Hydrogen-bond-driven coformer selection, high-throughput screening, and computational prediction to boost aqueous solubility of poorly soluble APIs via pharmaceutical cocrystals
The Biopharmaceutics Classification System (BCS) sorts drugs by aqueous solubility and intestinal permeability. BCS Class II compounds (low solubility, high permeability) and Class IV (low solubility, low permeability) dominate modern drug discovery pipelines as medicinal chemistry increasingly favors larger, more lipophilic, more potent molecules. When a molecule lacks an ionizable center suitable for salt formation — or when the available salts are hygroscopic, disproportionate in the GI tract, or simply do not raise solubility enough — cocrystal engineering offers a complementary, non-covalent route to modify the solid-state properties of the API without changing its pharmacology.
Solid-form strategies available to formulation scientists, and their limitations:
Polymorph screening: • Different packing arrangements of the SAME molecule can differ in solubility 2–3× at most (thermodynamically bounded by the free energy difference between polymorphs, typically <2 kcal/mol for enantiotropic pairs) • Does not change molecular structure or intrinsic hydrophilicity — a fundamentally limited lever
Salt formation: • Requires an ionizable acidic or basic group with a counter-ion of sufficiently different pKa (ΔpKa >~3 rule of thumb for a stable salt) • Roughly a quarter of drug-like molecules have no practically ionizable group, or the group is too weakly basic/acidic (pKa near neutral) for robust salt formation • Salts can suffer from hygroscopicity, disproportionation (reversion to free acid/base) in the GI tract or under humid storage, and limited counter-ion choices for chronic dosing
Amorphization (see companion topic: amorphous solid dispersions): • Can give large (10–100×) transient solubility gains but risks physical instability and recrystallization over shelf life — a fundamentally metastable, kinetically-trapped state requiring polymer stabilization
Cocrystals — the complementary approach: • A cocrystal is a crystalline material composed of two (or more) distinct molecular components — API and "coformer" — in a well-defined stoichiometry, held together in the same crystal lattice by non-ionic, non-covalent interactions (primarily hydrogen bonds) • Unlike salts, does NOT require full proton transfer between API and coformer, dramatically expanding the palette of usable partner molecules — carboxylic acids, amides, alcohols, aromatic N-heterocycles are all viable coformers regardless of the API's own ionization state • Modifies the crystal lattice energy directly: a cocrystal with a highly soluble, hydrophilic coformer typically has a LOWER lattice energy / higher solubility than the pure API crystal, since the new lattice no longer optimizes purely for the API's own self-complementary packing • Retains the manufacturing, stability, and regulatory advantages of a crystalline solid (unlike amorphous dispersions) while achieving a genuinely new solid form with distinct physicochemical properties — an entirely different XRPD fingerprint, melting point, and dissolution profile from the parent API
The central scientific challenge, addressed in the following stages, is: out of thousands of pharmaceutically acceptable small molecules, which ones will actually cocrystallize with a given API, and which of those will meaningfully improve solubility and dissolution rate?
Rather than screening every conceivable molecule, cocrystal design begins with supramolecular synthon theory: recurring, robust hydrogen-bond patterns (synthons) observed across the Cambridge Structural Database (CSD) predict which functional group pairs reliably form cocrystals. Combined with the ΔpKa rule (which distinguishes a cocrystal outcome from a salt outcome) and restriction to the GRAS/pharmaceutical-excipient coformer universe, this narrows thousands of candidate molecules to a rationally justified shortlist of 20–50 for experimental screening.
Supramolecular synthons — the building-block logic of cocrystal design:
Robust heterosynthons (API···coformer, most reliable, from CSD statistical analysis, Etter's rules and subsequent work by Aakeröy, Desiraden, and others): • Carboxylic acid···pyridine (COOH···N): forms in >70% of cases where both groups are present and unhindered — one of the most reliable synthons in the literature • Carboxylic acid···carboxamide (COOH···CONH2): common in API–coformer pairs, e.g., many cocrystals with nicotinamide, isonicotinamide • Phenol···pyridine, hydroxyl···carbonyl: weaker but still statistically favored • Homosynthons (API···API or coformer···coformer, e.g., carboxylic acid dimer COOH···HOOC) COMPETE with the desired heterosynthon — a good coformer choice disrupts the API's own homosynthon and replaces it with a heterosynthon to the coformer
The ΔpKa rule — cocrystal vs. salt boundary: • ΔpKa = pKa(conjugate acid of base) − pKa(acid) • ΔpKa < 0: cocrystal strongly favored (no significant proton transfer) • 0 < ΔpKa < 2–3: ambiguous/mixed region — may form a cocrystal, a salt, or a "cocrystal salt" continuum depending on solid-state environment (this ambiguity itself is now well documented crystallographically) • ΔpKa > 3: salt strongly favored (full proton transfer) • Practical use: computing ΔpKa for API + candidate coformer pairs BEFORE screening lets formulators deliberately steer toward cocrystal (low ΔpKa coformers) vs. salt (high ΔpKa) outcomes
GRAS and pharmaceutical-excipient coformer universe: • Restricting candidates to GRAS (US), EAFUS, or existing approved pharmaceutical excipients (citric acid, tartaric acid, saccharin, nicotinamide, urea, sugars, sugar alcohols) drastically reduces toxicology and regulatory burden, since the coformer itself must be qualified as safe for chronic human exposure • Common successful coformer classes: dicarboxylic acids (succinic, fumaric, malonic, glutaric), aromatic carboxylic acids (benzoic, salicylic), amides (nicotinamide, saccharin), sugars/polyols (sorbitol, xylitol) • Coformer molecular weight and melting point matter: low-MW, low-melting coformers tend to depress the cocrystal melting point (via a simple mixture-rule argument), often correlating with improved dissolution — a heuristic screening filter applied even before synthon analysis
Output of Stage 2: a ranked shortlist (typically 20–50 candidates) balancing synthon compatibility, ΔpKa-predicted cocrystal favorability, GRAS status, and complementary physicochemical properties (melting point, hydrophilicity) — the input list for the high-throughput experimental screen in Stage 3.
With a rational shortlist in hand, experimental cocrystal screening is run in parallel across multiple methods, since no single crystallization technique reliably discovers every thermodynamically accessible cocrystal. Liquid-assisted grinding (LAG) is fast and solvent-sparing; slurry conversion approaches thermodynamic equilibrium and best predicts the stable form; cooling/evaporative crystallization from solution provides material for full characterization. PXRD is the primary rapid readout distinguishing a genuine new cocrystal phase from a simple physical mixture.
Three complementary experimental screening techniques:
1. Liquid-assisted grinding (LAG) / solvent-drop grinding: • API + coformer (1:1 or other stoichiometry) ground together mechanically (mortar/pestle or ball mill) with a few drops (η ≈ 0.25–1 µL/mg) of solvent added as a catalytic liquid phase • Solvent choice affects outcome: low-polarity solvents (heptane, cyclohexane) often favor the thermodynamic product by allowing limited molecular mobility for reorganization; more polar solvents can promote solvate formation instead of the target cocrystal • Fast (20–60 minutes), minimal solvent use, easily parallelized across a 96-well grinding plate — the workhorse first-pass screening method in modern cocrystal discovery
2. Slurry conversion: • Excess solid API + coformer suspended in a small volume of solvent (often solvent-saturated to avoid net dissolution) and stirred for 24–72 hours at controlled temperature • Because solid-solid transformation to the thermodynamically most stable form occurs via a dissolution–recrystallization pathway, slurry conversion is considered the gold-standard method for identifying the STABLE cocrystal form (as opposed to a kinetically trapped metastable form that LAG can sometimes produce) • Also used downstream to confirm that a LAG-discovered hit is thermodynamically robust and not merely a transient co-grinding artifact
3. Cooling / evaporative solution crystallization: • API and coformer co-dissolved in a solvent (often chosen via solubility/ternary phase diagram considerations) then cooled or evaporated to induce crystallization • Produces larger, better-formed crystals suitable for single-crystal X-ray diffraction (definitive structure confirmation) and for generating enough material (mg–g scale) for full physicochemical characterization • Ternary phase diagrams (API–coformer–solvent) rationally guide the composition/solvent choice that favors cocrystal over separate API and coformer crystallization
PXRD-based hit triage (primary screening readout): • A physical mixture of API + coformer shows a PXRD pattern that is simply the SUPERPOSITION of the two pure-component patterns • A new cocrystal phase shows a DISTINCT PXRD pattern with new peaks not present in either starting material — the diagnostic, unambiguous signature of a new crystal phase (see companion topic: XRPD polymorph fingerprinting for the underlying diffraction principles) • DSC (differential scanning calorimetry) cross-check: a single, sharp melting endotherm at a temperature DIFFERENT from both pure components (not simply an intermediate eutectic melt with two events) supports a true cocrystal rather than a simple eutectic mixture • FTIR/Raman: shifts in characteristic API carbonyl or N-H/O-H stretch frequencies confirm new hydrogen-bonding environment consistent with the predicted heterosynthon
Typical outcome: of a rationally selected 40-coformer shortlist, 15–25% (6–10 coformers) yield a confirmed new cocrystal phase by PXRD — a substantially higher hit rate than unbiased random screening (typically <5%), demonstrating the value of the Stage 2 rational selection step.
Computational coformer prediction tools let formulators rank hundreds of candidate coformers by predicted cocrystallization propensity BEFORE committing bench time and material — particularly valuable when API supply is limited (common in early development). COSMO-RS (Conductor-like Screening Model for Real Solvents) predicts excess mixing enthalpies from quantum-mechanical surface charge distributions, while Hansen solubility parameter (HSP) distance offers a simpler, faster empirical miscibility screen, together enabling a 5–10× reduction in the experimental screening space.
COSMO-RS excess enthalpy prediction:
• Each molecule (API and candidate coformer) is first modeled quantum-mechanically (DFT, typically BP86/TZVP level) in a virtual conductor to generate its "σ-surface" — a map of screening charge density over the molecular surface • COSMOtherm statistical thermodynamics then combines the two σ-surfaces (in the correct API:coformer stoichiometric ratio) to predict the excess enthalpy of mixing, H_excess, for the hypothetical 1:1 (or other ratio) complex relative to ideal mixing • More negative H_excess = more favorable interaction between API and coformer = higher predicted cocrystallization propensity • Validated retrospectively across published cocrystal screening datasets (e.g., carbamazepine, theophylline, indomethacin coformer libraries): COSMO-RS correctly ranks successful cocrystallizers above non-cocrystallizing controls with good (though imperfect, ~70–80%) discrimination, sufficient for triage/prioritization rather than absolute go/no-go decisions • Computationally cheap once molecular σ-profiles are generated (minutes per pair on standard hardware), enabling combinatorial screening of hundreds to thousands of API–coformer pairs entirely in silico
Hansen solubility parameter (HSP) distance: • Each molecule is characterized by three partial solubility parameters: δD (dispersion forces), δP (polar forces), δH (hydrogen bonding) — together defining a point in 3D "Hansen space" (units: MPa^0.5) • HSP distance between API and coformer: Ra = √[4(δD1−δD2)² + (δP1−δP2)² + (δH1−δH2)²] • Empirically, cocrystal-forming pairs cluster at SMALL Ra (typically <5–7 MPa^0.5) — molecules with similar overall polarity/H-bonding character are more likely to co-crystallize and, importantly, to remain MISCIBLE rather than phase-separating during processing • HSP calculation is much faster than COSMO-RS (group-contribution methods, e.g., Hoftyzer–Van Krevelen or Hoy, give δD/δP/δH directly from molecular structure in seconds) making it useful as a first-pass filter before more expensive COSMO-RS or experimental screening • Also directly useful for predicting drug-polymer miscibility in amorphous solid dispersions (see companion topic), since the same HSP framework applies
Combined computational-experimental workflow: • Step 1: HSP-based fast filter reduces a coformer database (hundreds to a few thousand GRAS molecules) to the ~100 most HSP-compatible candidates • Step 2: COSMO-RS excess enthalpy ranks these 100 by predicted cocrystallization propensity, selecting the top 30–50 for experimental LAG/slurry screening (Stage 3) • Step 3: experimental hit rate on the computationally pre-filtered set is typically 20–30% higher than on a synthon-rules-only shortlist, and dramatically higher than random/exhaustive screening — directly translating to time and API-material savings during early formulation development • Machine-learning approaches (random forest, graph neural networks trained on the CSD cocrystal database) are an active area of extension, aiming to predict cocrystal formation probability directly from 2D molecular structure without requiring quantum-chemical calculation
The lead cocrystal candidate must be unambiguously confirmed as a genuine new crystal phase (not a salt, polymorph, or solvate), its solubility and dissolution advantage quantified against the parent API, and its regulatory classification established — since cocrystals, salts, and polymorphs are treated differently under FDA and ICH frameworks, directly affecting the regulatory pathway and intellectual property strategy for the final drug product.
Definitive solid-form characterization of the lead cocrystal:
Single-crystal X-ray diffraction (SC-XRD): • Gold standard: unambiguously determines unit cell, space group, and atomic-resolution hydrogen-bond geometry (D···A distances, D-H···A angles) confirming the predicted heterosynthon • Distinguishes true cocrystal (discrete, non-ionized API and coformer molecules in the asymmetric unit) from a salt (evidence of proton transfer: shortened C-O bonds in a carboxylate, protonated pyridinium N-H) or a solvate/hydrate (solvent molecule incorporated in the lattice)
Differential scanning calorimetry (DSC) and thermogravimetric analysis (TGA): • DSC: single sharp melting endotherm at a Tm distinct from both API and coformer confirms a true single-phase cocrystal (vs. a eutectic mixture showing a lower-temperature eutectic event followed by a second melt) • TGA: confirms absence of unexpected solvent/water loss (rules out solvate misassignment)
Solubility and dissolution testing: • Kinetic (apparent) solubility: shake-flask or dynamic method measuring concentration achieved from excess solid cocrystal over time — reflects the practically achievable supersaturation before the cocrystal itself converts back to the less soluble parent API (an important stability consideration, since many cocrystals are only KINETICALLY more soluble and can convert to the stable, less-soluble API form during the dissolution test itself) • Intrinsic dissolution rate (IDR): USP rotating-disk or stationary-disk apparatus, dissolution rate normalized to constant surface area (mg/min/cm²) — the most rigorous, surface-controlled comparison versus the parent API, avoiding confounding particle-size effects • Case-study-scale improvements: 2–10× kinetic solubility gain and 3–8× IDR improvement are typical for a well-selected cocrystal versus the parent crystalline API; larger gains (>10×) are occasionally seen for very poorly soluble parent forms
Regulatory classification (critical practical distinction): • FDA (2016 draft, finalized 2018 guidance "Regulatory Classification of Pharmaceutical Co-Crystals"): a cocrystal is classified as a drug product intermediate, not a new active ingredient — provided it dissociates into the API and coformer upon dissolution/administration, with the coformer itself independently GRAS or an approved excipient • This is DIFFERENT from a salt (which IS considered a distinct active moiety requiring its own NDA-supporting data package under most circumstances) and different from a polymorph (same molecule, different lattice — governed under ICH Q6A as a specification/control-strategy issue rather than a distinct entity) • Practical consequence: a cocrystal can often be developed under a 505(b)(2) pathway leveraging existing API safety/efficacy data, provided bioequivalence and solid-state stability are demonstrated — a materially faster and lower-risk regulatory route than developing a novel salt or covalent prodrug • ICH Q6A / Q3C: the coformer, even if GRAS as a food additive, still requires qualification for the intended route, dose, and duration of administration in the specific drug product context — GRAS status for food use does not automatically satisfy pharmaceutical excipient qualification
A well-documented industry case: the marketed cocrystal of a poorly soluble kinase inhibitor with a small dicarboxylic acid coformer (developed under FDA's cocrystal guidance framework) improved intrinsic dissolution rate roughly 6-fold and enabled a lower, more consistent oral dose. Because the cocrystal was classified as a drug product intermediate rather than a new active ingredient, the sponsor leveraged existing preclinical safety data for the parent API, substantially shortening development timeline versus what a novel salt or prodrug strategy would have required — illustrating why regulatory classification, decided early and correctly, is as consequential to program success as the chemistry itself.