Chris Lipinski earned his Ph.D. in physical organic chemistry from the University of California, Berkeley. Following his post-doctoral fellowship at the California Institute of Technology, Dr. Lipinski spent more than 30 years at Pfizer as a Senior Research Fellow where he developed the renowned, rule of 5. He is currently serving as a Scientific Advisor for Melior Discovery and consultant through his own company, Christopher A. Lipinski, Ph.D., LLC.
You are probably familiar with the 'rule of 5' (Ro5), the set of guidelines I introduced in 1997 to help reduce the attrition rate in drug discovery. It proved to be a useful strategy for selecting candidates with good oral bioavailability properties. However, some of the biggest challenges in drug discovery now require us to think beyond our current design space.
A major issue here is that many promising drug targets just don't favor small and lipophilic ligands. This is because the binding sites are often large, highly polar or featureless, so traditional Ro5 ligands won't bind with good affinity. Therefore, many drug developers are studying large ligands that are 'beyond the rule of 5' (bRo5).
It's unsurprisingly difficult to persuade bRo5 ligands to achieve acceptable oral pharmacokinetics. Indeed, drug attrition remains a significant problem and pharmaceutical companies invest heavily in unsuccessful candidates all too often. Given these difficulties, I wanted to know: do large targets always require bRo5 ligands, or could there be an alternative avenue to explore?
This is not a simple question to answer. However, it's now possible to harness the power of big data to reveal new insights, and so I wondered if this strategy could be applied to my problem. Here I describe how I used a new research platform and information tool to investigate this issue and gained valuable information that could shape the future of drug discovery.
How to modulate bRo5 targets: Could fragments be the answer?
Given the dramatic attrition rate of bRo5 ligands, I asked whether there could be an alternative approach to modulating large targets. In particular, I considered that fragments may provide an untapped drug resource here. Fragments have a low molecular weight (approximately 150-250) and are generating increasing interest in pharmaceutical R&D. Crucially, because these drugs are small, they don't tend to have the same problems of bRo5 agents in terms of poor oral bioavailability.
However, traditional drug discovery isn't really geared up to identify promising fragment candidates. This is because developers tend to focus on mechanistic screens, which assess affinity to a particular target, and fragments are often difficult to detect in these screens. Instead, they're more likely to be detected in phenotypic screens, where you're looking for drug effects.
Why is there such a difference between phenotypic and mechanistic screens in picking up fragments? In considering this question, I developed a novel hypothesis that fragments often show allosteric activity, interacting with the target away from the traditional binding site. This activity is frequently not identified in mechanistic screens, where you're looking for affinity to that particular site, but is better detected in phenotypic assays.
This really interests me as the answer could have significant implications for screening in drug discovery. However, addressing this question in the lab would be a difficult and time-consuming business. Therefore, I considered whether available scientific data could be used to gain some answers, and if so, how?
How a research platform became my lab partner for testing a novel hypothesis on big data
Thanks to recent developments in software, we can now begin to answer these complex questions using the huge resource of big data already available. The tool I selected was SciFindern, an advanced research platform based on CAS's content collection. I used this to conduct a series of filtered searches, focusing on anti-infective drug discovery as phenotypic assays dominate in this area. With these searches, I looked to determine whether allosteric activity is proportionally greater among low molecular weight anti-infectives as compared to CAS's entire biologically active chemistry database.
It was surprisingly easy to search and manipulate the data. Essentially, I was able to identify and compare two datasets of biologically active compounds associated with allosteric activity. The first of these was based on fragments (with a molecular weight of 150-250) that were also anti-infective agents and the second was based on the universe of known compounds (with a molecular weight of <5,000). When I compared the two datasets, the results were startling. The rate of allosteric mentions was 10-fold higher for the fragments.
To me, this is an incredibly thought-provoking result. It suggests that fragments are disproportionately likely to have allosteric effects identified on phenotypic screening. Since these effects aren't likely to be discovered in traditional mechanistic screening, we just don't know how common these small allosteric agents could be. Indeed, it's possible that they could provide a very promising solution to the problem of bRo5 targets.
In general, I'd say there are two key messages that I took home from this study. First, there's the exciting implication of the results: drug developers could potentially gain a lot of value from phenotypic fragment assays. Second, the remarkable message to draw from this study is the fact that it's now possible, and easy, to gain this kind of insight from big data. If we could gain other such insights in this way, we could save a lot of time and help advance drug discovery faster.
Accelerating research with advanced scientific information tools
In the process of my study, I learned a lot about the power of the latest research tools. In fact, SciFindern far exceeded my expectations of what I could achieve with a research and scientific information platform. Rather than being a simple search and retrieval tool, it became my 'lab partner' to help me address my hypothesis. In all my experience of working in drug discovery, I haven't found any other product able to do this.
With sophisticated scientific information tools from CAS combining advanced cheminformatics technology with comprehensive R&D data, we're gaining the ability to address novel medicinal chemistry hypotheses using the power of big data alone. In this way, we can develop valuable insights to optimize R&D efficiency, accelerating our progress in drug discovery.
See in detail how Dr. Lipinski used SciFindern to gain insights into drug discovery. Go beyond the rule of 5 and contact us to learn more about SciFindern, the platform that can become your lab partner and be the key to unlock your R&D productivity.