Title: Method for the preparation of a co-crystal

The present invention relates to a method for the preparation of a co-crystal from an active chemical ingredient (ACI) molecule, in particular an active pharmaceutical ingredient (API) molecule and an co-moiety. The invention also relates to a method for the selection of a co-moiety for co-crystallization with an active chemical ingredient (ACI) molecule, in particular an active pharmaceutical ingredient (API) molecule. The invention further relates to a method for the selection for a solvent for use in the preparation of a co- crystal . A pharmaceutical drug comprises an active ingredient, i.e. an active pharmaceutical ingredient. The active pharmaceutical ingredient is the chemical compound that is used for treatment of a disease or pain relief or the like. The pharmaceutical drug may comprise further compounds e.g. for enabling the drug to be formulated as a tablet or for dissolving the active pharmaceutical ingredient (diluents, excipients, disintegrants, formulation aids, coatings and the like). In particular, e.g. for dosing in a solid form such as a tablet, the active pharmaceutical ingredient may need to be provided in the form of a co-crystal of the API with a co- crystallizing moiety. This is achieved by bringing the API in contact with a co-crystal forming molecule, which hereinafter may be referred to as a co-former, co-moiety molecule or a co-moiety. Also, different co-crystals may exhibit different pharmacological behaviour, such as rate of dissolution, formulation behaviour, solubility etc which can influence the dose and dosage form. It is hence desirable to obtain co-crystals of active pharmaceutical ingredients. In WO2004078163, a large number of API's and possible co-crystal formers are described as an example, as well as methods for the preparation of some combinations of these co-crystals and the advantages of co-crystals. The document however, does not provide any insight on how to obtain co-crystals or how to select co-crystal formers.

In a known method for selecting a co-moiety for co- crystallization, the ACI is mixed with a known possible co-crystal forming moiety. The combination is subjected to conditions (in solution or solid form, i.e. ball milling) that may lead to the

formation of a co-crystal and it is determined whether or not a co- crystal has been formed between the ACI and the co-crystal forming moiety. This experiment is repeated for each co-crystal forming moiety. Given the number of possible co-crystal forming moieties (thousands) and the variety of methods that can be used for the formation of co-crystals (solution-based crystallization, ball milling, grinding, heating, melt crystallization) , and experimental parameters that can be varied (temperature profiles, solvent combinations, solvent / anti-solvent combinations, crystallization aids) there is an abundance of parameters and conditions available that can be varied. Based on these experiments, it is experimentally determined which compound may be applied as a co-moiety.

The above-mentioned experiments require a substantial amount of time and are therefore expensive. Moreover, for each new active ingredient the experiments need to be repeated, as even a small difference between two active ingredients, as small as exists between enantiomers, may result in different co-moieties required for co- crystallization. Furthermore, there is no governing theory that provides an adequate understanding of the processes that underlie the formation of co-crystals, let alone a theory that would enable the prediction of possible co-crystals.

A similar problem arises when attempting to find a solvent for crystallising the co-crystal. In particular, the solvent needs to be suitable for dissolving both the active ingredient and the co-moiety. Finding a suitable solvent again requires a large number of experiments, again not governed by theory but based on trial and error.

It is desirable to have a method for selecting at least one potentially suitable co-moiety and/or at least one potentially suitable solvent prior to performing the above-mentioned experiments. In accordance with the present invention, a method is provided in which a molecule of the active chemical ingredient (ACI) , in particular a molecule of the active pharmaceutical ingredient (API), is modelled using molecular modelling techniques such as molecular mechanics, semi-empirical or quantum mechanical modelling. The resulting molecular model represents at least a part of the physical and/or chemical properties of the ACI molecule.

Further, the method according to the present invention comprises providing a molecular model for each co-moiety of a set of known co-moieties. Each molecular model represents at least part of

the physical and/or chemical properties of a respective co-moiety from the set of known co-moieties.

Using the molecular model of the ACI molecule and the molecular models of the co-moieties, a combination of the ACI molecule and each co-moiety of the set of co-moieties may be investigated. In particular, it has been found that the possibility for forming a hydrogen bond between the ACI molecule and the co- moiety molecule largely increases the possibility to form a co- crystal pair. Therefore, in accordance with the present invention, for each co-moiety in the set of co-moieties, it is determined whether at least one hydrogen bond between the ACI molecule and the co-moiety is possible. Thereto, the molecular model of the ACI molecule and the molecular model of the co-moiety are compared. Any co-moiety not capable of hydrogen-bonding with the ACI molecule is excluded, thereby obtaining a set of possibly hydrogen-bonding co- moieties. Hence, the number of possible co-moieties is reduced and a smaller amount of experiments may suffice to find a suitable co- moiety.

In an embodiment, the determination whether a hydrogen bond is possible comprises determining which atoms in the molecular model of the ACI molecule are suitable as a hydrogen bond donor atom or a hydrogen bond acceptor atom; and determining which atoms in the molecular model of the co-moiety are suitable as a hydrogen bond donor atom or a hydrogen bond acceptor atom. A hydrogen bond is only possible if the at least one of the ACI molecule and the co-moiety comprises a donor atom and the other comprises an acceptor atom. The method according to the present invention may comprise determining a set of likely conformations of the ACI molecule, each conformation having an energy within a predetermined energy band and/or may comprise determining a number of likely conformations of each co-moiety in the set of co-moieties. The atoms in a molecule may take a position with respect to each other, thereby allowing that a molecule comprised of a certain set of atoms may take a number of shapes. These different shapes are known in the art as conformations. A conformation may be a likely conformation, if an energy level of the molecule is relatively low. More in particular, the conformation is likely, if a change of conformation would result in an increased energy level of the molecule, i.e. the energy level is a locally low energy level. Thus, a molecule may have a number of likely conformations. The actual conformation of the molecule may influence

the ability to form a co-crystal. Therefore, each likely conformation of the ACI molecule and/or a co-moiety is determined and used for comparison.

In an embodiment, the method comprises screening the set of known co-moieties for toxicity, in particular if the active chemical ingredient (ACI) is an active pharmaceutical ingredient (API) . It is noted that the co-moiety is preferably not toxic, but may be toxic dependent on the active pharmaceutical ingredient (API) and the actual resulting reaction of the body. For example, if a co-moiety is to be used in combination with a drug for chemotherapy, the co-moiety may exhibit a certain degree of toxicity, since the drug itself is in fact toxic. In addition, the co-moiety may be toxic, but only at levels that are not reached when compared with the targeted dose of the API. The toxicity screening may be performed prior to or after the above-described hydrogen bonding evaluation.

The molecular model of the ACI and the model of the co- moieties may comprise a set of parameters of the molecule representing a number of physical parameters and/or a number of chemical parameters. The model may also be a two-dimensional model providing constitutional and topological information about the actual molecule. In an embodiment the model is a three-dimensional model providing a geometrical representation of the actual molecule.

If the molecular model comprises geometrical information, in particular if the model comprises a three-dimensional representation, the method may further comprise steric comparison between the molecular model of the ACI molecule and the molecular model of the co-moiety for determining whether the ACI molecule and the co-moiety are sterically compatible. A co-moiety that is sterically incompatible is excluded. If the molecular model comprises geometrical information, in particular if the model comprises a three-dimensional representation, the method may further comprise electrical charge distribution comparison between the molecular model of the ACI molecule and the molecular model of the co-moiety for determining whether the ACI molecule and the co-moiety are electrically compatible.

In one aspect, the invention relates to a method for the preparation of a co-crystal by crystallising an ACI with a co-moiety that has been obtained or selected by the process of selecting a co- moiety as described herein elsewhere. In one embodiment of the present invention, the method comprises selecting and/or providing an

ACI. In one embodiment, the method comprises providing the one or more potentially suitable co-moieties. In one embodiment, the method comprises crystallising the co-crystal of the ACI and the co-moiety. In one embodiment, the method comprises providing the selected solvents. In one embodiment, the method comprises crystallising the co-crystal in the potentially suitable solvent.

Method for the preparation of a co-crystal from an active chemical ingredient (ACI) , in particular an active pharmaceutical ingredient, and a co-moiety comprising the steps of a) selecting and/or providing an ACI; b) providing a molecular model representing at least a part of the physical and chemical properties of the ACI molecule; c) providing a molecular model for each of a set of known co- moieties representing at least a part of the physical and chemical properties of each respective co-moiety; d) for each co-moiety in the set of step (c) , determining whether at least one non-covalent intermolecular bond between the ACI molecule and the co-moiety is possible by comparing the molecular model of the ACI molecule as provided in step (b) and the molecular model of the co- moiety as provided in step (c) ; where non-covalent intermolecular bonds comprise hydrogen bonds, dispersive interactions, pi-pi interactions and electrostatic interactions, e) excluding any co-moiety not capable of forming a non- covalent intermolecular bond with the ACI molecule, thereby obtaining a set of possibly non-covalent bonding co-moieties; and f) crystallising and isolating the co-crystal.

One aspect of the present invention is the method for selecting a co- moiety for co-crystallization with an active chemical ingredient, ACI, molecule, API, molecule, the method comprising:

(a) providing a molecular model representing at least a part of the physical and chemical properties of the ACI molecule;

(b) providing a molecular model for each of a set of known co-moieties representing at least a part of the physical and chemical properties of each respective co- moiety;

(c) for each co-moiety in the set of step (b) , determining whether at least one non-covalent intermolecular bond between the ACI molecule and the co-moiety is possible by comparing the molecular model of the ACI molecule as provided in step (a) and the molecular model of the co- moiety as provided in step (b) ; where non-covalent intermolecular bonds comprise hydrogen bonds, dispersive interactions, pi-pi interactions and electrostatic interactions, and (d) excluding any co-moiety not capable of forming a non- covalent intermolecular bond with the ACI molecule, thereby obtaining a set of non-covalent binding co- moieties .

In another aspect, the present invention provides a method for selecting a potentially suitable solvent for dissolving an active chemical ingredient, ACI, molecule, in particular an active pharmaceutical ingredient, API, molecule. Thus, the invention provides a method for selecting a solvent for dissolving an active chemical ingredient, ACI, molecule, in particular an active pharmaceutical ingredient, API, molecule, the method comprising:

(a) providing an ACI;

(b) providing a molecular model representing at least a part of the physical and chemical properties of the ACI molecule;

(c) providing a molecular model for each of a set of known solvents representing at least a part of the physical and chemical properties of each respective solvent;

(d) for each solvent in the set of step (c) , determining whether at least one hydrogen-bond between the ACI molecule and the solvent is possible by comparing the molecular model of the ACI molecule as provided in step (b) and the molecular model of the solvent as provided in step (c) ; (e) excluding any solvent not capable of hydrogen-bonding with the ACI molecule, thereby obtaining a set of non- hydrogen-bonding solvents; and (f) dissolving the ACI in the solvent. In a further aspect, the present invention provides a computer program comprising computer readable instructions for operating a

computer system, the instructions comprising instructions for performing at least one of the methods according to the present invention.

Further advantages, features, and measures will become apparent to those skilled in the art upon reading the description below in relation to the appended drawings, in which Fig. 1 shows a diagram illustrating an embodiment of the method according to the present invention.

The steps as illustrated in Fig. 1 are described in arbitrary order. Moreover, in a practical embodiment, the order as presented in Fig. 1 may be different, as is apparent to one skilled in the art. In step SIl, a library of known co-moieties is provided. The library may be a virtual library comprising descriptions of the co- moieties. Co-moieties can be described by a set of parameters that represent a number of physical parameters and/or a number of chemical parameters. Such a set is also called a set of 'molecular descriptors' . Typically each co-former or solvent can be characterised using a set of molecular descriptors. Molecular descriptors in themselves are known, and are used to describe single molecular species properties. Typical molecular descriptors can be based on bulk physical properties such as boiling points, critical temperature, vapour pressure, flash point, auto-ignition temperature, density, refractive index, melting point, octanol-water partition coefficient, aqueous solubility of liquids and solids, molecular mass, water-air partition coefficients, GC retention times and response factor, critical micelle concentration, and molecular descriptors such as partial charge and charge densities, dipole moment, molecular surface area, molecular volume, electrostatic potential, bond lengths, bond angle, heat of formation, hydrogen bonding ability, etc. See also Katritzky et al . Chem. Inf. Comput. Sci, 2000, 40, 1-18 and references cited therein on descriptor generation. Molecular descriptors can also include hash-key or structure fingerprints, constitutional descriptors such as functional group counts, fragment contribution, atomic contributions, topological descriptors such as connectivity indices, Wiener numbers and Balaban indices or geometric descriptors such as solvent excluded volume and WHIM descriptors. It will be clear that molecular descriptors can easily be generated by the skilled person based on the above non-limiting examples thereof.

Additionally, the library may also comprise an actual library of known co-moieties as may be used in accordance with the prior art for performing experiments in order to find a suitable co-moiety by trial and error. In step S12, for each co-moiety in the library of step SIl, one or more likely conformations of the co-moieties is determined. These likely conformations can be determined by developing and analysing a molecular model for each of the co-moieties in the library. The molecular model preferably comprises a three-dimensional representation of the molecular structure. Alternatively, these likely conformations may be determined, may be known or may be determined from a database or library.

In step S13, which may be optional, for each likely conformation of each co-moiety in the library, a molecular model is determined. The molecular model may comprise a three-dimensional representation of the molecular structure. The molecular model may further comprise a representation of the electronic properties of the molecule .

In step S14, the co-moieties are screened with respect to toxicity. A toxicity screening is preferred if the active ingredient is a pharmaceutical drug that is to be administered to a patient. If the active ingredient is not a pharmaceutical drug, a toxicity screening may be omitted. However, a screening for any other aspect may be performed instead. For example, a co-moiety may chemically react with another compound such as the active ingredient. If such a reaction is not desired, a screening may be performed.

The screening of step S14 may also include a screening of the characteristic numbers of the co-moieties. A number of co-moieties, or conformations of co-moieties, may be excluded based on these characteristic numbers. These numbers may be comprised in the molecular descriptors as described herein elsewhere.

As mentioned above, the order of the steps SIl - S14 may be altered. For example, a toxic co-moiety may be excluded prior to determining a number of conformations, e.g. because it is known that all conformations are toxic. In another embodiment, conformations of a co-moiety are determined using a model, e.g. a three-dimensional model. Thus, after step SIl, a step may be performed comprising providing a molecular model suitable for determining conformations of the co-moiety. The step S12 is then performed using the model provided. Such a model may as well be used for determining which

conformations are likely conformations. As explained above, a conformation is likely, if its energy is at a local minimum, meaning that a transition to another conformation would require energy. The more energy that would be required for a transition, the more stable the conformation is, and hence the more likely such a conformation is .

A number of method steps S31 - S34 may be similar to the above- described steps SIl - S14, being performed mutatis mutandis in relation to a solvent instead of in relation to a co-moiety. In step S21, the active ingredient, usually an active pharmaceutical ingredient (API) , but in general an active chemical ingredient (ACI), is provided. The active ingredient of step S21 may have a number of conformations, of which a number are likely conformations and another number are not likely conformations. Considering that a relatively large number of likely conformations may be present and further considering that the likely conformations have overlapping characteristics and/or structures, in step S23, a principle component analysis (PCA) and/or clustering may be performed in order to reduce the number of conformations that need to be examined. Using PCA and/or clustering, sufficient insight into the separate structural differences may be obtained, such that the behaviour of any other conformation may be theoretically predicted based on the results of the conformations selected in step S23. Of course, the PCA/ clustering may as well, or instead, be performed on any characteristic number or features of the conformations as compared to structural differences.

In step S24, for each active ingredient as selected in step S23, a molecular model is provided that will yield a variety of conformations. The model may comprise characteristic numbers, parameters, three-dimensional representations, and any other relevant information. Thus, molecular models of solvents, co-moieties and the active ingredient are provided and a variety of conformations of each are provided.

In order to determine which co-moiety may be a co-crystal forming compound in combination with the active ingredient, in steps S41 - S43, a conformation of the molecular model of the active ingredient and a conformation of the molecular model of one of the co-moieties are compared. In particular, in step S41, it is determined whether one or more hydrogen bonds may be formed between the active ingredient conformation and the co-moiety conformation.

For example, it may be expedient to first determine whether a hydrogen bond donor atom and/or a hydrogen bond acceptor atom is present in the co-moiety molecule and whether a hydrogen bond acceptor atom and/or a hydrogen bond donor atom is present in the active ingredient molecule, respectively, such that a donor-acceptor pair may be formed between the co-moiety and the active ingredient. Preferably, a number of hydrogen bonds may be formed, thereby providing a hydrogen bond complex.

In particular if a hydrogen bond complex may be formed based on the number of donor and acceptor atoms, a steric comparison S42 and/or an electric charge distribution comparison S43 may be performed in order to determine whether the acceptor atoms and the donor atoms may get near each other, preferably in one of the likely conformations such that the energy of the combination of conformations is relatively low. For example, the electric charge at a certain part of one of the molecules may repel a part of the other molecule such that the donor atom and the acceptor atom cannot engage in an electronic interaction and hence a hydrogen bond, or a partial hydrogen bond cannot be formed. It is observed that in the formation of crystals other interactions that are not hydrogen bonds may play a role that are commonly depicted as weak forces, these are, for example dispersive interactions, pi-pi interactions, interactions between electron-donating and electron-accepting elements such as halogens, functional groups such as tertiary amines, ketones, salts etc.. These interactions can also be taken into account using the same technical principles and consideration as outlined herein for hydrogen bonds .

It is noted that the steric and electric charge distribution comparisons may require a relatively large number of computations as they require a spatial comparison, preferably a three-dimensional comparison. In particular in case of large molecules, such as pharmaceutical drug molecules, the number of computations may be large. As the number of ACI, preferably API, conformations and the number of co-moieties, more specifically the number of conformations of the co-moieties, is relatively large, the number of computations may become prohibitive for performing the method cost-effectively. Therefore, it may be expedient to make a first selection based on characteristic numbers and the number of donor atoms and/or acceptor atoms such that a smaller number of conformations need to be compared spatially.

Further, a principle component analysis (PCA) and/or cluster analysis may be performed in step S44, thereby selecting a broad range (i.e. a broad parameter space) of possibly suitable co- moieties . In step S45, the remaining co-moiety conformations that appear the most promising co-moieties are used in experiments and their number is further reduced by trial and error.

In order to select a suitable solvent, in step S51, a hydrogen bonding comparison is performed. The step S51 is mutatis mutandis similar to the step S41 for selecting a co-moiety. For dissolving the active ingredient in a solvent, a steric comparison may not be of interest. Likewise, an electric charge distribution may not be of interest. Then, a PCA and/or a cluster analysis may be performed in step S52, substantially similar to step S44. After the theoretical selection, the remaining possibly suitable solvents are used in experiments in step S45.

It is noted that the method according to the present invention may be improved using the experimental results obtained in step S45. In particular, the present invention includes a method for improving the above-described method for selecting a co-moiety and/or selecting a solvent.

As mentioned above, the co-crystallization may be performed in the solvent after dissolving both the active ingredient and the co- moiety. Therefore, in an embodiment, the co-moiety and the solvent are compared in accordance with e.g. the steps S51 - S53. One of the purposes of this comparison is to determine whether the solvent will compete with the co-former to create a solvate of the API instead of a co-crystal of the API and the co-former.

If a suitable co-moiety is selected such that the co-moiety and the active ingredient may form a co-crystal and such that the co- moiety and the active ingredient may be dissolved in the solvent, and it is not likely that a solvate will be formed, a further examination may be performed in order to determine whether the co-crystal will crystallize in the solvent and precipitate such that a solid co- crystal remains.