Medicinal Chemistry at the Bedside
How the molecular properties in the Chemical Structures tool — lipophilicity, polar surface area, ionization, and protein binding — translate into dosing, drug interactions, and side effects
The Chemical Structures comparison tool reports a short list of numbers for every drug — molecular weight, cLogP, polar surface area, ionization at physiologic pH, aqueous solubility, and a highlighted protein-binding region. These are not trivia. Each property is a lever that a medicinal chemist deliberately tuned, and each one predicts something you will see clinically: how fast a drug enters the brain, how long it lingers, whether renal or hepatic impairment matters, how tightly it rides on albumin, and where it is likely to interact. This chapter walks through the panel one property at a time, connects each to bedside decisions, adds the druglikeness range bars and the transporters that carry drugs across membranes, and closes with how these levers are used on purpose during drug design.
The property panel, decoded
Everything in the tool's property rows maps onto one of three questions: Can the molecule get where it needs to go? How long does it stay? And what will it collide with along the way? The table below is the quick-reference version; the sections that follow expand the entries that matter most.
| Property (as shown) | What it measures | Why it matters clinically |
|---|---|---|
| Molecular weight | Size of the molecule in daltons | Small molecules (<500 Da) cross membranes and the blood–brain barrier passively; very large molecules need transporters or won't be oral. One of Lipinski's four cutoffs. |
| cLogP (lipophilicity) | Calculated fat-vs-water partitioning of the neutral molecule | Drives CNS penetration, volume of distribution, protein binding, and CYP metabolism. Too high predicts sedation, accumulation in fat, and off-target promiscuity. |
| Polar surface area (TPSA) | Total area of polar (O, N, and attached H) atoms | A gate on membrane and blood–brain-barrier crossing. TPSA > ~90 Ų sharply reduces brain entry; > 140 Ų limits oral absorption. |
| Ionization (pH 7.4) | Fraction charged in plasma, set by pKa | Only the neutral fraction crosses membranes freely. Governs pH-dependent absorption, ion trapping, and dialyzability. |
| Aqueous solubility | How much dissolves in water (mg/mL, logS) | Poorly soluble drugs have erratic or food-dependent absorption and are hard to formulate; it is the other half of the permeability/solubility trade-off. |
| Protein-binding region | Structural feature that anchors the drug to plasma proteins | Highly bound drugs have a small free (active) fraction, are displaceable, and are poorly dialyzed. See the low-albumin scenario below. |
| H-bond donors / acceptors, rotatable bonds | Count of N–H/O–H groups, of N and O atoms, and of freely rotating single bonds | The remaining Lipinski and Veber cutoffs (donors ≤ 5, acceptors ≤ 10, rotatable bonds ≤ 10). Shown as range bars; each hydrogen bond and each degree of flexibility is a tax on membrane crossing. |
| Membrane transport | Whether a carrier or efflux pump — not passive diffusion — governs where the drug goes | Flagged when a drug is a notable P-glycoprotein (ABCB1) substrate. Efflux at the blood–brain barrier can keep an otherwise brain-ready molecule peripheral. |
Lipophilicity: logP, cLogP, and the rule of five
logP is the base-10 logarithm of a drug's partition coefficient — the ratio of its concentration in octanol (a fat surrogate) to water at equilibrium, measured for the neutral form. A logP of 0 means the molecule splits evenly; +3 means it favors fat 1,000-to-1; a negative value means it prefers water. cLogP is the same quantity calculated from the structure rather than measured in a flask, by summing empirically derived fragment and atom contributions (the tool reports cLogP because it is generated directly from each SMILES string). The two usually agree within a few tenths; they diverge for unusual scaffolds, intramolecular hydrogen bonds, or zwitterions, where the fragment method has no good rule. When they disagree meaningfully, the measured logP is the reference standard and cLogP is the estimate.
Lipophilicity is the single most predictive property because it correlates with so many downstream behaviors: higher logP means faster passive membrane crossing, deeper CNS penetration, larger volume of distribution, tighter plasma-protein binding, and more avid CYP450 metabolism — but also more sedation, more fat accumulation with a longer effective half-life, more off-target receptor promiscuity, and higher risk of poor aqueous solubility. Good CNS drugs live in a moderate window, roughly logP 2–4.
Lipinski's Rule of Five
A 1997 rule of thumb from Christopher Lipinski for predicting whether an orally administered small molecule will be reasonably absorbed. Poor oral absorption becomes likely when two or more of the following are true (each threshold is a multiple of five, hence the name):
• Molecular weight > 500 Da
• cLogP > 5
• Hydrogen-bond donors > 5 (count of N–H and O–H)
• Hydrogen-bond acceptors > 10 (count of N and O)
How to use it: it is a filter, not a law. It predicts passive oral absorption of the neutral molecule; it does not apply to compounds that are actively transported, injectables, or naturally large classes (many antibiotics and all biologics break it intentionally). TPSA (< 140 Ų) and rotatable-bond count are common modern add-ons. For CNS drugs, the working targets are tighter still: MW under ~450, TPSA under ~90 Ų, and enough lipophilicity to cross but not so much that the drug is sedating or promiscuous.
Druglikeness at a glance: the range bars
The Chemical Structures tool now renders a compact set of range bars beneath the property rows — one track each for molecular weight, cLogP, polar surface area, hydrogen-bond donors, hydrogen-bond acceptors, and rotatable bonds — with a marker showing where the molecule falls against the accepted thresholds. Green is the favorable zone, amber is borderline, red is over the limit. Below the bars, three chips summarize the verdict: whether the molecule passes Lipinski's rule of five, whether it clears the Veber oral-bioavailability criteria, and whether its profile predicts good, limited, or poor entry into the brain.
The point of the bars is not to pass or fail a drug but to make the trade-offs visible. A molecule can sail through Lipinski and still be a poor CNS agent because its polar surface area is too high; another can violate a weight cutoff yet work fine because it is actively transported or given by injection. Two criteria round out the classic rule of five:
Veber's rules — the flexibility and polarity add-ons
Lipinski predicts absorption from size, lipophilicity, and hydrogen bonding, but says nothing about molecular flexibility. In 2002, Veber and colleagues at GSK found that oral bioavailability also tracks two further descriptors: rotatable bonds ≤ 10 (a floppy molecule pays a larger entropic penalty to bind and to cross membranes) and polar surface area ≤ 140 Ų. A drug that clears both Lipinski and Veber is very likely to be orally absorbable. The tool's "Veber" chip reflects that rotatable-bond and TPSA pair.
Hydrogen-bond donors and acceptors, concretely
Every hydrogen bond a drug makes with water is one it must break to slip through the greasy interior of a membrane. Donors (N–H, O–H) cost the most, which is why Lipinski caps them at five and CNS drugs usually keep them at three or fewer. Acceptors (any N or O) are capped at ten. You can read these almost straight off the 2D structure: count the blue nitrogens and red oxygens for acceptors, and the ones carrying an H for donors. Most psychotropics sit comfortably inside both limits — part of why they can be dosed orally and still reach the brain.
Polar surface area: the CNS gate — and staying out on purpose
Polar surface area deserves its own section because it is the property that most cleanly separates drugs that reach the brain from those that don't. As a rule of thumb, a TPSA under about 90 Ų is compatible with good CNS penetration; 90–140 Ų still allows oral absorption but limits brain entry; and above 140 Ų passive intestinal absorption falls off. Most psychotropics are engineered well under 90 — aripiprazole, sertraline, and buspirone all sit low — which is exactly why they are centrally active.
The more instructive examples are the drugs pushed to high PSA on purpose, to keep them out of the brain. Peripheral restriction is a deliberate design strategy, and three pairs make it vivid:
Sedating vs non-sedating antihistamines
First-generation diphenhydramine has a low polar surface area and slips easily across the blood–brain barrier — hence the sedation, and its use as a sleep aid. The second-generation agents fexofenadine and loratadine were designed with higher polarity (and, for fexofenadine, P-glycoprotein efflux) so they stay peripheral, blocking histamine in the airway and skin without the drowsiness. Same target, opposite CNS strategy — largely written in the polar surface area.
Domperidone vs metoclopramide
Both are D₂ antagonists used for nausea and gastroparesis. Metoclopramide crosses into the brain and can cause the full range of central D₂ effects — sedation, extrapyramidal symptoms, tardive dyskinesia. Domperidone, with higher polar surface area and strong P-glycoprotein efflux, largely stays outside the blood–brain barrier, so it causes far fewer central effects (at the cost of a QT liability from its peripheral action). The clinical difference is mostly a difference in membrane crossing.
Loperamide — the opioid you can buy off the shelf
Loperamide is a potent µ-opioid agonist, yet at normal doses it produces no euphoria or analgesia because a high polar surface area plus avid P-glycoprotein efflux keep it out of the CNS — it acts only on gut opioid receptors to slow motility. This is also why massive overdoses (taken to defeat the efflux pump or to self-treat withdrawal) are dangerous: once P-glycoprotein is overwhelmed, the drug reaches the heart and brain, causing life-threatening QT prolongation and arrhythmia.
Ionization and pKa: only the neutral form travels
The pKa is the pH at which a drug is half ionized. Combined with the local pH (via the Henderson–Hasselbalch relationship), it fixes the charged fraction the tool shows in the "Ionization (pH 7.4)" row. This matters because biological membranes are lipid, and only the uncharged form diffuses across them freely. Most psychotropics are weak bases (the amine nitrogen you can see highlighted in the 2D structure), so they are substantially protonated and positively charged in plasma — a brake on how much drug is free to cross into the brain at any instant.
Three practical consequences follow. First, pH-dependent absorption: raising gastric pH with a proton-pump inhibitor or antacid changes the ionized fraction of pH-sensitive drugs and can blunt absorption. Second, ion trapping: a drug that crosses a membrane as the neutral form and then becomes charged on the far side (because the pH differs) gets stuck there — the basis of urinary alkalinization to speed elimination of acidic overdoses, and of drug accumulation in acidic compartments. Third, dialyzability: small, water-soluble, minimally protein-bound, low-volume-of-distribution drugs (lithium is the classic) are efficiently removed by hemodialysis, whereas large, lipophilic, highly bound ones are not.
Because the concept is easier to hold onto through cases than through equations, three clinical situations show ionization actually changing a decision:
1 · pH-dependent absorption: PPIs and weak-base drugs
Some drugs need an acidic stomach to dissolve and be absorbed. Raising gastric pH with a proton-pump inhibitor or antacid leaves them undissolved and drops their absorption — a true efficacy-threatening interaction rather than a nuisance. The classic offenders are weak-base drugs with pH-dependent solubility: the antifungal ketoconazole, the HIV drug atazanavir, and several oncology tyrosine-kinase inhibitors (dasatinib, erlotinib). In psychiatry the effect is usually milder, but it is a reminder to ask why a co-prescribed acid suppressant might be quietly undercutting another drug.
2 · Ion trapping in overdose: why we alkalinize the urine
Salicylate (aspirin) and phenobarbital are weak acids. In the renal tubule, raising urine pH with a bicarbonate infusion converts more of the filtered drug to its charged, anionic form — which cannot diffuse back across the tubular membrane and is therefore "trapped" in the urine and excreted faster. Urinary alkalinization is a front-line technique in salicylate and phenobarbital poisoning, and it is pure Henderson–Hasselbalch: shift the pH, shift the ionized fraction, shift the elimination.
3 · Why the local anesthetic fails in an abscess
Lidocaine is a weak base (pKa ≈ 7.9). It must cross the nerve membrane in its neutral form to work, then re-ionize inside to block the sodium channel. Infected, inflamed tissue is acidic, which pushes lidocaine toward its charged form outside the nerve — so less crosses and the block is weak or fails. It is the everyday clinical face of ionization: the same drug, the same dose, defeated by a two-unit shift in local pH.
Protein binding — and the low-albumin problem
In plasma, a drug exists in equilibrium between a fraction bound to proteins (mainly albumin for acidic and neutral drugs; α1-acid glycoprotein for many bases) and a free fraction. Only the free (unbound) fraction is pharmacologically active — it is the part that can cross membranes, hit receptors, be metabolized, and be filtered by the kidney. The "protein-binding region" the tool highlights is the lipophilic/anionic patch of the molecule that docks into albumin. Highly bound drugs (say, > 90%) carry a large reservoir on protein and a correspondingly tiny free fraction.
The scenario: a highly protein-bound drug in a patient with low albumin
The question: what happens to a drug that is 95%+ protein-bound when the patient is hypoalbuminemic — malnourished, cirrhotic, nephrotic, burned, critically ill, or elderly?
The mechanism: fewer binding sites means a larger fraction of the drug is unbound. For a drug that was 95% bound, dropping to 90% bound doubles the free fraction from 5% to 10%. Because only free drug is active, the pharmacologic effect (and toxicity risk) rises even though the total measured concentration is unchanged.
The catch that traps clinicians: most laboratory assays report total drug level (bound + free). In hypoalbuminemia, the total level can sit squarely "in range" while the free, active level is high enough to cause toxicity. Phenytoin is the textbook example — a normal total phenytoin in a hypoalbuminemic patient can mask a toxic free level, which is why you correct the reported level (Sheiner–Tozer equation) or, better, order a free phenytoin directly. Valproate behaves similarly, and its binding is also saturable, so free fraction climbs at higher doses too.
The nuance: for most drugs the body compensates. A larger free fraction is also more available for clearance and distribution, so at steady state the free concentration often returns toward baseline while the total falls. The transient toxicity window matters most for drugs that are highly bound, have a narrow therapeutic index, and are given IV or titrated fast. Displacement drug interactions (e.g., adding a second highly bound drug) follow the same logic and are clinically important mainly for that same narrow-index, highly bound subset — not as a general rule.
What makes a drug favor renal excretion?
Whether a drug leaves mostly through the kidney (unchanged in urine) or through the liver (metabolized first) is largely predictable from the same properties on the panel. Renal excretion is favored by the mirror image of the features that favor hepatic metabolism.
| Favors renal excretion (unchanged) | Favors hepatic metabolism |
|---|---|
| Low lipophilicity / hydrophilic (low logP, high TPSA) | Lipophilic (high logP, low TPSA) |
| Small molecular weight | Larger, more complex structures |
| Charged / highly ionized at physiologic pH | Neutral or weakly ionized |
| Low plasma protein binding (free to be filtered) | High protein binding |
| Water-soluble; small volume of distribution | Poorly water-soluble; large volume of distribution |
The logic is mechanical. The glomerulus filters small, unbound, water-soluble molecules directly into urine; because the filtrate is water, a hydrophilic charged drug that can't diffuse back across the tubular membrane stays in the urine and is excreted. A lipophilic drug, by contrast, is largely protein-bound (not filtered), and any that is filtered simply diffuses back into the blood across the tubule — so the body must first make it water-soluble through hepatic phase I (CYP oxidation) and phase II (glucuronidation, etc.) metabolism before it can leave. Active tubular secretion (via OAT/OCT transporters) adds a second renal route for some charged drugs and is where interactions like probenecid act.
When a water-soluble drug still gets metabolized
The renal-versus-hepatic table above is a reliable first pass, and its core heuristic — hydrophilic drugs tend to be excreted unchanged, lipophilic drugs must be metabolized first — holds for most of the formulary. Truly water-soluble, low-logP psychotropics do bypass the cytochrome P450 system and leave through the kidney essentially intact: lithium, gabapentin, pregabalin, topiramate (largely), and metformin outside psychiatry are the clean examples. But two caveats keep the rule honest.
First, "not metabolized by CYP" is not the same as "not metabolized." Polar drugs frequently skip phase I oxidation only to be conjugated in phase II — glucuronidation, sulfation, acetylation. Morphine is the archetype: relatively hydrophilic, barely touched by CYP, yet almost entirely cleared by UGT glucuronidation (one product, morphine-6-glucuronide, is itself active). Lamotrigine, lorazepam, and oxazepam are psychiatric examples cleared chiefly by glucuronidation rather than CYP — which is why they are comparatively free of CYP-based interactions.
Second, there are genuine outliers where a water-soluble drug is a CYP substrate anyway, because CYP recognizes a specific oxidizable functional group, not overall lipophilicity. The xanthines are the teaching cases: caffeine and theophylline both have negative logP values — they are freely water-soluble — yet both are cleared almost entirely by CYP1A2. This is exactly why a smoker (who induces CYP1A2) clears caffeine, theophylline, and also clozapine and olanzapine faster — and why stopping smoking can push clozapine levels up sharply. Ethanol is another small, water-soluble molecule with real oxidative metabolism (alcohol dehydrogenase plus CYP2E1). The lesson for the panel: read logP and TPSA to predict the usual route, but remember that a reactive handle on the molecule can send even a hydrophilic drug through the liver.
Crossing the membrane by carrier: the transporter routes
Everything so far assumes drugs cross membranes by passive diffusion, which is why size, lipophilicity, charge, and polar surface area predict so much. But a large and clinically important set of drugs cross only because a dedicated transporter carries them — and a second set are actively pumped back out. These carriers explain drugs that "shouldn't" work on paper (too polar, too large) and several that stay out of the brain despite looking like they should get in. Two superfamilies do most of the work: the SLC (solute carrier) uptake transporters and the ABC (ATP-binding cassette) efflux pumps.
LAT1 — the amino-acid gate every psychiatrist should know
LAT1 (SLC7A5), the large-neutral-amino-acid transporter, ferries amino acids across the blood–brain barrier — and a striking number of neuroactive drugs are built as amino-acid mimics precisely to hitch a ride on it. Gabapentin and pregabalin (leucine-like), L-DOPA (the Parkinson's prodrug that crosses where dopamine cannot), baclofen, and methyldopa all enter the brain via LAT1. This also creates a food interaction: a high-protein meal floods LAT1 with dietary amino acids and can blunt L-DOPA's effect by competition.
The other uptake carriers (SLC family)
PEPT1 (SLC15A1) is a peptide transporter exploited by prodrugs such as valacyclovir to boost oral absorption. MCT1 (SLC16A1), a monocarboxylate transporter, carries valproate and γ-hydroxybutyrate (GHB) across the barrier. OATP transporters (SLCO family) pull statins and other anions into hepatocytes for clearance — the site of many statin interactions. OCT and MATE transporters handle organic cations like metformin in the kidney (and are where the cimetidine–metformin interaction lives), and GLUT1 supplies glucose to the brain. The monoamine transporters SERT, NET, and DAT are also SLC-family carriers (SLC6) — but here they are the antidepressant and stimulant targets rather than delivery routes.
P-glycoprotein and the efflux pumps (ABC family)
Working against uptake, the ABC efflux transporters use ATP to pump drugs back out of cells and out of the brain. P-glycoprotein (P-gp / ABCB1 / MDR1) is the dominant one at the blood–brain barrier; it is why loperamide stays out of the CNS, and it actively refluxes many antipsychotics (risperidone, paliperidone, aripiprazole) and antidepressants — which is why ABCB1 polymorphisms have been studied as predictors of central drug exposure and response. The tool now flags these with a "Membrane transport" note on the affected structures. BCRP (ABCG2) and the MRP family (ABCC) provide additional efflux at the gut, liver, kidney, and barrier. Because many CYP3A4 inhibitors and inducers also modulate P-gp, a single interacting drug can shift brain and gut exposure independently of metabolism.
Truly large molecules: receptor-mediated transcytosis
Neither passive diffusion nor the small-molecule carriers can move an antibody or a therapeutic protein across the blood–brain barrier — they are simply too big. Biologics that need central action instead use receptor-mediated transcytosis, piggy-backing on the barrier's own receptors: the transferrin receptor and the insulin receptor are the two "molecular Trojan horse" routes now used to shuttle engineered antibodies into the brain. It is a reminder that the same barrier described by polar surface area for a small molecule becomes an entirely different engineering problem for a large one.
Intentional design: what chemists tune, and why
None of these properties is accidental. Modern psychopharmacology is largely the story of taking a molecule that works and re-engineering its structure to keep the target activity while dialing away liabilities. The tool's built-in lineages make this concrete — a few recurring design moves:
Strip the scaffold to remove off-target binding (tricyclic → SSRI)
The flat, fused three-ring core of amitriptyline and imipramine incidentally blocks H₁, muscarinic, and α₁ receptors and destabilizes cardiac conduction — the source of TCA sedation, dry mouth, orthostasis, and lethal overdose. Discarding the tricyclic scaffold for an open-chain phenylpropylamine (fluoxetine) keeps serotonin-reuptake inhibition but sheds those liabilities. Design goal: selectivity and overdose safety.
The chiral switch: keep the active half (citalopram → escitalopram)
Citalopram is a 50/50 mix of two mirror-image forms; the R-enantiomer is nearly inactive at the serotonin transporter and even partially opposes the active S-form. Purifying to the single S-enantiomer (escitalopram) gives fuller target occupancy per milligram and a cleaner dose–response. Design goal: potency and tolerability from stereochemistry alone.
Promote the active metabolite (venlafaxine → desvenlafaxine)
Venlafaxine is converted by CYP2D6 to its active O-desmethyl metabolite. Marketing that metabolite as its own drug (desvenlafaxine) removes a metabolism step, flattening the variability that CYP2D6 genetics and interactions introduce. Design goal: more predictable exposure across patients.
Tune lipophilicity for depot delivery (esterified long-acting injectables)
Attaching a fatty-acid chain to a drug (fluphenazine decanoate, paliperidone palmitate, aripiprazole lauroxil) makes it dramatically more lipophilic and poorly water-soluble on purpose, so an oil-depot or crystal dissolves and releases slowly over weeks. Design goal: a long dosing interval and steady levels for adherence.
Prodrugs and peripheral restriction
A prodrug is deliberately inactive until the body converts it — lisdexamfetamine is dextroamphetamine bonded to lysine, cleaved only after absorption, which smooths the concentration curve and reduces abuse potential. Conversely, chemists sometimes add a permanent charge (a quaternary amine, e.g., methylnaltrexone) so a drug can't cross the blood–brain barrier, keeping its action peripheral. Design goals: controlled activation, abuse deterrence, and confining effects to where they're wanted.
Read together, these show the same panel of properties being pushed in opposite directions for different ends: lipophilicity up for a depot but down for a renally cleared, dialyzable drug; ionization added to keep a molecule out of the brain but removed to let one in; a metabolite eliminated for predictability or exploited for a prodrug. When you open two structures side by side in the Chemical Structures tool and read off the deltas in MW, cLogP, TPSA, and ionization, you are reading the chemist's intent.
Bringing it to the bedside
You don't need to compute any of this — the tool does — but a habit of glancing at four numbers pays off. cLogP and TPSA tell you whether a drug reaches the brain and how sedating or fat-accumulating it will be. Ionization plus protein binding tell you the active free fraction, the displacement- and albumin-sensitivity, and the dialyzability. The renal-vs-hepatic pattern that emerges from logP, size, and charge tells you whether to worry about GFR or about CYP interactions and liver function. Structure is destiny, and most of that destiny is legible in half a dozen numbers.