Optimization and comparison of statistical tools for the prediction of multicomponent forms of a molecule: the antiretroviral nevirapine as a case study

• Published in: CrystEngComm 2020-11, Vol.22 (43), p.746-7474
• Main Authors: Nunes Costa, Rogeria, Choquesillo-Lazarte, Duane, Cuffini, Silvia Lucía, Pidcock, Elna, Infantes, Lourdes
• Format: Article
• Language: English
• Subjects: Benzoic acid; Crystallography; Crystals; Data centers; Dihydroxybenzoic acid; Optimization; Pharmaceuticals; Ranking; Screening; Statistical analysis; Statistical prediction
• ISSN: 1466-8033; 1466-8033
• DOI: 10.1039/d0ce00948b

In the pharmaceutical area, to obtain structures with desired properties, one can design and perform a screening of multicomponent forms of a drug. However, there is an infinite number of molecules that can be used as co-formers. Aiming to avoid spending time and money in failed experiments, scientists are always trying to optimize the selection of co-formers with high probability to co-crystallize with the drug. Here, the authors propose the use of statistical tools from the Cambridge Crystallographic Data Centre (CCDC) to select the co-formers to be used in a pharmaceutical screening of new crystal forms of the antiretroviral drug nevirapine (NVP). The H-bond propensity (HBP), coordination values (CV), and molecular complementarity (MC) tools were optimized for multicomponent analysis and a dataset of 450 molecules was ranked by a consensus ranking. The results were compared with CosmoQuick co-crystal prediction results and they were also compared to experimental data to validate the methodology. As a result of the experimental screening, three new co-crystals - NVP-benzoic acid, NVP-3-hydroxybenzoic acid, and NVP-gentisic acid - were achieved and the structures are reported. Since each tool assesses a different aspect of supramolecular chemistry, a consensus ranking can be considered a helpful strategy for selecting co-formers. At the same time, this type of work proves to be useful for understanding the target molecule and analyzing which tool may exhibit more significance in co-former selection. A methodology is proposed to assess the propensity to obtain multicomponent forms of an API based on the combination of modified statistical analytical tools to order the possible co-formers in a ranking index.