Limits...
Identification and optimization of classifier genes from multi-class earthworm microarray dataset.

Li Y, Wang N, Perkins EJ, Zhang C, Gong P - PLoS ONE (2010)

Bottom Line: A variety of toxicological effects have been associated with explosive compounds TNT and RDX.We have developed an earthworm microarray containing 15,208 unique oligo probes and have used it to profile gene expression in 248 earthworms exposed to TNT, RDX or neither.This study demonstrates that the machine learning approach can be used to identify and optimize a small subset of classifier/biomarker genes from high dimensional datasets and generate classification models of acceptable precision for multiple classes.

View Article: PubMed Central - PubMed

Affiliation: School of Computing, University of Southern Mississippi, Hattiesburg, Mississippi, United States of America.

ABSTRACT
Monitoring, assessment and prediction of environmental risks that chemicals pose demand rapid and accurate diagnostic assays. A variety of toxicological effects have been associated with explosive compounds TNT and RDX. One important goal of microarray experiments is to discover novel biomarkers for toxicity evaluation. We have developed an earthworm microarray containing 15,208 unique oligo probes and have used it to profile gene expression in 248 earthworms exposed to TNT, RDX or neither. We assembled a new machine learning pipeline consisting of several well-established feature filtering/selection and classification techniques to analyze the 248-array dataset in order to construct classifier models that can separate earthworm samples into three groups: control, TNT-treated, and RDX-treated. First, a total of 869 genes differentially expressed in response to TNT or RDX exposure were identified using a univariate statistical algorithm of class comparison. Then, decision tree-based algorithms were applied to select a subset of 354 classifier genes, which were ranked by their overall weight of significance. A multiclass support vector machine (MC-SVM) method and an unsupervised K-mean clustering method were applied to independently refine the classifier, producing a smaller subset of 39 and 30 classifier genes, separately, with 11 common genes being potential biomarkers. The combined 58 genes were considered the refined subset and used to build MC-SVM and clustering models with classification accuracy of 83.5% and 56.9%, respectively. This study demonstrates that the machine learning approach can be used to identify and optimize a small subset of classifier/biomarker genes from high dimensional datasets and generate classification models of acceptable precision for multiple classes.

Show MeSH

Related in: MedlinePlus

Classification accuracy of the 248 earthworm samples using an increasing number of classifier genes optimized by SVM (a) or clustering (b).The weighted average accuracy and the accuracy for each of the three classes (control, RDX and TNT) are shown for each set of genes (1∼39 genes in 5(a) or 1∼30 genes in 5(b)). One gene (the next highest ranked gene) at a time was added to the previous gene set to generate a new gene set (see also Figure 3(a)).
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2965664&req=5

pone-0013715-g005: Classification accuracy of the 248 earthworm samples using an increasing number of classifier genes optimized by SVM (a) or clustering (b).The weighted average accuracy and the accuracy for each of the three classes (control, RDX and TNT) are shown for each set of genes (1∼39 genes in 5(a) or 1∼30 genes in 5(b)). One gene (the next highest ranked gene) at a time was added to the previous gene set to generate a new gene set (see also Figure 3(a)).

Mentions: Two different algorithms, SMO and K-mean clustering, were employed to optimize the number and set of genes from the 354 ranked classifier genes. Composition of the classifier gene set had a significant influence on classification accuracy (Fig. 4). Using SMO, as few as 16 top ranked genes classified 81% of the 248 samples into correct classes (Fig. 4a). Starting at the 250th gene, the inclusion of additional classifier genes not only did not improve the classification accuracy for the TNT and the RDX classes as well as the weighted average accuracy, but deteriorated the accuracy for the control class (Fig. 4a). Similarly, with the clustering approach, the top ranked 31 genes correctly clustered 66% of the samples, while addition of other genes did little, if any, to improve the accuracy of either individual classes or the weighted average (Fig. 4b). Clearly, individual classifier genes vary remarkably in its contribution to the change of classification accuracy, which also depends on the choice of machine learning algorithms. The iterative optimization process effectively removed many genes that made no or negative contribution to the classification performance. As a result, this process produced a SVM- and a clustering-optimized subset consisting of 39 and 30 genes, respectively (Fig. 5).


Identification and optimization of classifier genes from multi-class earthworm microarray dataset.

Li Y, Wang N, Perkins EJ, Zhang C, Gong P - PLoS ONE (2010)

Classification accuracy of the 248 earthworm samples using an increasing number of classifier genes optimized by SVM (a) or clustering (b).The weighted average accuracy and the accuracy for each of the three classes (control, RDX and TNT) are shown for each set of genes (1∼39 genes in 5(a) or 1∼30 genes in 5(b)). One gene (the next highest ranked gene) at a time was added to the previous gene set to generate a new gene set (see also Figure 3(a)).
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2965664&req=5

pone-0013715-g005: Classification accuracy of the 248 earthworm samples using an increasing number of classifier genes optimized by SVM (a) or clustering (b).The weighted average accuracy and the accuracy for each of the three classes (control, RDX and TNT) are shown for each set of genes (1∼39 genes in 5(a) or 1∼30 genes in 5(b)). One gene (the next highest ranked gene) at a time was added to the previous gene set to generate a new gene set (see also Figure 3(a)).
Mentions: Two different algorithms, SMO and K-mean clustering, were employed to optimize the number and set of genes from the 354 ranked classifier genes. Composition of the classifier gene set had a significant influence on classification accuracy (Fig. 4). Using SMO, as few as 16 top ranked genes classified 81% of the 248 samples into correct classes (Fig. 4a). Starting at the 250th gene, the inclusion of additional classifier genes not only did not improve the classification accuracy for the TNT and the RDX classes as well as the weighted average accuracy, but deteriorated the accuracy for the control class (Fig. 4a). Similarly, with the clustering approach, the top ranked 31 genes correctly clustered 66% of the samples, while addition of other genes did little, if any, to improve the accuracy of either individual classes or the weighted average (Fig. 4b). Clearly, individual classifier genes vary remarkably in its contribution to the change of classification accuracy, which also depends on the choice of machine learning algorithms. The iterative optimization process effectively removed many genes that made no or negative contribution to the classification performance. As a result, this process produced a SVM- and a clustering-optimized subset consisting of 39 and 30 genes, respectively (Fig. 5).

Bottom Line: A variety of toxicological effects have been associated with explosive compounds TNT and RDX.We have developed an earthworm microarray containing 15,208 unique oligo probes and have used it to profile gene expression in 248 earthworms exposed to TNT, RDX or neither.This study demonstrates that the machine learning approach can be used to identify and optimize a small subset of classifier/biomarker genes from high dimensional datasets and generate classification models of acceptable precision for multiple classes.

View Article: PubMed Central - PubMed

Affiliation: School of Computing, University of Southern Mississippi, Hattiesburg, Mississippi, United States of America.

ABSTRACT
Monitoring, assessment and prediction of environmental risks that chemicals pose demand rapid and accurate diagnostic assays. A variety of toxicological effects have been associated with explosive compounds TNT and RDX. One important goal of microarray experiments is to discover novel biomarkers for toxicity evaluation. We have developed an earthworm microarray containing 15,208 unique oligo probes and have used it to profile gene expression in 248 earthworms exposed to TNT, RDX or neither. We assembled a new machine learning pipeline consisting of several well-established feature filtering/selection and classification techniques to analyze the 248-array dataset in order to construct classifier models that can separate earthworm samples into three groups: control, TNT-treated, and RDX-treated. First, a total of 869 genes differentially expressed in response to TNT or RDX exposure were identified using a univariate statistical algorithm of class comparison. Then, decision tree-based algorithms were applied to select a subset of 354 classifier genes, which were ranked by their overall weight of significance. A multiclass support vector machine (MC-SVM) method and an unsupervised K-mean clustering method were applied to independently refine the classifier, producing a smaller subset of 39 and 30 classifier genes, separately, with 11 common genes being potential biomarkers. The combined 58 genes were considered the refined subset and used to build MC-SVM and clustering models with classification accuracy of 83.5% and 56.9%, respectively. This study demonstrates that the machine learning approach can be used to identify and optimize a small subset of classifier/biomarker genes from high dimensional datasets and generate classification models of acceptable precision for multiple classes.

Show MeSH
Related in: MedlinePlus