API
Solubility in Pharmaceutical R&D: Predictions and Reality
David P. Elder, PhD* and Christoph Saal, PhD†
*GlaxoSmithKline R&D, United Kingdom †
Merck KGaA, Germany
Elder DP, Saal C. Solubility in pharmaceutical R&D: predictions and reality. Pharm Outsourcing. 2015;16(2):36-41.
Note: This article originally appeared in the January/February 2014 issue of American Pharmaceutical Review.
Introduction
Lipinski’s seminal paper on experimental and computational (in silico) methods to estimate the solubility and permeability of drug candidates was published in 1997.1
The iconic “Rule of 5” predicted that absorption
was adversely impacted when the calculated LogP (cLogP) was >5, when molecular weight (MW) was >500, when there are >5 H-bond donors or >10 H-bond acceptors. The related concept of “drug-likeness” importantly focuses on both potency and physicochemical attributes, using tools such as lipophilic efficiency2
or ligand efficiency.3
Drug-likeness and related concepts have been widely used in the pharmaceutical industry to attempt to reduce the very high attrition rates currently seen with unprecedented pharmacological targets. Unfortunately, both combinatorial and high-throughput chemistry tends to favor leads with higher MW, higher cLogP and lowered solubility.1
Therefore, successful drug discovery strategies appear to be a balance between trying to optimize both “hydrophobicity-driven potency and hydrophilicity-driven biopharmaceutics properties.”4,5
Consequently,
an over-reliance on optimizing potency to the detriment of physicochemical properties will yield sub-optimal ADMET (absorption, distribution, metabolism, excretion, and toxicity) properties and reduce the likelihood of clinical success.6
However, the lack of optimal
physicochemical attributes can often be offset using sophisticated formulation strategies,7,8
and deficiencies in these properties can
often be rate limiting to the progression of drug candidates. Hence, computational methods that can qualitatively predict physicochemical properties, eg, solubility, before a compound is synthesized based on molecular structure, are an essential requirement for drug discovery.
In this article, the authors will provide an overview of approaches to predict and measure solubility in the pharmaceutical R&D environment with the goal of providing relevant information at the appropriate stage of the process.
Relationships between Solubility and
Physicochemical Properties There are significant numbers of computational methods reported in the literature for predicting solubility from underlying molecular properties.
These include electronic and topological evaluations, hydrophilicity/ hydrophobicity assessments, molecular surface area calculations, etc.9 In addition, other researchers have attempted to predict solubility using fragment-based models,10
being computational models utilizing molecular topology,11 contribution approaches,12
and E-state indices.13
with the 3 most favored approaches group
However, these
computational methods need to be able to cope with significant numbers of compounds and filter out ‘non-drug-like’ compounds. The methods must help to focus chemistry initiatives on programs with improved physicochemical attributes, thereby enhancing productivity. Importantly, it should be clearly understood that early discovery methodologies provide qualitative and not quantitative outcomes.14
The challenges inherent in solubility prediction were graphically highlighted by the recent Solubility Challenge. An academic research group15
measured the equilibrium solubilities of 100 drug-like
molecules at a fixed temperature and ionic strength. Using this training dataset, they requested other research groups to predict, using their own preferred computational approach(es), the intrinsic solubilities of an additional 32 drug-like molecules. The training set was selected to represent a wide range of chemical space with MW ranging from 115 (proline) to 645 (amiodarone), that had pKas between 1 and 12. The intrinsic solubilities spanned over 7-orders of magnitude from the poorly soluble (amiodarone) to the highly soluble (acetaminophen), with a relatively even spread of intermediate values.
The authors received 100+ entries to the Solubility Challenge.16 Participants used the entire spectrum of available computational tools, and this challenge therefore provided a holistic overview of our ability to predict aqueous solubility. The authors could not recommend the best approach(es), rather a number of methodologies that were equally successful at predicting aqueous solubility were identified. Some participants were surprised that the simple models were superior to the more complex methodologies.17
Faller and Ertl10 went further, claiming
that the advantages of the complex models were debatable when compared with the simple cLogP correlations.18
Hewitt et al17 indicated
that data quality was pivotal to successful predictivity and even with the “high quality” dataset provided in the “Challenge,” questions were still raised about data quality, and it is critical to understand the applicability domain (ie, the chemical space where the model works best). Understandably, predictions made outside of this domain will
Pharmaceutical Outsourcing | 36 | March/April 2015
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78