Principles of Drug Design: How to Leverage Ligand-Receptor Molecular Interactions

Maricel Torrent- Principal Research Scientist, Computational Drug Discovery, Discovery Research, AbbVie

Abstract

Ligands bind to receptors by a collection of molecular forces. Experienced drug designers have learned how to engineer the right type of molecular interactions in every given situation. Each type of interaction is governed by a different set of principles, including distance and geometry requirements. The ultimate goal is to maximize the contact with the intended receptor while minimizing binding to non-intended receptors. Not a single type of interaction can accomplish this goal but rather a balanced mix of multiple types of interactions. Hydrophobic and polar interactions are the most common types, used early in the drug design process to rapidly increase affinity for the target. Specialized interactions offer smaller energy gains. They are introduced later into more advanced leads to achieve proper multiparametric optimization and ensure developability down the road.

Introduction

Based on the fact that about 60% of the human body is water, any molecule entering the human body, including drugs, is expected to interact with water, at least some of the time. Water can help with transportation of a drug molecule, necessary to reach the intended receptor or target in the body. Non-intended receptors can be distracting, and they might prevent drug molecules from getting to their desired, final destination, or at least cause these drug molecules to become partially stuck or lost along the way. Recent studies suggest that every drug hits, on average, 11.5 known targets,¹ almost twice the amount reported about a decade earlier (6 known targets).² The drug designer’s job is to make the interactions of the drug with the desired receptor(s) stronger/better than any other interaction, whether with water or with other receptors.³ Drug designers must have, therefore, a working knowledge of the variety of molecular interactions that are possible. They also want to learn the principles that govern these interactions and how to apply these principles in every situation.

Early Metaphor for Ligand-Receptor Interactions

The first reasonably accurate illustration of ligand-receptor interaction was given in 1726 by author Jonathan Swift in his book Gulliver’s Travels.⁴ The hero in the novel, Lemuel Gulliver, ends up caught by the citizens of Lilliput and ultimately bound by many ropes (individually weak), but which collectively were able to constrain him to the ground. Likewise, ligands bind to receptors by a collection of molecular forces, each individually equal to a few (1-2) log units of energy and each easily broken. It is when these molecular forces are combined that they produce the nanomolar affinity (8-10 log units) drugs typically exhibit for their receptors.

Types of Molecular Interactions

In the early stages of drug discovery campaigns, most hits are commonly in the low millimolar to high micromolar range. As hits get further optimized and become leads, their affinity for the target of interest usually reaches the nanomolar range, and, in some select scenarios, even the picomolar range.

Table 1 lists some of the most common molecular interactions in drug design, together with a fold-change value. Fold change here represents a typical change in IC50 value when a particular interaction is achieved. Higher change than listed is rare, but possible. In practice, however, a not uncommon outcome tends to be simply no change. Many planned interactions, once implemented, result in zero gain in affinity due to a combination of countless approximations in paper and of suboptimal execution.

Hydrophobic Interactions

Some types of interactions are notably easier to engineer than others. Take, for instance, hydrophobic interactions. Lipophilic interactions are relatively straightforward to successfully implement, in part because they are governed solely by a distance requirement. As long as the distance between the two closest heavy atoms in the ligand and the receptor remain within the 3.5-5.0 Å range, chances are that the two entities will feel each other’s presence, and produce an effective interaction, translated empirically into a favorable IC50 or Kd. Because lipophilic interactions are so easy to implement, even inexperienced drug hunters can quickly advance their hits to greater levels of potency by simply adding more greasy building blocks (typically rings) to their starting scaffolds. Unfortunately, high levels of hydrophobicity in small molecules can lead to a large number of problems down the (drug development) road. Greasy molecules have high logP, which indicates the preference of a compound to dissolve better in an organic solvent such as octanol than in water when uncharged. High logP is often associated with high clearance, high volume of distribution, high receptor promiscuity, high hERG activity, low solubility, and high phospholipidosis as well as other undesirable attributes and toxicity parameters.^5-8

Newer generations of drug designers and medicinal chemists continue to be advised to treat compounds with logP > 4 with caution. Over the last several decades, medicinal chemists synthesized compounds that relied too heavily on hydrophobicity to drive potency, which led to significant challenges in drug developability.⁹ One way for medicinal chemists to attenuate this trend is to improve the balance of hydrophobic and polar interactions in newly synthesized compounds.

Polar Interactions and Hydrogen-Bonding Interactions

Polar interactions are necessary to rebalance solubility and also to reduce promiscuity. Engineering polar interactions into a molecule in a productive way requires some mastery. Unlike hydrophobic interactions, hydrogen-bonding interactions (the most popular type of polar interactions) are governed by, not just one, but two requirements: a distance and a geometry requirement. First, the distance between the two closest heavy atoms in the ligand and the receptor has to be within the optimal range of 2.8-3.2 Å (or at least within the acceptable range of 2.5-3.5 Å) for a hydrogen-bond interaction to be effective. In addition, hydrogen-bonding interactions do also have to fulfill a geometry requirement, namely, that the bond angle between the hydrogen-bond acceptor atom, the hydrogen atom, and the hydrogen-bond donor atom is as close as possible to an ideal 180-degree angle. A 30° deviation from linearity still retains good binding energy. A 50° deviation from linearity diminishes strength of the binding energy. A 90° deviation gives almost complete loss of binding energy, due to drastic reduction of atomic orbital overlap. Because hydrogen bonds have strict distance and angle requirements, they are not very forgiving of design shortcomings. With this type of molecular interactions, there is limited room to maneuver beyond ideal parameters. Paying great attention to detail and accuracy in the design of hydrogen-bonding interactions and moving forward only with designs that strictly stay within the optimal ranges can improve the outcomes.

Risk in Designing Hydrogen-Bonding Interactions

The harsh reality in drug design is that even hydrogen-bonds designed to be geometrically ideal frequently fail to give any gain in affinity. The root of this problem is the lack of awareness and inclusion of what already exists prior to the formation of the ligand-receptor complex. The design of hydrogen bonds needs to be considered as an exchange of hydrogen bonds with water, rather than a gain of hydrogen-bonds. Thus, both the ligand and the receptor are likely to be each surrounded by many water molecules prior to encountering their partner. All (or at least some of) these water molecules will need to be displaced and replaced with the partner. For the design experiment to be successful, the newly formed interactions will have to exceed in number or then be stronger in total than the sum of the original interactions.

Figure 1 illustrates a common scenario where the outcome (namely, no net gain in number of interactions made) does not match the initial expectation (namely, assumption that one net interaction will be made in total, which is false). Panel A suggests the net formation of one hydrogen-bond interaction. However, as shown in Panel B, the reality is that there are the same number of hydrogen-bond interactions (two) in both the reactants and in the products. Panel C illustrates a real-life scenario in drug design where an opportunity might be (mistakenly) identified to create a stronger ligand-receptor interaction swapping a pyridine group in the ligand with an N-oxide pyridine moiety. Panel D reveals why instead of an affinity gain, this particular design concept leads to an affinity loss. Pyridine N-oxides are such strong H-bond acceptors that the solvated state typically has, not just one but, two bound waters. Since no waters are capable of fitting into a tight active site, there is a net loss of one hydrogen bond interaction.

Electrostatic Interactions

As seen from Table 1, the type of interaction that has the ability to produce the largest gain in affinity is the electrostatic interaction. These interactions are present whenever a positively charged group in the ligand (or in the protein) falls in the vicinity of a negatively charged group in the protein (or in the ligand). Common positively charged groups include amines and other N-containing groups (lysine, arginine, in proteins), amongst others. Common negatively charged groups include carboxylates and other O-containing groups (glutamate, aspartate, in proteins). Similar to hydrophobic interactions, electrostatic interactions only need the distance requirement to be satisfied for them to be productive. Given the high potential gain and the relatively low eff ort to introduce them, it is tempting to engineer more electrostatic interactions in small molecules than can actually be accommodated. Charged molecules present a number of challenges later in the pipeline, especially in terms of developability. Designed ligand-receptor electrostatic interactions typically lead to molecules with diminished bioavailability or other pharmacological issues. Carboxylates and other anions, for example, can exhibit albumin binding. Cations can exhibit hERG binding. Both anions and cations can drastically reduce cell permeability. Thus, in the transport across a phospholipid bilayer by passive diffusion, the permeability of the neutral form of a molecule is about 108 greater than that of the charged form.¹⁰ Also, much less energy is required to desolvate neutral species than charged species. To reduce most of these risks, drug designers and medicinal chemists are advised to investigate isoelectronic/ isosteric replacements. For instance, neutral groups such as sulfones make excellent carboxylate replacements, and amines can be often replaced with alcohols.

Other Interactions in the Drug Designer’s Toolbox

Hydrophobic interactions and hydrogen-bonding interactions (covered in the previous sections) are the most frequent type of interactions in drug design, and, therefore, addressed first in this short review. These two types of molecular interactions are also the top two most common non-covalent interactions observed in protein-ligand complexes extracted from the Protein Data Bank.¹¹ More specialized interactions are needed in the designer’s toolbox to access the full spectrum of scenarios encountered when optimizing drug molecules. More specialized interactions include π-π interactions, cation-π interactions, halogen bonding interactions, and weak hydrogen bonding interactions, amongst others. As seen from Table 1, the affinity gain that can be reached when these more specialized interactions are implemented is very often inferior to the gains produced with the top two most frequent types of interactions. Specialized interactions tend to be introduced later in lead optimization campaigns. They are also used more sparingly due to their unique nature. For instance, introduction of chlorine or bromine atoms on a molecule quickly increases the molecular weight, reducing the binding efficiency. Therefore, addition of heavier halogen atoms is very purpose-specific and often applied only when other ways to accomplish the desired result have failed. Fluorine is an exception. Due to its smaller weight, fluorine atoms tend to be added to molecules more generously, especially to block a metabolic hotspot or to alter the basicity/acidity of a neighboring group. Only 6.5% of the drugs approved in the 1960s contained fluorine; by the mid-2010s, this number had increased to about 20% of the approved drugs,¹² indicating that profitable fluorinated molecules with good efficiency/safety ratios led to a higher tolerance for mono- and polyfluorinated designs over time.

Final Thoughts

The best drug designs, as displayed in the majority of approved drugs to date, are those that embody a diverse set of molecular interactions, each present in the molecule for a very specific reason. Not a single type of molecular interaction will suffice in any scenario; instead, a balanced mix of multiple types of interactions are required. As molecules grow and evolve along the drug discovery path, drug designers are advised to question at each step whether the new addition or change made will translate into a new or improved interaction that ultimately benefits the final deliverable. Along the drug optimization race, every atom in the molecule needs to be periodically scrutinized and its value assessed. Elements in the design that do not directly contribute to improving the interaction with the desired target offer anchoring points for other, non-intended receptors to redirect the drug elsewhere.

References

1. Peon A, Naulaerts S, Ballester PJ. Sci Rep. 2017;7(1):3820-3830.
2. Mestres J, Gregori-Puigjane E, Valverde S, Sole RV. Nat. Biotechnol. 2008;26(9):983-984.
3. Mensa-Wilmot K. Mol. Pharm. 2021;100(1):1-6.
4. Gulliver’s Travels. Encyclopedia Britannica information page. Available at: https://www. britannica.com/topic/Gullivers-Travels. Accessed February 25, 2023.
5. Lombardo F, Berellini G, Obach RS. Drug Metab. Dispos. 2018;46(11):1466-1477.
6. Leeson PD, Bento AP, Gaulton A, et al. J. Med. Chem. 2021;64(11):7210-7230.
7. Walters WP, Green J, Weiss JR, Murcko MA. J. Med. Chem. 2011;54(19):6405-6416.
8. Waring MJ. Expert Opin. Drug Discov. 2010;5(3):235-248.
9. Hann MM. Med. Chem. Comm. 2011;2(5):349-355.
10. Avdeef A. Curr. Top. Med. Chem. 2001;1(4):277-351.
11. Ferreira de Freitas R, Schapira M. Med. Chem. Comm. 2017;8(10):1970-1981  
12. Ilardi EA, Vitaku E, Njardarson JT. J. Med. Chem. 2014;57(7):2832-2842.

Author Biography

Maricel Torrent is a Molecular Modeler with 20+ years of experience in Drug Discovery. Primary strengths in Computer-Aided Drug Design, both structure-based and ligand-based, as well as data mining and analysis. Inspiring leader, sought mentor, coach, and team player. Creative co-inventor with proven results across various top 10 pharmaceutical companies, including Merck, Abbott, and AbbVie. International speaker. Presenter at numerous scientific conferences across the globe. Co-author of more than 80 peer-reviewed scientific articles; two book chapters.

Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special
offers from American Pharmaceutical Review delivered to your inbox!
Sign up now!