Development Process

In the Cancer Mutation To Drug version 2, we ...

Drug Target Predictor is a web server for predicting target proteins of giving compounds. We performed compound-target protein prediction based on the fingerprint of compound and protein binding pocket. The web server used the SMILES of the compound for drug side effect prediction and multiple query can be submitted in one request.

In the Drug Target Predictor version 1, we used random forest (RF) method to perform drug target protein prediction. The benchmark data contain 3743 protein binding pocket information, including all amino acid residues and binary fingerprint features of binding pockets. We also provide RF model file and source code for offline analysis.

In the Drug Side Effect Predictor version 2, we removed the method of neural networks (NN) due to its lower accuracy. The final model use two methods including support vector machine (SVM) and random forest (RF) to perform drug side effect prediction. The benchmark data contain 25756 paired drug-side effect interactions and 25756 randomly generated unpaired drugs-side effect interactions. We also provide SVM model file, RF model file and source code for offline analysis.

Sequence-Based CPI Predictor is a web server for sequence-based compound-protein interaction prediction. All human protein sequences were downloaded from UniProt database, and the output will show all possible binding proteins to the query compound. The web server used the SMILES of the compound for compound-protein interaction prediction and multiple query can be submitted in one request.

In the Sequence-Based CPI Predictor version 1, we used random forest (RF) method to perform sequence-based compound-protein interaction prediction. The benchmark data contain 831617 paired compound-protein interactions and 831617 randomly generated unpaired compound-protein interactions. We also provide benchmark data and source code for offline analysis.

Cancer Mutation To Drug is a web server for anti-cancer drugs prediction based on gene mutation. Our aim is to find the optimal personalized drug(s) to treatment cancer based on personalized genetic background. The input parameter is a list of gene symbols and mutations. The official gene symbol and refSNP number can be found from the NCBI-Gene and NCBI-SNP database.

In the Cancer Mutation To Drug version 1, we construct a composite network of "biomarkers-targets-drugs" by integrating cancer biomarkers, targets and drugs data. Base on the complicated network, we construct a predictive personalized cancer therapeutic model using gene mutation to predict ant-cancer drugs. The benchmark data contain 234278 mutations of 18066 genes and 14834 gene-protein-drug pairs which can download from our web. We also provide source code for offline analysis.

Drug Side Effect Predictor is a web server for predicting side effects of giving compounds. Side effects are defined as drug-induced adverse reactions were more than 1% in the study population. The web server used the SMILES of the compound for drug side effect prediction and multiple query can be submitted in one request.

In the Drug Side Effect Predictor version 1, we used a combination of three methods including support vector machine (SVM), random forest (RF) and neural networks (NN) to perform side effect prediction. The benchmark data contain 1744 different side effects and their corresponding drugs. We also provide SVM model file, RF model file, NN model file and source code for offline analysis.

Anti-Fungi Predictor is a web server for predicting new anti-fungal chemical compounds. The web server is rapid and with high accuracy (AUC > 0.95), which enables us to screen tens of millions of compounds and discover new anti-fungal agents. It used the SMILES of the compound for anti-fungal activity prediction and multiple query can be submitted in one request.

In the Anti-Fungi Predictor version 1, we used a combination of two methods including support vector machine (SVM) and random forest (RF) to perform anti-fungal activity compound prediction. The benchmark data contain 426 active anti-fungal compounds and 205 inactive anti-fungal compounds which can download from our web. We also provide SVM model file, RF model file and source code for offline analysis.

In the Anti-HIV Predictor version 2, we removed the method of relative frequency-weighted Tanimoto coefficient (RFW-TC) due to its lower accuracy. The final model use two methods including support vector machine (SVM) and random forest (RF) to perform anti-HIV activity compound prediction. The benchmark data contain 9584 active anti-HIV compounds and 23998 inactive anti-HIV compounds which can download from our web. We also provide SVM model file, RF model file and source code for offline analysis.

Anti-Bacteria Predictor is a web server for predicting new anti-bacterial chemical compounds by integrating a variety of high-throughput data. The web server is rapid and with high accuracy (AUC > 0.94), which enables us to screen tens of millions of compounds and discover new anti-bacterial agents. It used the SMILES of the compound for anti-bacterial activity prediction and multiple query can be submitted in one request.

In the Anti-Bacteria Predictor version 1, we used a combination of two methods including support vector machine (SVM) and random forest (RF) to perform anti-bacterial activity compound prediction. The benchmark data contain 1097 active anti-bacterial compounds and 578 inactive anti-bacterial compounds which can download from our web. We also provide SVM model file, RF model file and source code for offline analysis.

Anti-Cancer Predictor is a web server for predicting new anti-cancer chemical compounds. The web server is rapid and with high accuracy (AUC > 0.95), which enables us to screen tens of millions of compounds and discover new anti-cancer agents. It used the SMILES of the compound for anti-cancer activity prediction and multiple query can be submitted in one request.

In the Anti-Cancer Predictor version 1, we used a combination of two methods including support vector machine (SVM) and random forest (RF) to perform anti-cancer activity compound prediction. The benchmark data contain 14217 active anti-cancer compounds and 23952 inactive anti-cancer compounds which can download from our web. We also provide SVM model file, RF model file and source code for offline analysis.

Anti-HIV Predictor is a web server for predicting the anti-HIV activity of given compounds with high performance (AUC > 0.96), which enables us to screen tens of millions of compounds and discover new anti-HIV agents. The web server used the SMILES of the compound for anti-HIV activity prediction and multiple query can be submitted in one request.

In the Anti-HIV Predictor version 1, we used a combination of three methods including relative frequency-weighted Tanimoto coefficient (RFW-TC), support vector machine (SVM) and random forest (RF) to perform anti-HIV activity compound prediction. We also provide benchmark data and source code for offline analysis.