Understanding the solid-state properties of an active pharmaceutical ingredient (API) is critical in formulation development
and a finished drug product. Patricia Van Arnum, senior editor, asked several leading experts to share their insight into
this field: Menno A. Deij, head of technology development at Avantium Pharma (Amsterdam); Chris Frampton, chief scientific
officer of SAFC Pharma's Pharmorphix Solid State Services (Cambridge, UK); and Noel Hamill, team leader in the physical sciences
group at Almac (Craigavon, UK).
Patricia Van Arnum
» Can you outline the major issues involved in polymorphism and advances in crystal-structure prediction?
A lot of effort goes into understanding the crystalline solid phase of the API. The pharmaceutical industry is required
to screen for polymorphism, that is, the ability of a molecule to exist as two or more crystalline phases that have different
arrangements and/or conformations of the molecules in the crystal lattice (1). At room temperature, there is only one polymorph
that fits the definition of the most stable; all other polymorphs are metastable and may convert to the most stable form (2).
Properties that can differ among solid forms include color, melting point, spectral properties, solubility, crystal shape,
water sorption and desorption properties, particle size, hardness, drying characteristics, flow and filterability, compressibility,
and density. These variations can lead to differences in dissolution rate, oral absorption, bioavailability, toxicology and
clinical results, hence both safety and efficacy are impacted (3, 4). The potential formation of multiple polymorphs of an
API is particularly challenging when producing solid oral- dosage drug products. To make solid oral-dosage forms with crystalline
APIs, pharmaceutical companies prefer to use the most stable polymorph to prevent the conversion to a less soluble polymorph,
which can affect the efficacy of a drug product (5).
In an ongoing effort over the past 20 years, the academic community has pursued the prediction of crystal polymorphs starting
from the molecular structure (6). Ideally, a researcher would like to search through the complete range of crystal structures
and evaluate their free energies accurately as a function of temperature and pressure to obtain the global minimum energies
of optimal crystal structures. The crystal-structure prediction process starts with the generation of millions of trial crystal
structures, followed by the optimization of these structures and ranking according to their lattice energy as a first approximation
to the lattice free energy. The crystal structure with the lowest lattice energy is expected to correspond to the most stable
polymorph that can be found experimentally. The differences between the stable and metastable forms is typically on the order
of only a few kJ/mol.
In the past, the lattice-energy ranking was based on molecular mechanics and generic forcefields. In this approach, molecules
are modeled as balls and springs, representing the atoms and bonds. The forcefield is a large table that describes the balls'
and springs' intramolecular behavior during bond stretching, bond bending, out-of-plane bending, and torsional rotation. The
intermolecular interactions are described as charge–charge interactions and induced dipole–dipole interactions (van der Waals
interactions). These forcefields are based on empirical data with fitted parameters. Forcefields are meant to be generic and
able to describe a lot of different situations. Because the problem of crystal-structure prediction requires very accurate
energy rankings, it turns out that the precision and accuracy of these forcefields is not sufficient.
The most recent developments, however, introduced by Marcus Neumann and his company, Avant-garde Materials Simulation, do
not use generic forcefields. Advanced quantum-mechanical calculations (dispersion-corrected density functional theory or d-DFT)
are used instead to develop a forcefield that is customized specifically for the molecule at hand (7, 8). This forcefield
is then used in the geometry optimization of the thousands of crystal structures generated. The most promising structures
with the lowest lattice energy are reranked with d-DFT, which results in an energy ranking that is of unprecedented accuracy
and precision, giving the required confidence in the computational results.