Table 2.

Non-hydrogen bondbased methods for co-former screening and formation prediction of multicomponent solid forms.

Prediction Method	System	API	Co-former	Preparation Method	Key Findings	Ref.
HSP (for both multicomponent crystal forms and co-amorphous systems)	multicomponent crystal forms	indomethacin	nicotinamide, saccharin, 4,4′-bipyridine cinnamic acid	liquid-assisted grinding, reaction crystallization	Indomethacin was miscible with 21 out of 33 co-formers according to HSP results. Indomethacin formed co-crystals with 4 different co-formers, 2 of which were newly developed.	[48]
	multicomponent crystal forms	paliperidone	benzamide, nicotinamide, para hydroxy benzoic acid	solvent evaporation	Paliperidone formed co-crystals with benzamide, nicotinamide, and para hydroxy benzoic acid, as HSP predicted using three group contribution methods.	[49]
	co-amorphous	tadalafil	repaglinide	solvent evaporation	HSP difference between tadalafil (26.01 MPa0.5) and repaglinide (26.01 MPa0.5) is less than 7 MPa0.5, confirming their good miscibility. There were no molecular interactions in the system.	[50]
	co-amorphous	florfenicol	oxymatrine	solvent evaporation	The Δδt value of florfenicol and oxymatrine is 3.87 MPa0.5, indicating miscibility.	[66]
	co-amorphous	norfloxacin	saccharin, naproxen, indomethacin, l-phenylalanine, l-arginine, l-tryptophan	dry ball mill	Δ$\overline{\delta }$ and Δδt criteria were used to predict the molecular miscibility between norfloxacin and 17 co-former candidates Norfloxacin can form co-amorphous with 6 co-formers. Van Krevelen criterion is more suitable for assessing molecular miscibility in the formation of norfloxacin co-amorphous.	[67]
COSMO-RS (only for multicomponent crystal forms)	multicomponent crystal forms	carbamazepine	dl-mandelic acid, dl-tartaric acid, indomethacin	liquid-assisted grinding	21 out of 75 co-former candidates were investigated, where 3 new systems and 9 already known systems were obtained. Only using ΔHex as a criterion for selecting co-formers is not sufficient Considering both the fusion entropy ΔSm and the excess enthalpy ΔHex, the results have a better prediction.	[51]
	multicomponent crystal forms	caffeine, theophylline	8 phenolic acids	liquid-assisted grinding	8 new co-crystals were discovered on caffeine and theophylline. Caffeine and theophylline showed a linear correlation in the mixing enthalpies and co-crystal forming abilities.	[52]
	multicomponent crystal forms	posaconazole	4-aminobenzoic acid, l-malic acid, succinic acid, fumaric acid, ferulic acid, maleic acid, citric acid, l-hydroxy-2-naphthoic acid, gentisic acid, salicylic acid, l-lactic acid, adipic acid, 3,4-dihydroxybenzoic acid	high-throughput slurry, liquid-assisted grinding, reaction crystallization	COSMOquick was used to reduce a list of about 140 potential co-formers to 28 candidates. 13 new posaconazole co-crystals (7 anhydrous, 5 hydrates, and 1 solvate) were successfully prepared.	[53]
	multicomponent crystal forms	clotrimazole	3,5-dinitrosalicylic acid, 3,5-dinitrobenzoic acid indole-6-carboxylic acid, syringic acid, 3-nitrobenzoic acid, 1,4-naphthalenedicarboxylic acid, pyromellitic acid, 2,3-dihydroxybenzoic acid, 1,2,4-benzenetricarboxylic acid	liquid-assisted grinding, slurry suspension, solvent evaporation	14 out of 21 potential co-formers formed multi-component crystal forms with clotrimazole, including 5 reported cases, 3 new co-crystals and 6 new salts.	[54]
	multicomponent crystal forms	2-hydroxybenzylamine	succinic acid, p-aminobenzoic acid, p-nitrobenzoic acid, o-nitrobenzoic acid, p-toluic acid, 2,3-dihydroxybenzoic acid, 3,4-dihydroxybenzoic acid, p-nitrophenol, 5-hydroxyisophthalic acid	liquid-assisted grinding	21 out of 40 potential co-formers were characterized as new solid phases. 9 multicomponent single crystals were obtained and characterized.	[55]
MC (only for multi-component crystal forms)	multicomponent crystal forms	artemisinin	resorcinol, orcinol	liquid-assisted grinding	Only 2 out of 75 co-formers (3%) led to the formation of a co-crystal. Low success rate was due to artemisinin lacking strong hydrogen bond donors or acceptors.	[45]
	multicomponent crystal forms	leflunomide	pyrogallol, 3-hydroxybenzoic acid, 2-picolinic acid, 2-aminopyrimidine	liquid-assisted grinding	Structure and intermolecular interactions of leflunomide were analyzed using Isostar and Mercury to identify favorable functional groups for co-crystallization, resulting in 5 new co-crystals.	[46]
	co-amorphous and multi-component crystal forms	sulfamethoxazole	acetamide, propionamide, isonicotinamide, 2-hydroxypyridine, pyrazine, imidazole, oxalic acid dihydrate, N-hydroxysuccinimide, 1,2-di(4-pyridyl)ethylene, 1,2-di(4-pyridyl)ethylene, 1,3-di(4-pyridyl)propane, 4,4′-bipyridine, 4-phenylpyridine, benzamidine, carbamazepine, deoxycholic acid, hexamethylenetetramine, sodium deoxycholate	neat grinding, solvent evaporation	CSD motif search offered 39 potential candidates and MC was used to screen the compatibility. 13 new co-crystals, 1 salt and 4 co-amorphous systems were identified experimentally.	[47]
lattice energy (only for multicomponent crystal forms)	multicomponent crystal forms	carbamazepine	isonicotinamide	cooling crystallization, slurry	Carbamazepine-isonicotinamide formed co-crystals due to lower or comparable lattice energy of pure components. Carbamazepine-picolinamide could not form co-crystal due to less stable lattice energies than pure components.	[43]
	multi-component crystals	pentoxifylline	aspirin, salicylic acid, benzoic acid	neat and liquid-assisted grinding, solvent evaporation	The experimental and in silico screening of co-crystals yielded consistent results. FlexCryst software (www.flexcryst.com (accessed on 23 July 2023)) suggests that the prediction of co-crystal formation must satisfy ΔG ≥ −3 kJ/mol for feasibility, instead of ΔG ≥ 0 kJ/mol.	[44]
Artificial intelligence (for both multicomponent crystal forms and co-amorphous systems)	multicomponent crystal forms	diclofenac	iIsonicotinamide, 2-pyrrolidinone, 4,4′-Bipyridine.	neat grinding	An extreme gradient boosting model was developed using 1000 co-crystallization cases and 2083 chemical descriptors. 3 new co-crystals of diclofenac were discovered.	[56]
	multicomponent crystal forms	captopril	l-proline, sarcosine	liquid-assisted grinding	A random forest-based co-crystal prediction model was created using a dataset of positive samples from CSD and negative samples from randomly paired molecules. 2 captopril co-crystals verified the effectiveness of the model.	[57]
	multicomponent crystal forms	norfloxacin	nicotinamide, 4,4′-vinylenedipyridine	neat grinding	COSMO-SVM and 3D-CNN machine learning models were established. 2 new norfloxacin co-crystals were predicted and fabricated.	[58]
	co-amorphous	folic acid	nicotinamide, l-isoleucine, anthranilic acid, citric acid, theophylline, theobromine.	neat grinding	The gradient boost model achieved a predictive accuracy of over 73%. 6 novel co-amorphous forms of folic acid were predicted and discovered.	[59]
	co-amorphous	glycopyrronium bromide	budesonide, ethambutol	neat grinding	A molecular descriptor-based ML model was built with an accuracy of 79%. 2 successful co-amorphous and one failed case were confirmed by the model.	[60]
	multicomponent crystal forms	nimesulide	4,4′-bipyridine, trans-1,2-bis(4-pyridyl)ethylene, 1,2-bis(4-pyridyl)ethyne, 1,2-bis(4-pyridyl)ethane	liquid-assisted grinding, slurry	Co-crystal formation between nimesulide and pyridine analogues depends on various molecular descriptors of co-formers, with MEP having the greatest impact, followed by h_ema (sum of hydrogen bond acceptor strengths), Kier flex (molecular flexibility), and horizontal distance between two N atom projections.	[61]
	co-amorphous	carvedilol, mebendazole, carbamazepine, furosemide, indomethacin, simvastatin	20 natural amino acids	neat grinding	The PLS-DA model effectively separated co-amorphous and non-co-amorphous samples. Non-polar side chain amino acids were the most effective co-formers for the formation of co-amorphous systems, while polar amino acids were the least successful.	[62]