EVENTS | VIEW CALENDAR
An alternative view of drug-like properties
The terms "drug-like" and "drug-like properties" have gained common currency in the drug discovery community. They usually refer to the values of simple physicochemical and structural compound properties that successful drugs have in common. However, despite the frequency with which these terms are used, "drug-like" means different things to different people and will depend on a project's specific objective.
For example, one of the best-known rules is Lipinski's Rule of Five (RoF), which is based on four easily calculated properties:
However, it is important to note that this rule was defined for compounds absorbed through the human intestine, and it was never intended as a general definition of drug-likeness. The rules for compounds intended for other routes such as inhalation, topical or intravenous administration might be quite different.
Many other, similar rules have been proposed that define drug-like properties. For example, Veber, et al. (Journal of Medicinal Chemistry, 2002) found that the majority of compounds with good oral bioavailability in rats had less than 10 rotatable bonds (ROTB) and polar surface area (PSA) less than 140 ┼2. Others have explored different drug discovery objectives: Hughes, et al. (Bioorganic and Medicinal Chemistry, 2008) found that compounds with logP less than 3 and PSA greater than 75 ┼2 were six times less likely to exhibit adverse events in in-vivo tolerance studies than compounds failing to meet both of these criteria; Ritchie, et al. (Drug Discovery Today, 2009) looked at several "developability" requirements such as solubility, serum albumin binding and inhibition of the hERG ion channel and found relationships with the number of aromatic rings (AROM) in a compound, suggesting that AROM greater than 3 significantly increases the risk of compound attrition; and finally, Lovering, et al. (Journal of Medicinal Chemistry, 2009) related the "flatness" of compounds, as defined by the fraction of carbons that are sp3 hybridized (fSP3), to their success in clinical development.
These rules for drug-like properties are appealing because they are very simple to apply. The properties to which they relate are easily calculated, and it is easy to see when a compound meets the criteria in each case. They provide guidance regarding potential risk factors and indicate strategies for improvement, if an issue should be encountered.
However, due to the apparent simplicity of these rules, there may be a tendency to apply them as filters or hard cutoffs when selecting compounds. This is a risky approach because the simple compound properties on which the rules depend have only a weak correlation with the objectives to which they relate, such as oral bioavailability or toxicity.
For example, does a compound with a logP of 5.1 have a significantly lower chance of success than one with a logP of 4.9? By rigidly applying these criteria, a project team runs the risk of rejecting good compounds that represent potentially valuable opportunities. The RoF states that 80 percent of compounds with good oral absorption fail no more than one of the four criteria noted above, i.e., failing only one criterion should not be enough to reject a compound. Of course, this also means that 20 percent of orally absorbed drugs fail even this more relaxed definition. Similarly, the majority of non-oral drugs also meet the RoF. Therefore, the RoF helps to improve the odds of success when looking for an orally absorbed compound, but is far from a guarantee of success.
The dangers of hard filters are further exacerbated by the fact that some of the properties on which these rules are based have significant uncertainty; for example, a prediction from a good model of logP has an uncertainty of approximately 0.4 log units. Therefore, drawing conclusions based on differences of less than this range would be inappropriate.
Moving away from hard cutoffs
Instead of applying hard cutoffs, it would be better to rank compounds according to the similarity of their properties to the majority of drugs. In a recent paper, Bickerton, et al. (Nature Chemistry, 2012) proposed a new metric, the Quantitative Estimate of Drug-likeness, or QED. To derive this, the authors generated "desirability functions'" for eight properties commonly used to define drug-likeness: MW, logP, HBA, HBD, PSA, AROM, ROTB and ALERTS (the number of matches to undesirable functionalities). A desirability function maps the value of a property onto a scale between 0 and 1, where a desirability of 1 indicates an ideal value of the property and a desirability of 0 corresponds to a completely unacceptable outcome. The desirability functions used in QED were fitted to the distributions of the eight properties for a set of oral drugs, so that a desirability of 1 was assigned to the property values of oral drugs that occur most commonly, and 0 to property values that are not observed. The QED for a new compound can then be calculated from the desirability of the eight properties by taking their geometric mean, giving an overall value between 0 and 1 that indicates the similarity of the compound to the majority of oral drugs. Using this, compounds can be ranked according to their drug-likeness, instead of simply being labeled drug-like or non-drug-like.
The authors of the QED paper showed that the QED metric correlated well with the subjective opinions of medicinal chemists regarding the attractiveness of compounds for undertaking further chemistry. They also showed that the QED performed well in distinguishing oral drugs from other small-molecule ligands taken from the Protein Data Bank.
Drug-like or likely to be a drug?
QED, like many other definitions of drug-likeness, considers the properties that drugs have in common. The assumption is that a compound with similar properties to successful drugs will have a lower risk than compounds with one or more significantly different properties. This makes intuitive sense, because there is no precedence for the success of a compound with radically different properties. However, having a similar value of a property does not necessarily increase the chance of a compound being a successful drug; if the distribution of a property for successful drugs is the same as that for all compounds that have been explored, the property will not provide any information about the chance of success of a compound. In other words, we would like to identify the properties that make successful drugs different from other compounds and hence increase the chance of success.
To achieve this, the property distributions of successful drugs can be compared with those of unsuccessful compounds explored in the search for a drug. A branch of mathematics called Bayesian probability allows for rigorous comparison of the probability of a drug having a property value, with the probability of an unsuccessful compound having the same value, and hence estimate the relative likelihood of success of a compound with that property value.
The relative likelihood can identify the properties that are most important in distinguishing drugs from non-drugs. For example, if oral drugs are compared with other compounds explored in the course of drug discovery projects, then of the eight properties listed above for QED, the properties that best distinguish these sets of compounds are MW, PSA and ROTB. Low values of these properties can increase the likelihood of success by more than a factor of two. Conversely, this analysis indicates that HBA and HBD have a limited impact on the chance of success of a compound as an oral drug.
Furthermore, the relative likelihoods derived from the eight properties may be combined into a single metric by taking their geometric mean to calculate a new metric, the Relative Drug Likelihood (RDL), which can be used to rank compounds in a similar manner to QED. When trained to distinguish oral drugs, as described above, the RDL outperforms rules and metrics based only on similarity of properties.
Words of caution
These analyses of property trends across a wide range of chemistries can provide some general guidance on appropriate compound properties. However, the most relevant information will come from comparison of successful and unsuccessful compounds explored for a specific objective, such as a target or therapeutic class. The properties of compounds intended for use as an antibiotic provide little information about what makes a successful kinase inhibitor and may even be misleading. Therefore, when sufficient data is available, metrics such as QED or RDL should be "trained" to identify the specific requirements for a given project.
Finally, it must be emphasized that high similarity to successful drugs or relative likelihood is far from a guarantee of success, and in some cases, it may be necessary to explore "outside of the box," particularly for new target classes such as protein-protein interactions. All the evidence suggests that the absolute chance of any single compound becoming a successful drug is very low. Therefore, all of these rules and metrics should be used only as guidelines, and they should be given appropriate weight when making decisions. They are most useful early in a project when there are many compounds from which to choose. Once data on structure-activity relationships is available, this provides much better information to guide the design and selection of compounds. Combining these data to identify compounds with the optimal balance of potency, physicochemical, ADMET and safety characteristicsŚfor example, using a multiparameter optimization approach (Current Pharmaceutical Design, 2012), will enhance the chance of success much more than by considering simple drug-like properties.
Dr. Matthew Segall is the CEO of Optibrium Ltd., a software developer based in Cambridge, England. Segall's prior positions include the associate director of Amitro, ArQule Inc. and Inpharmatica and senior director of BioFocus DPI's ADMET division. He has an M.Sc. degree in computation from the University of Oxford and a Ph.D. in theoretical physics from University of Cambridge.