New method analyzes how drugs affect gene networks

Modern drug development is a lengthy, lavish, and laborious process largely because biological systems are complex and dynamic.

According to the California Biomedical Research Association, it takes about 12 years and $359 million for a drug to complete its journey from inception in a research lab to patients. Only 10 percent of pre-clinical drugs successfully reach human trials. Of these, only one in five reaches the market.

Wei Wu, a computational biologist at the Ray and Stephanie Lane Center for Computational Biology at Carnegie Mellon, believes that “if we understand the changes introduced by a drug at the molecular level, we can design smarter drugs that have a higher success rate.” To do so, Wu and her team have developed a computational method to analyze expression data from breast cancer cells collected by collaborators at Lawrence Berkeley National Laboratory.

Their method, detailed in the online scholarly journal PLOS Computational Biology, enables researchers to understand how a drug influences changes in gene networks. This knowledge can provide insight as to why some drugs appear to work at first, but might fail at curing a disease in the long run.

In a cell, proteins of specific shapes and sizes interact in a vast, tightly regulated network that determines their production, function, and eventual destruction. Many diseases, such as cancer, are caused by proteins that are made incorrectly, allowing them to escape regulation and negatively impact biological processes. Modern drug development aims to identify the misguided proteins and use chemicals to alter their functions and eventually cure diseases.

Unfortunately, this is easier said than done. Altering a biological pathway can have unexpected consequences. Cells will adapt by activating other pathways, or other proteins in the same pathway, to compensate for and overcome the effect of the drug. These changes can introduce new symptoms or diseases, or simply render the drug useless.

Using breast cancer cells, the group's method harnesses new techniques to identify the adaptations at the molecular level and explain the effects of a drug.

At Berkeley Lab, Wu worked under Mina Bissell, a breast cancer biologist. Bissell developed a unique 3-D cell culturing technique to grow cancer cells in a lab while maintaining their native 3-D structure in a patient. Older methods consisted of growing cancer cells on the flat 2-D surface of a petri dish. “This technique is vital because the interactions between cells in a 3-D environment can largely impact the protein networks within the cell,” Wu said.

Exposing these cells to different cancer drugs mimics the treatment in a patient. Bissell and her team use microarrays to measure the expression of thousands of genes in five different states: normal cells, cancer cells, and three types of reverted-cured cells after the treatment.

Researchers can use this expression data to make inferences about active protein pathways. Unfortunately, with previous methods, a large amount of expression data was required to infer the correct pathway, and researchers are limited by the number of microarrays they can afford and have time to carry out. To overcome this problem, computational biologists have pooled the data from multiple cell states to determine significant pathways.

Wu said that in the group's study, the goal is to differentiate the pathways in the different cell states, so simply pooling all data does not suffice. To solve this problem, she collaborated with Eric P. Xing, a professor of machine learning at Carnegie Mellon, and together they developed Treegl.

Wu explained, “Treegl is a powerful computational method that pools the data from the different cell states while also identifying similarities and differences. By identifying the similarities, the method discovers statistically significant pathways while only requiring three microarrays for each of the five cell states.” More importantly, the method identifies the differences in the active pathways between the different cell states, highlighting the effects of the drug at the molecular level.

A potential approach for cancer drugs that has garnered interest involves inhibiting MMP proteases, which naturally break down other proteins in the cell.

Wu described that with their approach, “after treatment with MMP inhibitors, the newly active pathways showed how cells were able to compensate for the lack of MMP, thus resisting the effect of the drug. This provides a potential explanation as to why this treatment approach shows a low success rate within patients.”

The earliest tangible records of cancer in humans date back to the Egyptians, who observed cancerous overgrowths but were unable to determine their causes. The past few decades have seen a burst of knowledge in molecular biology, enabling researchers to better understand cancer and make significant strides toward developing viable cures.

Sidney Farber, widely regarded as the father of modern chemotherapy, saw promise in using powerful chemical agents to defeat cancer. Countless times, he saw the children he treated improve from a combination of drugs to only fall ill once again to a cancer that adapted to the drugs and returned stronger than before.
Wu strongly believes that the work of her team and her collaborators has immense potential and real-world application, including the ability to understand the effects of the drugs they test and combine drugs to overcome a disease’s ability to adapt.

After losing many patients, Farber admitted that fighting cancer would be an arduous task, requiring a multidisciplinary approach. His vision has now become a reality, as this new method combines the power of computation with modern biological techniques to devise strategic approaches to defeat the most challenging diseases.