building material to start with and develop novel blocks with drug-
like properties. Next, developed blocks are screened for the initial
desired properties such as biological activity, pharmacokinetic prop-
erties at an initial stage of the project. Further, screened molecular
blocks with desirable properties are prolonged by either growing or
linking. A set of ligands with desirable properties can be obtained
by the iteration of the above steps. Hence frequently, the term de
novo or fragment-based drug designing are used interchangeably.
To avoid confusion, this chapter only uses the term de novo,
instead of using two separate names of similar approaches.
Computer aided de novo drug discovery process has comple-
mented the experimental combinatorial chemistry by providing
cost-effective and time saving approach to explore vast chemical
space. Cox-2, factor Xa, CDK4, carbonic anhydrase II, and HIV
protease are among few scientifically proven examples [5–9]. A de
novo molecule design software is challenged by virtually infinite
chemical search space to start with. The estimated number of
chemically feasible molecules is in order of 10^60 to 10^1000 for the
selection of promising candidates [10–12]. Despite boost in
computational power and advent of high-throughput screening
(HTS), exhaustive searching is not feasible for such a large chemical
space. Hence, the search in de novo design process focuses on
principle of local optimization rather than concentrating on global
optimization and does not systematically construct, evaluate, and
compare each and every individual compound. In such case, the
covered space is called “practical” optimum. Therefore, most of the
software work in nondeterministic way and rely on stochastic struc-
ture optimization. As chemists from different background will most
likely consider different molecules as “lead molecule,” similarly
repeated runs from same software that rely on stochastic optimiza-
tion will generate different lead molecules. Hence, it becomes
important to include as much chemical structure search space as
possible. There are two important aspects of a de novo software;
search algorithm and the scoring function. In a way, search algo-
rithms mimic a medicinal chemist and a scoring function analogi-
cally performs as an assay to evaluate the activity. Ideally,
computational approaches or in silico experiments provide routes
that facilitate identification of high quality molecular structures
with easiest possible synthetic process. The chances of identifying
“promising candidate” without getting lost into the vast chemical
space depend on two design strategies: positive design and the
negative design strategy. Positive design strategy limits the search
space within small region of molecules with high probable drug-like
molecules. Negative design defines the “unwanted” region by
identifying adverse properties and unwanted structures [13, 14].
Once a library of potential molecules is generated, the mole-
cules can further be engineered to avoid issues that prevent a
chemical series from delivering low quality molecules. Such124 Shashank P. Katiyar et al.