543
16 Cheminformatics: At the Crossroad of Eras
approach [29] to infer a compound’s toxicity from its closest chemical and biological analogues.
With several thousand nanotechnology-based consumer products available on
the market, nanotechnology is drawing worldwide attention for its applications in
various industrial areas, such as material science, medical research, and cosmetics
[30–32]. Importantly, a significant portion of these efforts is directed towards the
development of “green” products intended to achieve efficient and less polluting
energy sources [33, 34]. In this context, cheminformatics has a role to play by: (i)
facilitating the access, storage, search, and integration of all experimental results
currently distributed in literature, databases, and other sources; (ii) achieving externally predictive QSAR models to compute nanomaterials’ properties based on their
structural characteristics; and (iii) boosting the development and testing processes
by identifying the most promising nanomaterials that require focused experimental
investigations. The latter point is especially of importance due to the concerns about
the safety of certain nanomaterials and the development of nanomedicine [33, 35,
36]. In a recent proof-of-concept study, our research group introduced the terminology of Quantitative Nanostructure-Activity Relationships (QNARs) [37] that employs classical machine-learning methods for establishing links between chemical
descriptors and various measured activities of nanomaterials. Published studies and
developing trends in computational modeling of nanomaterials also have been summarized in a recent review [38].
With the forecasted, concomitant increase of both computing power and availability of additional chemical biological data, many new types of applications will
appear and expand in the coming years. Integrated meta-databases are likely to
emerge, combining structure-based, ligand-based, experimental and predicted properties, integrating the concepts and levels of systems chemical biology. Novel types
of chemical descriptors (especially at 2.5D, 3.5D and 4D levels) will be developed
to better take into account target–ligand interactions, notably molecular flexibility.
One can also speculate on the development of system fingerprints to characterize,
analyze, and compare complex chemical biological systems.
On a technical point of view, the performances of computer CPUs have improved as prognosticated by Moore’s law. However, the most time-consuming
calculations, such as molecular dynamics, quantum-chemical computation, and
molecular docking, must still be performed on supercomputers and large clusters
including thousands of CPUs in parallel. Interestingly, parallel calculations can also
be conducted on modern video cards, also called graphics processing units (GPU)
computing [39]. The most powerful individual workstations equipped with several
high-end GPUs can incorporate up to 10,000 CUDA cores ( e.g., 4-way SLI configuration with Nvidia’s Titan cards), allowing massively-parallel calculations as
long as the software code incorporates specific CUDA instructions. However, to
date, very few prototypes of cheminformatics software tools have been adapted to
take advantage of GPU-computing capabilities. GPU-computing could represent an
interesting solution to rapidly process, explore, and screen extremely large datasets
of 10
9
compounds and more.
16 Cheminformatics: At the Crossroad of Eras
approach [29] to infer a compound’s toxicity from its closest chemical and biological analogues.
With several thousand nanotechnology-based consumer products available on
the market, nanotechnology is drawing worldwide attention for its applications in
various industrial areas, such as material science, medical research, and cosmetics
[30–32]. Importantly, a significant portion of these efforts is directed towards the
development of “green” products intended to achieve efficient and less polluting
energy sources [33, 34]. In this context, cheminformatics has a role to play by: (i)
facilitating the access, storage, search, and integration of all experimental results
currently distributed in literature, databases, and other sources; (ii) achieving externally predictive QSAR models to compute nanomaterials’ properties based on their
structural characteristics; and (iii) boosting the development and testing processes
by identifying the most promising nanomaterials that require focused experimental
investigations. The latter point is especially of importance due to the concerns about
the safety of certain nanomaterials and the development of nanomedicine [33, 35,
36]. In a recent proof-of-concept study, our research group introduced the terminology of Quantitative Nanostructure-Activity Relationships (QNARs) [37] that employs classical machine-learning methods for establishing links between chemical
descriptors and various measured activities of nanomaterials. Published studies and
developing trends in computational modeling of nanomaterials also have been summarized in a recent review [38].
With the forecasted, concomitant increase of both computing power and availability of additional chemical biological data, many new types of applications will
appear and expand in the coming years. Integrated meta-databases are likely to
emerge, combining structure-based, ligand-based, experimental and predicted properties, integrating the concepts and levels of systems chemical biology. Novel types
of chemical descriptors (especially at 2.5D, 3.5D and 4D levels) will be developed
to better take into account target–ligand interactions, notably molecular flexibility.
One can also speculate on the development of system fingerprints to characterize,
analyze, and compare complex chemical biological systems.
On a technical point of view, the performances of computer CPUs have improved as prognosticated by Moore’s law. However, the most time-consuming
calculations, such as molecular dynamics, quantum-chemical computation, and
molecular docking, must still be performed on supercomputers and large clusters
including thousands of CPUs in parallel. Interestingly, parallel calculations can also
be conducted on modern video cards, also called graphics processing units (GPU)
computing [39]. The most powerful individual workstations equipped with several
high-end GPUs can incorporate up to 10,000 CUDA cores ( e.g., 4-way SLI configuration with Nvidia’s Titan cards), allowing massively-parallel calculations as
long as the software code incorporates specific CUDA instructions. However, to
date, very few prototypes of cheminformatics software tools have been adapted to
take advantage of GPU-computing capabilities. GPU-computing could represent an
interesting solution to rapidly process, explore, and screen extremely large datasets
of 10
9
compounds and more.