Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Chemical analysis
Reexamination Certificate
2001-02-28
2003-12-30
Brusca, John S. (Department: 1631)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Chemical analysis
C365S094000
Reexamination Certificate
active
06671627
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates generally to the field of combinatorial chemistry and more particularly to a method and computer program product for designing combinatorial arrays.
2. Related Art
Historically, drug discovery has been based on a serial and systematic modification of chemical structure guided by the “similar property principle”, i.e., the assumption that structurally similar compounds tend to exhibit similar physicochemical and biological properties. New therapeutic agents are typically generated by identifying a lead compound and creating variants of that compound in a systematic and directed fashion. The first phase of the process, known as lead generation, is carried out by random screening of large compound collections. Such compound selections may include, natural product libraries, corporate banks, etc. The second phase of the process, known as lead optimization, represents the rate-limiting step in drug discovery. This step involves the elaboration of sufficient structure-activity relationship (SAR) around a lead compound and the refinement of its pharmacological profile. Prior to the arrival of combinatorial chemistry, this process involved a simple prioritization of synthetic targets based on preexisting structure-activity data, synthetic feasibility, experience, and intuition.
Advances in synthetic and screening technology have recently enabled the simultaneous synthesis and biological evaluation of large chemical libraries containing hundreds to tens of thousands of compounds. With the expansion of the knowledge base of solid- and solution-phase chemistry and the continuous improvement of the underlying robotic hardware, combinatorial chemistry has moved beyond its traditional role as a source of compounds for mass screening and is now routinely employed in lead optimization and SAR refinement. This has led to the conceptual division of combinatorial libraries into (1) exploratory or universal libraries which are target-independent and are designed to span a wide range of physicochemical and structural characteristics and (2) focused or directed libraries which are biased toward a specific target, structural class, or known pharmacophore.
Two different methods are known for designing combinatorial experiments. The first is called “singles” or “sparse array” and refers to a subset of products that may or may not represent all possible combinations of a given set of reagents. The second is called a “full array” or simply “array” and represents all the products derived by combining a given subset of reagents in all possible combinations as prescribed by the reaction scheme.
The combinatorial nature of the two problems is vastly different. For singles, the number of possibilities that one has to consider (the number of different k-subsets of an n-set) is given by the binomial
C
s
=
n
!
(
n
-
k
)
!
k
!
(
1
)
In contrast, the number of different k
1
x k
2
x . . . k
R
arrays derived from an n
1
x n
2
x . . . n
R
R-component combinatorial library is given by
C
a
=
∏
i
=
1
R
n
i
!
(
n
i
-
k
i
)
!
k
i
!
(
2
)
For a 10×10 two-component combinatorial library, there are 10
25
different subsets of 25 compounds (singles) and only 63,504 different 5×5 arrays. For a 100×100 library and a 100/10×10 selection, those numbers increase to 10
241
and 10
26
for singles and arrays, respectively. Note that in this context, the term “array” is basically equivalent to reagent selection based on the properties of the products and does not necessarily refer to the physical layout and execution of the experiment. Although arrays are generally inferior in terms of meeting the design objectives, they require fewer reagents and are much easier to synthesize in practice.
Conventional methods for generating combinatorial arrays include reagent-based methods, and product-based methods based on stochastic sampling. Reagent-based methods examine each variation site independently of the other sites in the combinatorial library. A variation site is a point on a combinatorial core that allows the introduction of multiple building blocks into the combinatorial structure. For example, in a two-component combinatorial library, the design is carried out by selecting a list of reagents exhibiting certain desired properties from R
1
and a list of reagents exhibiting certain desired properties from R
2
. Then the selected reagents from R
1
are combined with the selected reagents from R
2
to produce a list of products, which constitute the combinatorial array. Although this method simplifies the selection process by examining each reagent pool independently of all the others, the properties of the products, which are of most concern, are never explicitly considered. Instead, this method is based on the hope that the selected reagents will result in a list of products that possess the desired properties.
The selection of reagents based on the properties of the products is conventionally accomplished using algorithms that are stochastic in nature. For example, in a two-component combinatorial library, a number of reagents from both R
1
and R
2
are randomly selected. The selected reagents from R
1
are combined with the selected reagents from R
2
. The resulting products are evaluated against some design objective. For example, if the design objective is to obtain products that are similar in nature to a known drug molecule, then the products are evaluated based on their similarity to that known drug molecule, resulting in some numerical value. Then one of the selected reagents at one of the variation sites is chosen at random and is replaced by another randomly chosen reagent from the pool of candidate reagents at that site. Better or worse results may occur. This process is repeated until some convergence criterion or time limit is met, for example until the resulting similarity values to the known drug molecule can no longer be improved. Thus, stochastic sampling methods require a significant amount of trials until a satisfactory solution is obtained.
What is needed is an algorithm for designing combinatorial arrays that capitalizes on the presence of optimal substructure when the objective function is decomposable or nearly decomposable to individual molecular contributions and allows the selection of optimal or nearly optimal arrays in an expedient fashion. What is further needed is a method for designing combinatorial arrays based on similarity to one or more reference compounds, or predicted activity and/or selectivity against one or more biological targets according to one or more QSAR, pharmacophore, or receptor binding models, degree of matching against one or more queries or probes, containment within certain property bounds, etc.
SUMMARY OF THE INVENTION
The present invention solves the above stated problem by providing a greedy method and computer program product for designing combinatorial arrays. The greedy method is particularly well suited in situations where the objective function is decomposable or nearly decomposable to individual molecular contributions. The invention makes use of a heuristic that allows the independent evaluation and ranking of candidate reagents in each variation site in a combinatorial library. The greedy method, when executed, is convergent and produces combinatorial arrays in an expedient manner. The combinatorial array solutions produced by the greedy method are comparable to, and often better than, those derived from the substantially more elaborate and computationally intensive stochastic sampling techniques. Typical examples of design objectives that are amendable to this approach include similarity to one or more reference compounds, predicted activity or selectivity according to one or more structure-activity or receptor binding models, degree of matching to one or more queries or probes, containment within certain molecular property bounds, and many others.
According to the greedy method of the present invention, an array of reagents from a combina
Agrafiotis Dimitris K.
Lobanov Victor S.
Salemme Francis Raymond
3-D Pharmaceuticals, Inc.
Brusca John S.
LandOfFree
Method and computer program product for designing... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and computer program product for designing..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and computer program product for designing... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3174055