Efficient and biologically relevant consensus strategy for Parkinson’s disease gene prioritization

Cruz-Monteagudo, Maykel; Borges, Fernanda; Paz-y-Miño, Cesar; Cordeiro, M. Natália D. S.; Rebelo, Irene; Perez-Castillo, Yunierkis; Helguera, Aliuska Morales; Sánchez-Rodríguez, Aminael; Tejera, Eduardo

doi:10.1186/s12920-016-0173-x

BMC Medical Genomics

Table 4 Classification performance of the ML classification algorithms used to identify PD relevant sets of genes

From: Efficient and biologically relevant consensus strategy for Parkinson’s disease gene prioritization

ML Classification Algorithm	Training set			LOO CV			5-Fold CV			Test set
ML Classification Algorithm	Acc.	Se.	Sp.	Acc.	Se.	Sp.	Acc.	Se.	Sp.	Acc.	Se.	Sp.
functions.SimpleLogistic	1.000	1.000	1.000	0.827	0.860	0.781	0.827	0.814	0.844	0.704	0.750	0.636
rules.MODLEM	1.000	1.000	1.000	0.813	0.837	0.781	0.760	0.767	0.750	0.778	0.750	0.818
rules.PART	0.987	0.977	1.000	0.653	0.674	0.625	0.747	0.721	0.781	0.741	0.750	0.727
trees.ADTree	1.000	1.000	1.000	0.853	0.860	0.844	0.787	0.721	0.875	0.741	0.750	0.727
trees.BFTree	0.973	1.000	0.938	0.853	0.884	0.813	0.747	0.744	0.750	0.741	0.750	0.727
trees.FT	1.000	1.000	1.000	0.800	0.837	0.750	0.867	0.884	0.844	0.741	0.813	0.636
trees.LADTree	1.000	1.000	1.000	0.840	0.884	0.781	0.827	0.814	0.844	0.889	0.875	0.909
trees.LMT	1.000	1.000	1.000	0.813	0.860	0.750	0.773	0.767	0.781	0.741	0.813	0.636
trees.SimpleCart	0.973	1.000	0.938	0.827	0.837	0.813	0.747	0.721	0.781	0.741	0.750	0.727
meta.AdaBoostM1	1.000	1.000	1.000	0.840	0.884	0.781	0.880	0.907	0.844	0.926	1.000	0.818
meta.AttributeSelectedClassifier	0.960	0.977	0.938	0.680	0.721	0.625	0.760	0.767	0.750	0.852	0.875	0.818
meta.ClassificationViaRegression	0.960	0.977	0.938	0.813	0.814	0.813	0.733	0.698	0.781	0.815	0.938	0.636
meta.Decorate	1.000	1.000	1.000	0.893	0.860	0.938	0.867	0.837	0.906	0.963	1.000	0.909
AVERAGE	0.989	0.995	0.981	0.808	0.832	0.777	0.794	0.782	0.810	0.798	0.832	0.748

Acc. = accuracy or overall classification rate; Se. = sensitivity or true positives rate (% of PD samples correctly classified); Sp. = specificity or true negatives rate (% of HC samples correctly classified)
functions.SimpleLogistic: Classifier for building linear logistic regression models [104]; rules.MODLEM: Class for building and using a MODLEM algorithm to induce rule set for classification [105]; rules.PART: Class for generating a PART decision list [106]; trees.ADTree: Class for generating an alternating decision tree [107]; trees.BFTree: Class for building a best-first decision tree classifier [108]; trees.FT: Classifier for building ‘Functional trees’, which are classification trees that could have logistic regression functions at the inner nodes and/or leaves [109]; trees.LADTree: Class for generating a multi-class alternating decision tree using the LogitBoost strategy [110]; trees.LMT: Classifier for building ‘logistic model trees’, which are classification trees with logistic regression functions at the leaves [104, 111]; trees. SimpleCart: Class implementing a classification and regression tree with minimal cost-complexity pruning [112]; meta.AdaBoostM1: Metaclassifier class for boosting a nominal class classifier using the Adaboost M1 method [113]; meta.AttributeSelectedClassifier: Metaclassifier class where dimensionality of training and test data is reduced by attribute selection before being passed on to a classifier http://weka.sourceforge.net/doc.dev/weka/classifiers/meta/AttributeSelectedClassifier.html; meta.ClassificationViaRegression: Metaclassifier class for doing classification using regression methods [114]; meta.Decorate: Meta-learner for building diverse ensembles of classifiers by using specially constructed artificial training examples [115, 116]

Back to article page

ISSN: 1755-8794

Contact us

Submission enquiries: bmcmedicalgenomics@biomedcentral.com
General enquiries: ORSupport@springernature.com