Prediction of Drug-Induced Liver Toxicity Using SVM and Optimal Descriptor Sets

Abstract

Drug-induced liver toxicity is one of the significant safety challenges for the patient’s health and the pharmaceutical industry. It causes termination of drug candidates in clinical trials and also the retractions of approved drugs from the market. Thus, it is essential to identify hepatotoxic compounds in the initial stages of drug development process. The purpose of this study is to construct quantitative structure activity relationship models using machine learning algorithms and systematical feature selection methods for molecular descriptor sets. The models were built from a large and diverse set of 1253 drug compounds and were validated internally with 10-fold cross-validation. In this study, we applied a variety of feature selection techniques to extract the optimal subset of descriptors as modeling features to improve the prediction performance. Experimental results suggested that the support vector machine-based classifier had achieved a better classification accuracy with reduced molecular descriptors. The final optimal model provides an accuracy of 0.811, a sensitivity of 0.840, a specificity of 0.783 and Mathew’s correlation coefficient of 0.623 with an internal validation set. Furthermore, this model outperformed the prior studies while evaluated in both the internal and external test sets. The utilization of distinct optimal molecular descriptors as modeling features produce an in silico model with a superior performance.

Drug-Induced Immune Thrombocytopenia Toxicity Prediction Based on Machine Learning

“…Pervious research demonstrated the importance of dimensionality reduction for toxicity prediction, which can enhance prediction performance, increase interpretability, and reduce computational complexity [ 16 , 29 ]. Thus, eliminating redundant and irrelevant features is considered desirable for our prediction modeling.…”

“…For instance, Mansouri et al developed a number of models to predict the logarithmic acid dissociation constant pKa of compounds using a series of classic machine learning-based methods, including extreme gradient boosting (XGBoost), k-nearest neighbor (k-NN), and support vector machine (SVM), and these models outperformed the corresponding commercial models [ 14 ]. Jaganathan and co-workers used the SVM classifier to generate a prediction model of hepatotoxic compounds, and the optimal model outperformed the models reported in the previous studies [ 15 ], achieving an overall prediction accuracy of 81.1% and 75.6% on the internal and external validation sets, respectively [ 16 ]. In addition, various excellent structure–activity relationship (SAR) models have been developed for quickly determining the properties of compounds by their structures.…”

ACP-ADA: A Boosting Method with Data Augmentation for Improved Prediction of Anticancer Peptides

“…The AAINDEX database’s 544 physicochemical attributes were retrieved, returning a total of 531 physicochemical characteristics to represent each residue in the peptide sequence; any physicochemical qualities for any of the amino acids that were removed are indicated with “NA”. The AAINDEX descriptor can be used to encode peptides of the same length [ 32 ]. When Lx is set to 40, the AAINDEX descriptor for a peptide of length 40 produces a feature vector with a dimension of 21,240, which is excessively high and results in a dimension disaster.…”

Rapid Detection of Direct Compound Toxicity and Trailing Detection of Indirect Cell Metabolite Toxicity in a 96-Well Fluidic Culture Device for Cell-Based Screening Environments: Tactics in Six Sigma Quality Control Charts

“…In the framework of precision medicine, DILI is a subgroup of liver injury and is one reason that new drug compounds are removed from late-stage clinical trials and after market entry [65][66][67]. Typically, static cultures or animal studies are deemed unreliable because DILI can be induced by parental drugs or subsequent cell-generated metabolites [55], indicating DILI as a longer-term chronic hepatotoxicity injury model with explicit cell signaling implications, whereas short-term in vitro cultures of hepatocytes, in static setups, are limited due to omitted cell signaling dynamics.…”