540
D. Fourches
250,000 compounds against several panels of targets ( e.g., GPCRs, kinases). The
development of these facilities is leading to the public release of massive amounts
of chemical and biological data in the form of published materials and/or deposited
entries in online databases, which often mix raw and unstructured data with more
standardized and processed ones. After being curated and integrated, these very
large datasets can be considered for in-depth cheminformatics studies, including
selective hit fishing, SAR analysis, QSAR modeling, and study-dependent combinations of similarity searches, molecular docking, and pharmacophore analysis
for virtual screening and/or molecular design. One can also note that phenotypic
screens (both in vitro and in vivo) can be analyzed using cheminformatics techniques.
Screening is thus one of the most important driving forces for the field of chemogenomics, as it can ultimately allow the actual exploration of all possible small
molecule-target interactions [8]. For drug discovery purposes and side effect assessment and tuning, filling the gaps in this overall molecule-target interaction
matrix enables the analysis of compounds’ polypharmacology. The latter can be
at the origin of drug’s high efficacy and/or drug’s undesired effects caused by offtarget binding [9]. Although chemogenomics is developing and expanding based
on such revolutionary experimental platforms, assays, and protocols, it is clear that
computer-aided analysis and modeling will have an increasingly important role to
(i) curate, integrate, and exploit these huge amounts of data; (ii) develop predictive approaches for prognosticating the complex profile of multi-level activities for
novel molecules; and (iii) screen extremely large libraries of virtual compounds
using both ligand-based QSARs and structure-based scoring functions for better
prioritizing promising molecules to be synthesized and tested first.
This novel era of data profusion is drastically modifying the field of cheminformatics, from its most practical and technical aspects to its ultimate role of
chemical exploration and discovery. Utilizing chemical biological databases freely-available on the internet is becoming a common, everyday practice allowing
researchers to obtain very diverse information at a glance. This diverse information includes:
1. Three-dimensional structures of proteins using the Protein Data Bank (PDB,
http://www.rcsb.org/pdb/), which is very close to incorporating a stunning
100,000 entries,
2. Experimental activity measurements (mainly in the form of Ki and IC 50 values) using the ChEMBL database, which contains more than 1.5M compound
records: 9,350 biological targets and 734,000 assays (version 17, August 2013).
More than 31M compound entries are also available in PubChem [10], the largest repository of molecules and their associated activities against biological
endpoints, including HTS data deposited by pharmaceutical companies ( e.g.,
antimalaria screen of ~13,000 compounds by GSK),
3. Expert-verified structures (26M unique compounds available as of September
2013) and associated physico-chemical data using the Chemspider webportal [11];
a list of 2.7M purchasable compounds can be found in the ZINC database [12],
D. Fourches
250,000 compounds against several panels of targets ( e.g., GPCRs, kinases). The
development of these facilities is leading to the public release of massive amounts
of chemical and biological data in the form of published materials and/or deposited
entries in online databases, which often mix raw and unstructured data with more
standardized and processed ones. After being curated and integrated, these very
large datasets can be considered for in-depth cheminformatics studies, including
selective hit fishing, SAR analysis, QSAR modeling, and study-dependent combinations of similarity searches, molecular docking, and pharmacophore analysis
for virtual screening and/or molecular design. One can also note that phenotypic
screens (both in vitro and in vivo) can be analyzed using cheminformatics techniques.
Screening is thus one of the most important driving forces for the field of chemogenomics, as it can ultimately allow the actual exploration of all possible small
molecule-target interactions [8]. For drug discovery purposes and side effect assessment and tuning, filling the gaps in this overall molecule-target interaction
matrix enables the analysis of compounds’ polypharmacology. The latter can be
at the origin of drug’s high efficacy and/or drug’s undesired effects caused by offtarget binding [9]. Although chemogenomics is developing and expanding based
on such revolutionary experimental platforms, assays, and protocols, it is clear that
computer-aided analysis and modeling will have an increasingly important role to
(i) curate, integrate, and exploit these huge amounts of data; (ii) develop predictive approaches for prognosticating the complex profile of multi-level activities for
novel molecules; and (iii) screen extremely large libraries of virtual compounds
using both ligand-based QSARs and structure-based scoring functions for better
prioritizing promising molecules to be synthesized and tested first.
This novel era of data profusion is drastically modifying the field of cheminformatics, from its most practical and technical aspects to its ultimate role of
chemical exploration and discovery. Utilizing chemical biological databases freely-available on the internet is becoming a common, everyday practice allowing
researchers to obtain very diverse information at a glance. This diverse information includes:
1. Three-dimensional structures of proteins using the Protein Data Bank (PDB,
http://www.rcsb.org/pdb/), which is very close to incorporating a stunning
100,000 entries,
2. Experimental activity measurements (mainly in the form of Ki and IC 50 values) using the ChEMBL database, which contains more than 1.5M compound
records: 9,350 biological targets and 734,000 assays (version 17, August 2013).
More than 31M compound entries are also available in PubChem [10], the largest repository of molecules and their associated activities against biological
endpoints, including HTS data deposited by pharmaceutical companies ( e.g.,
antimalaria screen of ~13,000 compounds by GSK),
3. Expert-verified structures (26M unique compounds available as of September
2013) and associated physico-chemical data using the Chemspider webportal [11];
a list of 2.7M purchasable compounds can be found in the ZINC database [12],