form, which could warrant investigating experimentally. The assessment of the
stability of a drug’s solid form can be more complex if it is formulated as a
co-crystal, an approach now often undertaken to improve a physical property of
the compound. Using the data from the CSD on robust supramolecular synthons, a
crystal engineering approach has been used to screen libraries of approved
co-formers for co-crystal design. Understanding whether a drug molecule is likely
to crystallise with a solvate or hydrate is still a challenge, and recent work has used
machine learning in an attempt to predict the likelihood for formation
[175, 176]. Machine learning approaches, using artificial neural networks, have
also been the subject of work to ascertain other solid-state properties, e.g. melting
point, lattice energy and crystal density [177].
Information from crystal structures is also valuable in assessing properties
impacting pharmaceutical manufacturing, such as predicting the ease of compressing
a potential drug form into a stable tablet. This method identifies slip planes between
packed layers of molecules and considers the ease with which they could move over
each other [178, 179]. The morphology of crystals has been shown to have an impact
on the flowability of the material during manufacturing and is influenced by the
growth rate of crystal faces [180]. This growth of a crystal face is influenced by the
interactions between an attaching molecule and the molecules in the crystal, which
has led to a variety of approaches to improve performance, by modifying the crystal
habit, using different solvent systems, dopants and crystallisation conditions
[181, 182].
3.4.2 Examples of Data-Driven Materials Discovery
MOFs have seen a dramatic rise this century [98], as shown above in Fig. 7. This
interest in MOFs is partially due to their range of potential applications in gas
storage, separation, catalysis, chemical sensors and drug delivery. MOF researchers
have made extensive use of the CSD, and several groups have generated databases of
structures, according to their own criteria, in order to then apply computational
screening methods [183–185]. With these approaches in mind, a CSD subset of
MOF structures, available through Conquest, has been assembled and is regularly
updated as the CSD grows [97]. This subset can be used to characterise useful
features such as classifying framework types [186], pore limiting diameter (PLD)
and largest cavity diameter (LCD) [187], absorption, flexibility and other physical
properties [188]. It has also been used to study the limits of hydrogen storage [189],
gas adsorption [190] and separation of different gases [191, 192]. The key feature of
all MOF materials is their high porosity, and there are now materials based on
covalent organic frameworks (COFs), hydrogen-bonded frameworks (HOFs) and
molecular cages that exhibit this high porosity. The exchange and identity of guest
molecules in the pores of these materials has been of interest, and one application
involving the encapsulation of guest molecules in the framework or cage has allowed
their structure determination when it has not been possible to crystallise them
independently, known as the crystalline sponge method [151, 193].
120
S. J. Coles et al.
stability of a drug’s solid form can be more complex if it is formulated as a
co-crystal, an approach now often undertaken to improve a physical property of
the compound. Using the data from the CSD on robust supramolecular synthons, a
crystal engineering approach has been used to screen libraries of approved
co-formers for co-crystal design. Understanding whether a drug molecule is likely
to crystallise with a solvate or hydrate is still a challenge, and recent work has used
machine learning in an attempt to predict the likelihood for formation
[175, 176]. Machine learning approaches, using artificial neural networks, have
also been the subject of work to ascertain other solid-state properties, e.g. melting
point, lattice energy and crystal density [177].
Information from crystal structures is also valuable in assessing properties
impacting pharmaceutical manufacturing, such as predicting the ease of compressing
a potential drug form into a stable tablet. This method identifies slip planes between
packed layers of molecules and considers the ease with which they could move over
each other [178, 179]. The morphology of crystals has been shown to have an impact
on the flowability of the material during manufacturing and is influenced by the
growth rate of crystal faces [180]. This growth of a crystal face is influenced by the
interactions between an attaching molecule and the molecules in the crystal, which
has led to a variety of approaches to improve performance, by modifying the crystal
habit, using different solvent systems, dopants and crystallisation conditions
[181, 182].
3.4.2 Examples of Data-Driven Materials Discovery
MOFs have seen a dramatic rise this century [98], as shown above in Fig. 7. This
interest in MOFs is partially due to their range of potential applications in gas
storage, separation, catalysis, chemical sensors and drug delivery. MOF researchers
have made extensive use of the CSD, and several groups have generated databases of
structures, according to their own criteria, in order to then apply computational
screening methods [183–185]. With these approaches in mind, a CSD subset of
MOF structures, available through Conquest, has been assembled and is regularly
updated as the CSD grows [97]. This subset can be used to characterise useful
features such as classifying framework types [186], pore limiting diameter (PLD)
and largest cavity diameter (LCD) [187], absorption, flexibility and other physical
properties [188]. It has also been used to study the limits of hydrogen storage [189],
gas adsorption [190] and separation of different gases [191, 192]. The key feature of
all MOF materials is their high porosity, and there are now materials based on
covalent organic frameworks (COFs), hydrogen-bonded frameworks (HOFs) and
molecular cages that exhibit this high porosity. The exchange and identity of guest
molecules in the pores of these materials has been of interest, and one application
involving the encapsulation of guest molecules in the framework or cage has allowed
their structure determination when it has not been possible to crystallise them
independently, known as the crystalline sponge method [151, 193].
120
S. J. Coles et al.