Evaluation of novel candidate variations and their interactions related to bipolar disorders: Analysis of GWAS data

Abstract

BackgroundMultifactor dimensionality reduction (MDR) is a nonparametric approach that can be used to detect relevant interactions between single-nucleotide polymorphisms (SNPs). The aim of this study was to build the best genomic model based on SNP associations and to identify candidate polymorphisms that are the underlying molecular basis of the bipolar disorders.MethodsThis study was performed on Whole-Genome Association Study of Bipolar Disorder (dbGaP [database of Genotypes and Phenotypes] study accession number: phs000017.v3.p1) data. After preprocessing of the genotyping data, three classification-based data mining methods (ie, random forest, naïve Bayes, and k-nearest neighbor) were performed. Additionally, as a nonparametric, model-free approach, the MDR method was used to evaluate the SNP profiles. The validity of these methods was evaluated using true classification rate, recall (sensitivity), precision (positive predictive value), and F-measure.ResultsRandom forests, naïve Bayes, and k-nearest neighbors identified 16, 13, and ten candidate SNPs, respectively. Surprisingly, the top six SNPs were reported by all three methods. Random forests and k-nearest neighbors were more successful than naïve Bayes, with recall values >0.95. On the other hand, MDR generated a model with comparable predictive performance based on five SNPs. Although different SNP profiles were identified in MDR compared to the classification-based models, all models mapped SNPs to the DOCK10 gene.ConclusionThree classification-based data mining approaches, random forests, naïve Bayes, and k-nearest neighbors, have prioritized similar SNP profiles as predictors of bipolar disorders, in contrast to MDR, which has found different SNPs through analysis of two-way and three-way interactions. The reduced number of associated SNPs discovered by MDR, without loss in the classification performance, would facilitate validation studies and decision support models, and would reduce the cost to develop predictive and diagnostic tests. Nevertheless, we need to emphasize that translation of genomic models to the clinical setting requires models with higher classification performance.

Bayesian and Machine Learning Models for Genomic Prediction of Anterior Cruciate Ligament Rupture in the Canine Model

“…A training set is used to calculate frequencies of genotypes in case or control individuals, and this information is used to calculate the probability of an unknown individual’s classification. NB is known for being simple and computationally efficient, but it is prone to miscalibration when features are high in number, as is the case with SNP datasets ( Acikel et al 2016 ). Though it has been theoretically outclassed by ensemble machine learning methods, NB is still an excellent baseline for comparing classifiers ( Acikel et al 2016 ).…”

“…NB is known for being simple and computationally efficient, but it is prone to miscalibration when features are high in number, as is the case with SNP datasets ( Acikel et al 2016 ). Though it has been theoretically outclassed by ensemble machine learning methods, NB is still an excellent baseline for comparing classifiers ( Acikel et al 2016 ). The R package ‘e1071’ ( Dimitriadou et al 2017 ) was used for NB implementation.…”

“…Instead, KNN compares the unknown input with classification of the k -nearest data points and uses the features of these neighbors to classify the unknown input. If multiple classifications are possible, a majority vote is applied ( Acikel et al 2016 ). However, KNN also struggles when the number of inputs is very large.…”

Machine learning for genetic prediction of psychiatric disorders: a systematic review

“…acquired Bipolar Disorder Only (BDO) participants [28]; Li et al report using the Bipolar and Related Disorders (BARD) subset [29]. Controls, obtained through KN, are described under "Clinical Procedures" of the relevant dbGaP entry, and by other studies [17].…”

Machine learning algorithm performance evaluation in structural magnetic resonance imaging-based classification of pediatric bipolar disorders type I patients

“…For distinguishing BD patients from normal subjects, commonly used ML methods include logistic regression (LR) ( Pirooznia et al, 2012 ), support vector machine (SVM) ( Schnack et al, 2014 ), random forest classifier (RF) ( Besga et al, 2015 ; Chuang and Kuo, 2017 ), naïve Bayes (NB) ( Struyf et al, 2008 ), k-nearest neighbor (kNN) ( Struyf et al, 2008 ; Acikel et al, 2016 ), and so on. Currently, there are relatively few studies on using structural magnetic resonance imaging (sMRI) for BD discrimination, which mainly focused on gray matter (GM) and white matter (WM) density as features to train classifiers with different accuracy rates.…”