Using SMOTE to Deal with Class-Imbalance Problem in Bioactivity Data to Predict mTOR Inhibitors

Abstract

Machine learning algorithms give sub-optimal performance in the presence of class-imbalanced dataset. Mammalian target of rapamycin (mTOR) is one of the serine/threonine protein kinase, and plays an integral role in autophagy pathway. Autophagy is a cellular pathway for recycling of macromolecules (proteins, lipids, and organelles), which enables eukaryotic cells to adapt metabolism to survive during adverse growth conditions. Targeting mTOR through therapeutic interventions of autophagy pathway establishes mTOR a promising pharmacological target for autophagy modulation in cancer. The bioactivity dataset of mTOR in ChEMBL, a compound bioactivity database maintained by European Bioinformatics Institute, shows disproportionate distribution of active and inactive classes. The predictive models based on this skewed dataset are biased towards prediction of majority class. Hence, we have used Synthetic Minority Over-sampling TEchnique to deal with class-imbalance problem in bioactivity datasets. We have built and evaluated predictive models based on four commonly used classifiers using both class-imbalanced and class-balanced bioactivity datasets, and compared their performance based on various metrics like accuracy, sensitivity, specificity, F1-measure, and AUC . We observe that the classification models based on balanced dataset generally outperform those that are based on class-imbalanced dataset, irrespective of the classifiers used for classification task. We conclude that predictive models trained over class-balanced dataset can be used for screening large compound bioactivity datasets to predict mTOR inhibitors-like compounds.

Boosting methods for multi-class imbalanced data classification: an experimental review

“…For solving these problems, Chawla proposed the Synthetic Minority Over-sampling technique (SMOTE), which creates synthetic samples from the minority class. The SMOTE samples are linear combinations of two similar samples from the minority class [29].…”

Network anomaly detection using deep learning techniques

“…To overcome this problem, Chawla [13] proposed the SMOTE technique, which generates synthetic samples from the minority class. Samples created by the SMOTE technique are a linear combination of two identical samples from the minority class [13, 26, 28]. The SMOTE over‐sampling algorithm works as follows: Let S represent the size of a small class, considering a sample j of the small size class, and x j denotes its feature vector such that, j ∈ {1, …, S }: Find k neighbours of the sample x j from all S (using the Euclidean Distance for example) and denoted it as x j ( near ) , near ∈ {1, …, k }. x j ( nn ) sample is selected randomly from the k neighbours, and the random number β 1 between 0 and 1 is generated to synthesise a new sample x j 1 as in the equation x j 1 = x j + β 1 ∗ ( x j ( nn ) − x j ). Step 2 is repeated M times to synthesise M new samples: x j ( new ) , new ∈ {1, …, M }. …”

“…Over-sampling techniques add some more instances to the minority class in the training set, the simplest method is random over-sampling [26] but the drawback is over-fitting [27]. To overcome this problem, Chawla [13] proposed the SMOTE technique, which generates synthetic samples from the minority class. Samples created by the SMOTE technique are a linear combination of two identical samples from the minority class [13,26,28].…”

“…To resolve an imbalance issue, there are many sampling techniques. One of the most commonly used of such techniques is the synthetic minority over-sampling technique (SMOTE), which is proposed by Chawla [13]. The SMOTE method generates additional synthetic samples from the minority class to balance the data.…”

Multitask CapsNet: An Imbalanced Data Deep Learning Method for Predicting Toxicants

“…Moreover, Tox21 is severely imbalanced; the volume of the inactive (negative/nontoxic) data set is much larger than that of the active (positive/toxic) data set. As a result, multitask deep learning models are unable to thoroughly explore the essence of the minority class data set consisting of the positive compounds. , To address this issue, several data augmentation approaches have been introduced in previous studies, such as the resampling effect and the synthetic minority oversampling technique (SMOTE). , Manual preprocessing is an essential ingredient of these data augmentation strategies, which may impact the objectivity and performance of the models to some extent. Data augmentation technologies are also not suitable for direct use in chemicals.…”