1. Drug Candidate Prediction

(1) Drug-drug interaction(DDI) prediction
















Drug-drug interaction (DDI) is a change in drug effect when taken with the other drugs. Identification of DDIs is important in drug discovery since those interactions can cause antagonistic and toxic effects as well as synergistic. Moreover, combination therapy has been focused on because of its advantages on drug resistance and efficacy compared to monotherapy. Thus, we develop a computational model to predict DDIs using omics data and network analysis. Machine learning algorithms specifically classification algorithms are adopted to construct prediction models with calculated features from network analysis and each data type property information.

(2) Drug repositioning prediction
















Drug repositioning is identifying new uses of existing drugs, which can contribute to a remarkable reduction in cost and time spent for traditional de novo drug discovery. Since drug repositioning deals with existing drugs which are already qualified their safety, it has attracted interests from researchers and pharmaceutical industry. There has been much effort on designing effective drug repositioning approaches including expression profile analysis, network analysis, pathway-based method and so on.


In our lab, we study on similarity-based method for developing prediction models for repositioning candidates with machine learning algorithms such as linear regression, SVM, and random forest.

2. Prediction of Drug-Target Interaction 












Drugs activate and inhibit biological activity of target protein by binding. Thus, identification of drug-target interaction plays important role in drug discovery. However, biological and chemical experiments are laborious so cost a lot. Therefore, identification of drug-target interaction by computational methods is good alternative model to reduce cost with accumulation of drug-target interaction data and advance in data science.

To construct prediction model, first, gather drug-target data from various source (DrugBank, KEGG, IUPHAR). Second, transform the data to vectors, which is understandable for computer. Finally, Build prediction model with machine learning methods (SVM, k-NN, KRM).

3. Drug Toxicity Prediction & Optimization

Many marketed drugs are withdrawn for the undesired adverse effects, in other words, toxicity. It causes a tremendous amount of monetary loss, therefore, drug toxicity assessment and optimization are the most crucial part during the drug discovery process. However, conventional experimental assessment and optimization are time-consuming tasks and cost-ineffective; thus the computational method is obviously promising as it reduces the cost and time significantly. As the accumulation of public experimental data, we have been studying the field of data-driven in silico prediction. In our lab, we are working on developing the frameworks to predict the compound’s toxicities from collecting the public data to making reliable and interpretable results. In particular, we have been fascinated by the deep learning approaches, therefore, we are recently focusing on developing deep learning-based frameworks which are apposite to the toxicity data. 

Computational Systems Biology Lab.

School of Electrical Engineering and Computer Science (EECS),

Gwangju Institute of Science and Technology (GIST), 123 Cheomdangwagi-ro, Buk-gu, Gwangju, 61005, Republic of Korea