Vol.2 No.1 2009

72/88

Research paper : Advanced in-silico drug screening to achieve high hit ratio (Y. Fukunishi et al.)−69−Synthesiology - English edition Vol.2 No.1 (2009) compared with random screening[16].The DSI method searches for compounds analogous to known active compounds using the protein-compound interaction matrix. Even different compounds that bind to the same protein are considered analogous (Fig. 5). The DSI method does not require the 3D structure of target proteins and can thus be applied to target proteins with unknown 3D structures such as G protein-coupled receptors (GPCRs). In addition, similarly to the MTS method, the DSI method can be combined with methods that correct scores to maximize the discovery rate of known compounds. As a result of applying the DSI method to a total of 14 target proteins including the proteins mentioned above and GPCRs and selecting the top 1 % of compounds predicted from the compound library, the discovery rate was improved approximately 70-fold on average compared with random screening[17].5 Degree of achievementWe have currently achieved more than 90 % of the initial objective. Our first compound DB was released in 2004 and immediately used for compound screening against TNF- converting enzyme. The MTS and DSI methods were applied using a protein-compound affinity matrix containing 182 proteins and 1 million compounds. Among 900 compounds subsequently purchased, 35 were found to be active compounds. The discovery rate was approximately 500-fold higher than the previously conducted screenings in which seven active compounds were obtained by randomly screening 100,000 compounds. In addition, no active compound was found after purchasing 700 compounds following screening by Glide, a commercially available software; hence, the discovery rate was dramatically improved by our methods. Since then, the compound DB has been annually renewed and the 2007 version is the latest. We have conducted direct screenings with respect to 10 target proteins over a period of six years and obtained active compounds with a probability range of a few to 20 %. This rate is several hundred to one thousand times higher than that achieved by random screening. Moreover, every year the compound DB and the protein-compound affinity matrix have been distributed to 10 to 20 institutions, primarily pharmaceutical companies, in Japan and overseas. The software and the compound DB have been partially released as myPresto[18] and LigandBox[19], respectively.6 Future workFirstly, our compound DB is not suited to screening of inhibitors of metalloproteinases containing metals such as zinc. The ion form of the molecules exhibits a predominant configuration under water; however, it will be different when the molecule binds to metals. For example, while a thiol (-SH) is normally configured as -SH under water, it is deprotonated and becomes -S- in the case of coordination with a metal. Changes in the ion form of molecules due to coordination with metals are observed in various functional groups. We found that the discovery rate strongly depends on the ion form of compounds through the VS of metalloproteinases. Accordingly, we plan to develop a compound DB for metalloproteinases.Secondly, our compound DB does not include inorganic compounds. Inorganic compounds such as metal complexes are considered to be unsuitable for drugs and are generally excluded from the compound DB. However, zinc complex was recently discovered to be an active compound with respect to insulin receptor protein, for which no active compound has previously been known except peptides, and this has attracted attention to inorganic compounds as novel therapeutic agents. The development of a DB for inorganic compounds is therefore necessary in order to examine the possible applications of inorganic compounds.Thirdly, distribution of our compound DB has depended solely on word-of-mouth publicity and it has not gained recognition by means of journal articles or websites. This is because our compound DB depends on catalog data provided by commercial firms. Catalog distribution is restricted to the marketing of reagents and advertisements of reagent vendors should be posted. For example, the free downloading of ZINC[20] was realized by posting advertisements of reagent vendors on university websites as a result of direct negotiations with reagent vendors. However, the advertising of private companies is prohibited at the National Institute of Advanced Industrial Science and Technology (AIST), and free downloading therefore cannot be realized. We are consequently distributing our compound DB on the assumption that AIST has compiled a database from catalogs that the users have independently obtained. It is also possible for incorporated associations, our collaborators who support -2.1-2.9-8.2-8.1-2.2-9.3-6.6-2.2-0.4-6.610-8.2-8.8-0.2-4.4-0.7-8.1-0.2-0.4-6.6-0.29-8.1-8.9-4.4-5.4-0.6-4.4-7.2-0.5-2.8-7.28-4.4-4.3-5.4-7.5-4.3-5.4-4.4-4.3-3.2-2.17-5.4-5.1-8.1-0.4-5.5-2.1-5.4-5.5-3.3-8.26-6.6-6.1-8.4-6.6-0.4-3.8-8.1-6.1-0.9-8.15-0.2-0.9-4.3-2.8-4.4-5.1-8.4-0.1-7.5-4.44-7.2-7.5-0.9-3.2-5.4-0.4-0.9-5.5-0.2-5.43-2.5-2.1-7.5-5.1-6.6-6.6-7.5-2.1-8.1-8.42-3.3-3.8-2.1-6.1-2.8-2.8-2.1-3.5-4.4-4.31-9.2-9.6-3.8-0.9-9.9-3.2-3.8-9.1-1.2-0.9TargetproteinKnownactivecompound987654321Compound databaseProtein groupHit compounds predicted by MTS methodHit compound predicted by DSI methodAnalogous compoundsFig. 5 Diagram of MTS and DSI methods. The numbers in the table indicate the scores. Higher scores are indicated by a deeper color.

※このページを正しく表示するにはFlashPlayer9以上が必要です