Current methods for drug development are one of the major factors standing between society and affordable health care. Often the possible side effects of a drug are not discovered until after massive amounts of money have been invested in their testing. This testing includes exposing hundreds of people to the drug in clinical trials. To recoup the losses incurred by late-stage drug failures, pharmaceutical companies increase the price of drugs that do find their way to market. Science is increasingly looking for effective methods to circumvent this inefficient process in which many people can be harmed. Research published earlier this month by scientists at the University of California, San Diego is exploring a new approach to identifying possible off-target drug effects before the lengthy, expensive and possibly dangerous clinical trials begin. "New techniques are required to understand how multiple protein-drug interactions affect clinical outcomes. Our chemical systems biology approach is developed for this purpose," says Xie, lead author on the paper.

The chemical systems approach to predicting off-target drug effects is a modification of an approach that has been used in the past known as chemogenomics. Most drugs act by binding a target protein, or receptor, and blocking a specific, unwanted activity.  Unfortunately, it is not uncommon for these small molecules to also bind other, off-target proteins - sometimes with disastrous effects. Chemogenomics attempts to predict off-target binding possibilities using known information about protein receptors and the structures that bind them. The drawback is that a lot is still unknown about how binding interactions are governed, and for novel drugs this method may not be an effective tool for predictio

The chemical systems approach developed by Xi and colleagues differs from chemogenomics because it focuses solely on the receptor. The first step is to characterize the 3D structure of the receptor. The 3D structure can then be compared to the known or predicted structure of a wide array of other receptors. When proteins with similar binding sites are identified, the interplay between these putative targets and the drug are further modeled to determine their atomic interaction. Targets that receive a high docking score-suggesting a more likely interaction- are included in a panel which then undergoes further analysis. Another important difference between the chemical systems approach and chemogenomics is that Xie and his colleagues then incorporate functional data about the targets, including their involvement in metabolic pathways, signal transduction pathways or gene regulation pathways. By integrating the roles of predicted targets the researchers can better predict the overall effect of the off-target interaction

To demonstrate their method the authors turned their attention to Torcetrapib, a once promising drug for the treatment of cardiovascular disease that was pulled out of the development process during phase III clinical trials after 15 years and $800 million dollars had been invested toward bringing it to market. It was not until this late stage of development that clinical trials showed the drug had deadly side effects in people with hypertension.

When Xie and his colleagues used Torcetrapib as the model for their chemical systems approach they came up with 273 putative targets. Most of these proteins belonged to six broad classes of molecules. One of these classes included the nuclear hormone receptors that control the Renin-Angiotension-Aldosterone System (RAAS), the main system for blood pressure regulation in the body. This result explained the fatal hypertension side effects that caused to the drug to be yanked from testing.Currently, the information gained from predictive models does not determine if a drug candidate is pursued. "I think that the major issue in applying a prediction model is not technical but cultural. It is expected that after 5-10 years we will have sufficient data to build complicated models. Today, even with limited data, the predictive model is able to give invaluable results. However, mindset in the pharmaceutical industry is resistant to such changes. I am not sure when they will change their mind," says Xie. Accurately predicting drug side effects will save millions of lives by making drugs more affordable and accessible to the general public and by decreasing the fatality rates during clinical trials. Torcetrapib foreshadows the promise that chemogenomics, the chemical systems approach and other predictive tools hold for revolutionizing medicine.

To read the article click here

or go to http://www.ploscompbiol.org/article/info%3Adoi%2F10.1371%2Fjournal.pcbi.1000387