banner by Kirsten Wade
The Bit-By-Bit Emergence of In-Silico Drug Design
by Noah Gafen
There are two general models in medicinal research today with which labs test the effectiveness of drug candidates. These models, in vitro and in vivo, have each proven invaluable in our efforts to cure and prevent disease. In vitro testing allows investigators to test treatment mechanisms in externally reproduced biological systems, whereas in vivo testing assesses their practicality in whole organisms, with full biological complexity intact. While both have been crucial to the success of drug research in recent times, they are unavoidably imperfect. In vitro studies focus on the bare bones of biochemical processes which is often hard to translate into a practical drug option. In vivo studies such as clinical trials are long, expensive, and occasionally inconclusive.
Fortunately, there is an emerging third research model that has the potential to revolutionize the process of drug discovery. In silico research, as you may guess from the name, is the way investigators have started to incorporate computational methods into the pharmaceutical industry. The idea is simple: use computers to model specific biological systems in which prospective drugs can undergo initial testing. In silico systems would not immediately toss the older models aside, but they would serve to test or even design drugs that could then undergo in vitro and in vivo testing before hitting the market. Computer simulations possess the speed and cost effectiveness that the current models so obviously lack. Ideally, a combination of these three models will dramatically increase the rate and efficiency of drug discovery.
The basis of computer modeling of biological systems is rooted in and limited by our understanding of physical biochemistry. The more we understand about the nature of how chemical reactions in our bodies drive physiological activity the more accurate the in silico model would be. For instance, many drugs are administered directly into a vein so that the drug can immediately circulate the body via the bloodstream. A computer model developed to test these candidate drugs would have to account for every location the drug will travel, the receptors or target proteins present in cells therein, their respective probabilities of recognizing the drug, and the potential physiological outcomes upon binding. It is necessary that the model be designed such that the specific chemical interactions and environments are understood and appropriately implemented.
While much of the applications of in silico modeling are currently hypothetical, there do exist early success stories. Dr. Georgios Stamatakos from the National Technical University of Athens has headed a collaborative effort to design the oncosimulator, an in silico model of malignant tumors. As a part of the larger Advanced Clinical-Genomic Trials on Cancer Project (ACGT), the oncosimulator accounts for the genetic identity of a specific patient’s tumor cells so that it can predict the cancer’s response to treatment by a variety of therapeutics. While the technology is young and untested clinically, it has been used alongside clinical treatment to predict the patient’s response to chemotherapy. During a trial of children with nephroblastoma, the oncosimulator successfully predicted the change in tumor size following treatment. Furthermore, the extensive data collected was used to help optimize the model’s ability to serve as a better predictor of future in vivo tests.
In silico endeavors are also being undertaken closer to home. The University of Pittsburgh has been using in silico models to help identify new drugs for treatment of patients with cystic fibrosis. A molecule present in the endoplasmic reticulum of mammalian cells called Hsp70 has been found to be crucial for the way cells regulate the folding of proteins, a process that goes awry in cystic fibrosis. 15-deoxyspergualin (DSG) has been shown to provide some counteraction by binding to Hsp70 as a weak activator but it is not nearly perfect. The National Cancer Institute had their Developmental Therapeutics Program use computational modeling to help identify compounds that shared structural similarity to DSG in the hopes that these compounds might also bind to Hsp70 but with better results. At the University of Pittsburgh, collaborative efforts led to the synthesis of more than 30 novel molecules that underwent subsequent testing. The result was the discovery of a number of molecules that actually inhibit the function of Hsp70. Additionally, while the majority were found to be allosteric inhibitors (molecules that bind and cause a functional change to their target molecules), two of them disrupted Hsp70 function by blocking the binding of its coactivator Hsp40. While this research may not have directly benefited cystic fibrosis treatment yet, the new Hsp70 inhibitors could yet provide medical utility as Hsp70 has a variety of roles that it plays in cell homeostasis.
The ultimate goal of those developing in silico models is to create a Virtual Physiological Human (VPH). Essentially simulated patients, such a model would be the closest thing we can ethically have to a “living cadaver” that can undergo uninhibited testing. While clinical trials have been good enough up to now, they have serious limitations. If a complete human being could be simulated, then we could “treat” millions of sick “patients” with any drug imaginable and see the effect each may have. The scale is clearly enormous, and better yet an in silico “clinical trial” could take place over the course of days as opposed to years. Ideally, fewer people will be stuck with a hopeless drug candidate or a placebo in a last resort clinical trial. However, a world in which we can achieve such an amazing breakthrough is still quite far off. As it stands, the computing power of even our most advanced supercomputers is insufficient for what a virtual clinical trial would require. There is no doubt, however, that our computational capabilities will continue to improve into the future. But for now, we will happily settle for the short-term benefits of in silico drug discovery that include a reduction in the number of hopeless clinical trials.
An astonishing amount of qualitative knowledge about human physiology has been accumulated over the years. To make practical use of this knowledge as a VPH, the knowledge must be translated into quantitative mathematical application. So much biochemistry has been characterized, but the challenge now is to convert our knowledge into what is essentially a video game where researchers can tinker with and study how our cells respond to a virtually unlimited variety of chemical input. If perfected, though it may take some time, it would be a gamechanger. No longer would clinical trials for promising drugs take years only to discover that the drug provides little efficacy. Instead, drug candidates can be screened beforehand to see if additional research pursuits would be worthwhile. The money and time saved by this not-so-distant technology would greatly improve our ability to make lifesaving therapeutics and in silico methods could prove to be a trailblazer into the frontier of human health.