CDRUG is a Web Server for predicting anticancer efficacy of chemical compounds. 
It is based on NCI60 database and is powerful for large scale prediction.
Each line should contain a compound's SMILE and a one-word name.
For example:
O=C(c3nccc(Oc2ccc(NC(=O)Nc1cc(c(Cl)cc1)C(F)(F)F)c(F)c2)c3)NC Regorafenib
Cn1cnc2c1c(=O)n(c(=O)n2C)C caffeine
c1c(ccc2C(=O)c3ccccc3C(=O)c12)OC 7920


1. HSCORE method can be downloaded(15.6M): HSCORE.rar.

2. Standard alone version of CDRUG can be downloaded(16.6M): cdrug_standard_alone.rar.

3. Benchmark dataset of CDRUG can be downloaded(0.3M): Benchmark.rar.

4. Cross-validation datasets of CDRUG can be downloaded: cross_validation_datasets.xls.

Description of CDRUG

1. CDRUG is rapid. A query with 1~20 compounds requires 35 seconds, while a query with 1000 compounds requires about 4 minutes.

2. CDRUG is effective. CDRUG identifies about 65% anticancer compounds at the false positive rate of 0.05.

3. Known limitation. Because CDRUG uses NCI-60 DTP database(in vitro), some "indirect" anticancer compounds such as angiogenesis inhibitors, EGFR inhibitors, may not hit a signifanct result in CDRUG. We may cover this limitation in the futher through carefully adding the in vivo datasets.

4. The maximum number of compound in a query is 1000. If you want to implement a huge dataset prediction, such as all pubchem compounds(>ten million), please do not hesitate to contact me.


Gong-Hua Li and Jing-Fei Huang.
CDRUG: a web server for predicting anticancer activity of chemical compounds. 
Bioinformatics 2012 Dec; 28(24):3334-3335


Dr. Gong-Hua Li, 
Kunming Institute of Zoology, Chinese Academy of Sciences,China