Comparison of different methods for variable selection

In chemistry for chemical analysis of a multi-component sample or quantitative structure-activity/property relationship (QSAR/QSPR) studies, variable selection is a key step. In this study, comparisons between different methods were performed. These methods include three classical methods such as forward selection, backward elimination and stepwise regression; orthogonal descriptors; leaps-and-bounds regression and genetic algorithm. Thirty-five nitrobenzenes were taken as the data set. From these structures quantum chemical parameters, topological indices and indicator variable were extracted as the descriptors for the comparisons of variable selections. The interesting results have been obtained. (C) 2001 Elsevier Science B.V. All rights reserved.

Selection for variables in quantitative structure-activity/property relationship studies - Comparison between methods of orthogonal descriptors and leaps-and-bounds regression analysis

In this paper, the comparison of orthogonal descriptors and Leaps-and-Bounds regression analysis is performed. The results obtained by using orthogonal descriptors are better than that obtained by using Leaps-and-Bounds regression for the data set of nitrobenzenes used in this study. Leaps-and-Bounds regression can be used effectively for selection of variables in quantitative structure-activity/property relationship(QSAR/QSPR) studies. Consequently, orthogonalisation of descriptors is also a good method for variable selection for studies on QSAR/QSPR.

Variable selection by orthogonal descriptors and prediction of R-f value of phenol and aniline derivatives

Orthogonal descriptors is a viable method for variable selection, but this method strongly depend on the orthogonalisation ordering of the descriptors. In this paper, we compared the different methods used for order the descriptors. It showed that better results could be achieved with the use of backward elimination ordering. We predicted R-f value of phenol and aniline derivatives by this method, and compared it with classical algorithms such as forward selection, backward elimination, and stepwise procedure. Some interesting hints were obtained.

Evidence for positive selection in phycoerythrin genes of red algae and cyanobacteria Prochlorococcus and Synechococcus

Because of the shortage of phycoerythrin (PE) gene sequences from rhodophytes, peBA encoding beta- and alpha-subunits of PE from three species of red algae (Ceramium boydenn, Halymenia sinensis, and Plocamium telfariae) were cloned and sequenced. Different selection forces have affected the evolution of PE lineages. 8.9 % of the codons were subject to positive selection within the PE lineages (excluding high-irradiance adapted Prochlorococcus). More than 40 % of the sites may be under positive selection, and nearly 20 % sites are weakly constraint sites in high-irradiance adapted Prochlorococcus. Sites most likely undergoing positive selection were found in the chromophore binding domains, suggesting that these sites have played important roles in environmental adaptation during PE diversification. Moreover, the heterogeneous distribution of positively selected sites along the PE gene was revealed from the comparison of low-irradiance adapted Prochlorococcus and marine Synechococcus, which firmly suggests that evolutionary patterns of PEs in these two lineages are significantly different.

A genetic linkage map of Pacific white shrimp (Litopenaeus vannamei): sex-linked microsatellite markers and high recombination rates

Pacific white shrimp (Litopenaeus vannamei) is the leading species farmed in the Western Hemisphere and an economically important aquaculture species in China. In this project, a genetic linkage map was constructed using amplified fragment length polymorphism (AFLP) and microsatellite markers. One hundred and eight select AFLP primer combinations and 30 polymorphic microsatellite markers produced 2071 markers that were polymorphic in either of the parents and segregated in the progeny. Of these segregating markers, 319 were mapped to 45 linkage groups of the female framework map, covering a total of 4134.4 cM; and 267 markers were assigned to 45 linkage groups of the male map, covering a total of 3220.9 cM. High recombination rates were found in both parental maps. A sex-linked microsatellite marker was mapped on the female map with 6.6 cM to sex and a LOD of 17.8, two other microsatellite markers were also linked with both 8.6 cM to sex and LOD score of 14.3 and 16.4. The genetic maps presented here will serve as a basis for the construction of a high-resolution genetic map, quantitative trait loci (QTLs) detection, marker-assisted selection (MAS) and comparative genome mapping.

Heat-shock protein70 gene expression in four hatchery Pacific Abalone Haliotis discus hannai Ino populations using for marker-assisted selection

Heat shock proteins (Hsps) are molecular chaperones that help organisms cope with stressful conditions. Here, we report on the growth rates and Hsp70 expressions in inbred and hybrid populations of abalone Haliotis discus hannai Ino. In abalone, inbred populations expressed more Hsp70 than hybrid populations at all temperatures, except at very high temperatures close to the physiological limit. At benign temperatures, there was a clear trend towards higher Hsp70 expression in inbred than hybrid populations, whereas at higher temperatures, a trend in the opposite direction was observed. The temperature of maximal Hsp70 expression (T-peak) varied with the population type. The T-peak of inbred populations (26 degrees C) was lower than that of the hybrid populations (28 degrees C). The maximal inducible Hsp70 of inbred populations was higher than that of hybrid populations. The results showed a trend towards higher expression in inbred population at a lower temperature. These results provide direct experimental evidence that hybrids can cope with the intrinsic stress even at non-stressful temperatures. The constitutive Hsp70 may therefore be used for marker-assisted selection in a breeding programme.

Divergent selection and realized heritability for growth in the Japanese scallop, Patinopecten yessoensis Jay

In order to improve the production and accurately estimate response to selection, divergent selection for growth in shell height was conducted in a cultured population of the Japanese scallop Patinopecten yessoensis. Applying the same selection intensity +/- 1.756 in upward and downward directions, three groups including two selected groups of Fast and Slow and one non-selected Control group were created, which were reared under the same environmental conditions at any stage. Differences always significantly existed among the three groups (P < 0.05), except for larvae at day 1 and at day 5, and in the order of Fast > Control > Slow. The average standardized response to selection (SR), realized heritability (h(R)2) and genetic gain (GG) was 0.473%, 0.269% and 7.85% for the Fast group and 0.381%, 0.217% and 6.60% for the Slow group respectively. Moreover, significant differences (P < 0.05) were detected between the fast and the slow lines in both SR and h(R)2, providing evidence for an asymmetric response in two directions. Performance in shell height is improved by 7.85% in the fast line after one generation selection, suggesting that mass selection for faster growth in a cultured population of the Japanese scallop is effective.

Effect of parental stock size on F-1 genetic structure in the bay scallop, Argopecten irradians (Lamarck, 1819)

Three F-1 families of the bay scallop, Argopecten irradians, were produced from one, two and 10 individuals. The genetic changes in these populations, which suffered recent and different levels of bottleneck, were analysed using amplified fragment length polymorphism (AFLP) techniques. In the parental stock, a total of 330 bands were detected using seven AFLP primer pairs, and 70% of the loci were polymorphic. All F-1 groups had a significantly lower proportion of polymorphic loci when compared with the initial stock, and loss of the rare loci and reduction in heterozygosity both occurred. The progeny of the larger population (i.e., N=10) exhibited a lesser amount of genetic differentiation compared with the progeny from N=2, which showed lesser differentiation than progeny from N=1. The effective population sizes (N-e) in N=1, 2 and 10 were estimated as 1.50, 1.61 and 2.49. Based on regression analysis, we recommend that at least 340 individuals be used in hatchery populations to maintain genetic variation.

Different responses to selection in two stocks of the bay scallop, Argopecten irradians irradians Lamarck (1819)

Two different stocks (A and B) of the bay scallop Argopecten irradialls irradians (Lamarck, 1819) were used to test mass selection on growth. Stock A was a descending stock from the initial introduction from U.S.A. in 1982, which had been cultured in China for about 20 years. Stock B was the third generation from a recent introduction from U.S.A. in 1999. Truncation selection was conducted by selecting the largest 11% scallops in shell length from Stock A and the largest 12.7% scallops from Stock B as parents for the respective selected groups. Before the removal of parents for truncation selection, equal numbers of scallops were randomly chosen from Stock A and B to serve as parents for the control groups. Offspring from the four groups were reared under the same hatchery, nursery, and grow-out conditions. Values of response to selection and realized heritability at larvae, spat and grow-out stages for Stock B were all significantly (P < 0.001) higher than its counterpart for Stock A. For Stock A, no significant response to selection was observed (P > 0.05) at any stage, and the realized heritability for shell length was 0.015 +/- 0.024 for larvae, 0.040 +/- 0.027 for spat, and 0.080 +/- 0.009 for grow-out, respectively. For Stock B, however, significant (P < 0.05) response to selection was observed, and the realized heritability for shell length was 0.511 +/- 0.010 for larvae, 0.341 +/- 0.022 for spat, and 0.338 +/- 0.015 for grow-out. On average, responses to selection at the three stages for Stock B was 30 x, 7.1 x, and 3 x higher than its counterpart for Stock A, respectively. Accordingly, realized heritability at above stages for Stock B was 33 X, 7.5 x, and 3.2 X higher than its counterpart for Stock A, respectively. Differences in response to selection and realized heritability between the two stocks are presumably due to differences in genetic variability. As the 20th generation from the initial introduction consisted of only 26 scallops, Stock A is known to be highly inbred, while inbreeding in Stock B is negligible. (C) 2004 Elsevier B.V. All rights reserved.

Sustained response to selection in an introduced population of the hermaphroditic bay scallop Argopecten irradians irradians Lamarck (1819)

Introduced species often start with limited genetic variability, which is problematic for selective breeding. The problem of inbreeding can be exasperated by hermaphroditism. The bay scallop Argopecten irradians irradians is a hermaphroditic species that has been introduced to and now supports a major aquaculture industry in China. Positive response to selection for fast growth was observed in one of the less inbred stocks in a previous study. In this study, we evaluated selection for the second generation to determine if response to selection can be sustained in this introduced population of a hermaphroditic species. Response to selection, realized heritability for the second generation, cumulative (over two generations), current (for the second generation) and residual (from the first generation) genetic gains were estimated by comparing three different types of lines: SS (selected for two generation), SC (selected for the first generation only) and CC (unselected for two generations). The SS line grew significantly faster (P < 0.05) than the other two lines, indicating that the second generation selection for faster growth is still effective. Response to selection and realized heritability for the second generation were 0.612 +/- 0.101 and 0.349 +/- 0.057, respectively, which are similar to those observed for the first generation. The cumulative, current, and residual gains were 17.56 +/- 5.30%, 10.63 +/- 2.46%, and 6.25 +/- 3.13%, respectively. The sustained response to selection for the second generation observed here suggests that considerable genetic variability exists in this population and that future efforts on selective breeding are likely to be fruitful. (c) 2005 Elsevier B.V. All rights reserved.

An application of the analytic hierarchy process and fuzzy logic inference in a decision support system for forage selection

Forage selection plays a prominent role in the process of returning cultivated lands back into grasslands. The conventional method of selecting forage species can only provide attempts for problem-solving without considering the relationships among the decision factors globally. Therefore, this study is dedicated to developing a decision support system to help farmers correctly select suitable forage species for the target sites. After collecting data through a field study, we developed this decision support system. It consists of three steps: (1) the analytic hierarchy process (AHP), (2) weights determination, and (3) decision making. In the first step, six factors influencing forage growth were selected by reviewing the related references and by interviewing experts. Then a fuzzy matrix was devised to determine the weight of each factor in the second step. Finally, a gradual alternative decision support system was created to help farmers choose suitable forage species for their lands in the third step. The results showed that the AHP and fuzzy logic are useful for forage selection decision making, and the proposed system can provide accurate results in a certain area (Gansu Province) of China.

(Semi-)Predictive Discretization During Model Selection

In this paper, we present an approach to discretizing multivariate continuous data while learning the structure of a graphical model. We derive the joint scoring function from the principle of predictive accuracy, which inherently ensures the optimal trade-off between goodness of fit and model complexity (including the number of discretization levels). Using the so-called finest grid implied by the data, our scoring function depends only on the number of data points in the various discretization levels. Not only can it be computed efficiently, but it is also independent of the metric used in the continuous space. Our experiments with gene expression data show that discretization plays a crucial role regarding the resulting network structure.

Learning and Example Selection for Object and Pattern Detection

This thesis presents a learning based approach for detecting classes of objects and patterns with variable image appearance but highly predictable image boundaries. It consists of two parts. In part one, we introduce our object and pattern detection approach using a concrete human face detection example. The approach first builds a distribution-based model of the target pattern class in an appropriate feature space to describe the target's variable image appearance. It then learns from examples a similarity measure for matching new patterns against the distribution-based target model. The approach makes few assumptions about the target pattern class and should therefore be fairly general, as long as the target class has predictable image boundaries. Because our object and pattern detection approach is very much learning-based, how well a system eventually performs depends heavily on the quality of training examples it receives. The second part of this thesis looks at how one can select high quality examples for function approximation learning tasks. We propose an {em active learning} formulation for function approximation, and show for three specific approximation function classes, that the active example selection strategy learns its target with fewer data samples than random sampling. We then simplify the original active learning formulation, and show how it leads to a tractable example selection paradigm, suitable for use in many object and pattern detection problems.

Model Selection for Solving Kinematics Problems

There has been much interest in the area of model-based reasoning within the Artificial Intelligence community, particularly in its application to diagnosis and troubleshooting. The core issue in this thesis, simply put, is, model-based reasoning is fine, but whence the model? Where do the models come from? How do we know we have the right models? What does the right model mean anyway? Our work has three major components. The first component deals with how we determine whether a piece of information is relevant to solving a problem. We have three ways of determining relevance: derivational, situational and an order-of-magnitude reasoning process. The second component deals with the defining and building of models for solving problems. We identify these models, determine what we need to know about them, and importantly, determine when they are appropriate. Currently, the system has a collection of four basic models and two hybrid models. This collection of models has been successfully tested on a set of fifteen simple kinematics problems. The third major component of our work deals with how the models are selected.