Exploring Rational Drug Design

Medicinal chemists strive to optimize molecules that fit snugly into their proposed targets. But in the quest for potency, we often overlook the local physics that govern drugs’ binding to these receptors. What if we could rationally predict which drugs bind well to their targets?

A new review, currently out on J. Med. Chem. ASAP, lays out all the computational backing behind this venture. Three computational chemists (David Huggins, Woody Sherman, and Bruce Tidor) break down five binding events from the point-of-view of the drug target: Shape Complementarity, Electrostatics, Protein Flexibility, Explicit Water Displacement, and Allosteric Modulation….whew!

Selectivity Strategies

Selectivity Strategies for Rational Design | Credit: Huggins, Sherman, Tidor; J. Med. Chem.

Note: Before we dive into this article, let’s clarify a few terms computational drug-hunters use that bench chemists think of differently: ‘decoy’ – a test receptor used to perform virtual screens; ‘ligand’ – the drug docking into the protein; ‘affinity / selectivity’ – a balance of characteristics, or how tightly something binds vs. which proteins it binds to; ‘allosteric’ – binding of a drug molecule to a different site on an enzyme than the normal active site. Regular readers and fans of compu-centric chem blogs such as The Curious Wavefunction and Practical Fragments will feel right at home!

We’ll start at the top. Shape complementarity modeling uses small differences in a binding pocket, such as a methylene spacer in a residue (say, from a Val to Ile swap) to dial-in tighter binding between a target and its decoy. The authors point out that selectivity can often be enhanced by considering a drug that’s literally too big to fit into a related enzymatic cavity. They provide several other examples with a ROCK-1 or MAP kinase flavor, and consider software packages designed to dock drugs into the “biologically active” conformation of the protein.

Electrostatic considerations use polar surface maps, the “reds” and “blues” of a receptor’s electronic distribution, to show how

Affinity Optimization

Affinity Optimization - Black dot represents the optimal minimum energy between Coulombic forces (green) and desolvation penalty (blue) | Credit: Huggins, Sherman, Tidor; J. Med. Chem.

molecular contacts can help binding to overcome the desolvation penalty (the energy cost involved in moving water out and the drug molecule in). An extension of this basic tactic, charge optimization screening, can be used to test whole panels of drugs against dummy receptors to determine how mutations might influence drug binding.

Because target proteins move and shift constantly, protein flexibility, the ability of the protein to adapt to a binding event, is another factor worth considering. The authors point out that many kinases possess a “DFG loop” region that can shift and move to reveal a deeper binding cavity in the kinase, which can help when designing binders (for a collection of several receptors with notoriously shifty binding pockets – sialidase, MMPs, cholinesterase – see p. 534 of Teague’s NRDD review).

But these shifting proteins also swim in a sea of water and other cytoplasmic goodies. This means that drug designers, whether they like it or not, must account for explicit water molecules. The authors even suggest a sort of “on-off” switch for including the bound water molecules, but contend that more efforts should be directed to accurate modeling of water in these protein settings.

Finally, the authors weigh the effects of allosteric binding, the potential for a modeled molecule to be highly selective for a site apart from where the protein binds its native ligand. The authors consider the case of a PTP1B ligand that binds 20Å away from the normal active site, at the previously mentioned “DFG loop.” Since this binding hadn’t been seen for related phosphatases, it could then be used to control selectivity for PTP1B.

In each section, the authors provide examples of modeling studies that led to the design of a molecule. Two target classes recur oftenCOX and HIV inhibitors throughout the review: HIV protease inhibitors (saquinavir, lopinavir, darunavir) and COX-2 inhibitors (celecoxib), which have all been extensively modeled.

Two higher-level modeling problems are also introduced: the substrate-envelope hypothesis, which deals with rapidly mutating targets, and tailoring molecules to take rides in and out of the cell using influx and efflux pumps in the membrane. Since different cell types overexpress certain receptors, we can use this feature to our advantage. This strategy has been especially successful in the development of several cancer and CNS drugs.

Overall, the review feels quite thorough, though I suspect regular Haystack readers may experience the same learning curve I did when adapting to the field-specific language that permeates each section. Since pictures are worth a thousand words, I found that glancing through the docking graphics that accompany each section helped me gain a crucial foothold into the text.

Author: See Arr Oh

See Arr Oh is a Ph.D. chemist working in industry

Share This Post On


  1. This doesn’t make it clear that the decoy is what you’re trying NOT to bind to; the object is to selectively bind to the target (producing the therapeutic effect) without binding to any decoys (producing presumably undesirable side effects).

  2. The first graphic seems perfectly clear. There is a big red “X” between the drug and the decoy, indicating you do not want to go down that path.