterion is recommended for model selection.

Area Under Curve↗
PubMedDate not supplied

Use of a SCID mouse model to select for a more aggressive strain of prostate cancer.

BACKGROUND: Prostate cancer is the most common non-cutaneous malignancy to affect men and has a propensity for metastasizing to bone. To better mimic the biology of metastatic prostate cancer, we have developed a model that utilizes both human prostate cancer and human bone in the SCID mouse. MATERIALS AND METHODS: Injection of a xenograft of human prostate cancer, LAPC-4, near a human bone core that was previously implanted within the hindlimb of a SCID mouse allowed for the selection of a more aggressive subset of cells, known as LAPC-4(2) (read as "LAPC-4 squared"). RESULTS: As compared to LAPC-4 cells, these "bone-selected" LAPC-4(2) cells form tumors more rapidly, develop PSA-positive serum at an earlier time-point as well as with higher levels, develop androgen independence, and metastasize to human bone after orthotopic injection. CONCLUSION: The selection of a more aggressive subset of prostate cancer cells that have developed androgen independence and the propensity to metastasize is of paramount importance as these are the cellular characteristics that are clinically linked to morbidity. Analysis of these "bone-selected" cells may lead to a better understanding of the molecular basis behind the conversion of low-grade prostate carcinoma to its more deadly, metastatic form.

Adult↗
PubMed1999

A computational model of selective deficits in first and second-order motion processing.

Recent neurological studies of selective impairments in first and second-order motion processing are of considerable relevance in elucidating the mechanisms of motion perception in normal human observers. We examine the stimuli which have been used to assess first and second-order motion processing capabilities in clinical subjects, and discuss the nature of the computations necessary to extract their motion. We find that a simple computational model of first and second-order motion processing is able to account for the data. The model consists of a first-order channel computing motion at coarse and fine scales, and a coarse scale second-order channel. The second-order channel is sensitive to motion information defined by variations in luminance, contrast, spatial frequency and flicker. When elements of the model are disabled, its performance on either first or second-order motion can be selectively impaired in line with the neurological data.

Brain Diseases↗
PubMed1981

Social selection in human populations. I. Modification of the fitness of offspring by an affected parent.

The concept of social selection for deleterious genes has been introduced by considering two alleles at one locus. A social selection model is constructed by assuming that the fitness of an individual is determined by his or her own as well as the parental phenotypes. It is shown that the equilibrium gene frequency depends on the loss of fitness of an individual due to the trait (gamma), due to affected parents (beta), and the probability that the heterozygote develops the trait (h). With mutational changes from the wild-type allele to the deleterious gene at a rate of alpha per generation, the equilibrium frequency of deleterious genes is approximately alpha/hs for 0 less than h less than or equal to 1 and square root alpha/s for h = 0, where s = gamma + beta(1 -- gamma)/2. Implications of the social selection model have been discussed for several diseases in man.

Alleles↗
PubMed1991

An optimization strategy for a biokinetic model of inhaled radionuclides.

Models for material disposition and dosimetry involve predictions of the biokinetics of the material among compartments representing organs and tissues in the body. Because of a lack of human data for most toxicants, many of the basic data are derived by modeling the results obtained from studies using laboratory animals. Such a biomathematical model is usually developed by adjusting the model parameters to make the model predictions match the measured retention and excretion data visually. The fitting process can be very time-consuming for a complicated model, and visual model selections may be subjective and easily biased by the scale or the data used. Due to the development of computerized optimization methods, manual fitting could benefit from an automated process. However, for a complicated model, an automated process without an optimization strategy will not be efficient, and may not produce fruitful results. In this paper, procedures for, and implementation of, an optimization strategy for a complicated mathematical model is demonstrated by optimizing a biokinetic model for 144Ce in fused aluminosilicate particles inhaled by beagle dogs. The optimized results using SimuSolv were compared to manual fitting results obtained previously using the model simulation software GASP. Also, statistical criteria provided by SimuSolv, such as likelihood function values, were used to help or verify visual model selections.

Administration, Inhalation↗
PubMedDate not supplied

[A mathematical model of selection, including the energy metabolism of organisms, as an instrument for researching the evolutionary process].

A system simulating the selection conditions for studying the consequences arising from the differences in metabolism rates of the individuals (E-system) has been described. The E-system is a constructive expression of indirect individual elimination (Shmal'gauzen, 1939). The notion of energy exchange and general nonselective elimination as conjugated evolutionary factors has been proposed. The partial elimination has been connected with the known ecological generalization, the temporal decrease of negative interactions. The study demonstrated the ability to use the E-system as a heuristic tool for studying the theory of evolution.

Animals↗
PubMed2006

A novel model of organic waste composting in Taiwan military community.

The relatively large quantities of waste generated daily in military barracks pose a significant impact on environments in this small island of Taiwan due to limited land and crowded populations. In order to find a suitable handling method to the barracks' characteristics in Taiwan, a comprehensive selection model for various barracks carrying out waste composting was established through personal interviews and questionnaire data analyses on the military barracks. From this, the basic data for this research was also built, to evaluate the experiences and management practices in these facilities. This study found that environmental responsibility and expertise are the primary concerns and considerations among all of the military units when doing waste composting. Support of the commander, organizational image and human resource/facilities are also concerns. To offer the military barracks a selection model and policy guidelines for the waste composting treatment, an itemized score evaluation standard and a selecting chart of composting recycling methods were designed by combining the practical experimental results with economic considerations, waste classification, and Delphi expert technique.

Conservation of Natural Resources↗
PubMed2004

On a hybrid method in dose finding studies.

OBJECTIVES: Combination of multiple testing and modeling techniques in dose-response studies. Use of hypotheses tests to assess the significance of the dose-response signal associated with a given candidate dose-response model. Estimation of target dose(s) following the previous model selection step. Illustration of the method with a real data example. METHODS: We assume a set of candidate models potentially reflecting the data generating process. The appropriateness of each individual model is evaluated in terms of contrast tests, where each set of contrast weights describes a specific dose-response shape. Optimum contrast weights are computed, which maximize the non-centrality parameters associated with the contrast tests. A reference set of appropriate candidate models is obtained while controlling the familywise error rate. A single model is then selected from this reference set using standard model selection criteria. The final step is devoted to dose finding by applying inverse regression techniques. This is illustrated for estimating the minimum effective dose. RESULTS: The method is as powerful as competing standard dose-response tests to detect an overall dose-related trend. In addition, the possibility is given to estimate one or more target doses of interest. The analysis of a real data example confirms the advantages of the proposed hybrid method. CONCLUSIONS: Combining multiple testing and modeling techniques leads to a powerful tool, which uses the advantages of both approaches: Rigid error control at the significance testing step and flexibility at the dose estimation step. The method can be extended to handle more general linear models including covariates and factorial treatment structures.

Clinical Trials as Topic↗
PubMed2001

Variable selection and Bayesian model averaging in case-control studies.

Covariate and confounder selection in case-control studies is often carried out using a statistical variable selection method, such as a two-step method or a stepwise method in logistic regression. Inference is then carried out conditionally on the selected model, but this ignores the model uncertainty implicit in the variable selection process, and so may underestimate uncertainty about relative risks. We report on a simulation study designed to be similar to actual case-control studies. This shows that p-values computed after variable selection can greatly overstate the strength of conclusions. For example, for our simulated case-control studies with 1000 subjects, of variables declared to be 'significant' with p-values between 0.01 and 0.05, only 49 per cent actually were risk factors when stepwise variable selection was used. We propose Bayesian model averaging as a formal way of taking account of model uncertainty in case-control studies. This yields an easily interpreted summary, the posterior probability that a variable is a risk factor, and our simulation study indicates this to be reasonably well calibrated in the situations simulated. The methods are applied and compared in the context of a case-control study of cervical cancer.

Analysis of Variance↗
PubMed1995

Elucidation of autoimmune disease mechanism based on testicular and ovarian autoimmune disease models.

This paper describes several selected models of autoimmune disease of the gonads. Based on these findings, I have reviewed current knowledge concerning the tolerance mechanisms that normally prevent gonadal autoimmunity, the potential events that can overcome such mechanism to trigger autoimmune diseases. In addition we also summarize the immunopathology of orchitis and our understanding of the mechanisms responsible for the immunopathology of the disease. Recent studies indicate that pathogenic T cells capable of eliciting autoimmune diseases in these organs develop in both the neonatal and adult thymuses and they persist in the normal peripheral immune system. However, the function of the pathogenic T cells in adult mice is normally under the control of regulatory T cells which maintain peripheral tolerance, and important phenotypic differences are being defined between these two functional CD4+ T cell subsets. When the clonal balance of these T cell subsets is tipped in favor of pathogenic T cells, autoimmune diseases of the gonads could ensue. Pathogenic T cells responsible for autoimmune oophoritis can be activated through stimulation by non-ovarian peptides that cross-react with self ovarian peptides at the level of the T cell receptor. This novel form of antigen mimicry depends in part on the sharing, between unrelated peptides, the few critical amino acids required for activation of pathogenic T cells. Antibodies can bind to the ovarian target antigens during the development of autoimmune orchitis and autoimmune oophoritis. However, the precise role of antibody in these autoimmune diseases has not been critically explored. In this study, I have described a novel mechanism of autoantibody induction. Immunization of female mice with a pure T cell peptide from ZP3 can lead to the production of antibodies against ZP3 domains outside the immunogenic ZP3 peptide. Evidently, endogenous antigens from normal and pathologic ovaries may reach peripheral immune tissues, and provide the antigenic stimulus to trigger an autoantibody response. This occurs at the same time when activation of ZP3 specific T cells is detected, and it is not simply a consequence of tissue injury. Importantly, the autoantibodies react with native antigenic determinants, and are potentially important in autoimmune disease pathogenesis.

Animals↗
PubMed1999

Residual vision in the blind field of hemidecorticated humans predicted by a diffusion scatter model and selective spectral absorption of the human eye.

The notion of blindsight was recently challenged by evidence that patients with occipital damage and contralateral field defects show residual islands of vision which may be associated with spared neural tissue. However, this possibility could not explain why patients who underwent the resection or disconnection of an entire cerebral hemisphere exhibit some forms of blindsight. We present here a model for the detection of intraocular scatter, which can account for human sensitivity values obtained in the blind field of hemidecorticated patients. The model demonstrates that, under controlled experimental conditions i.e. where the extraocular scatter is eliminated, Lambertian intraocular scatter alone can account for the visual sensitivities reported in these patients. The model also shows that it is possible to obtain a sensitivity in the blind field almost equivalent to that in the good field using the appropriate parameters. Finally, we show with in-vivo spectroreflectometry measurements made in the eyes of our hemidecorticated patients, that the relative drop in middle wavelength sensitivity generally obtained in the blind field of these patients can be explained by selective intraocular spectral absorption.

Blindness, Cortical↗