FIELD OF INVENTION:
fhe present invention is in the realm of bio/cheminformatics and in relation to in-silico optimization of chemical compounds. More particularly, the present invention provides a method and a system for filtering compounds with undesirable interactions, which are obtained by doing a similarity search of the query against an interactions database.
BACKGROUND AND PRIOR ART OF THE INVENTION:
Running 'Natural Language Processing' engine on scientific literature can yield information on interactions between biological entities like proteins and small molecules. Such an approach, when run on Medline abstracts in December 2005 yielded around 231,400 Protein-Small molecule and 110,850 small molecule-small molecule interactions. Clearly, there is a plethora of information available for analysis. However, the nature of the search, which is 'text' driven, limits such an approach. What is of immensely more use is to run a 'substructure' search using the query compound of interest against the small molecule interactions database. The resulting hits can then be analyzed to check if the query compound has potentially similar biological interactions. This gains significance in a drug discovery setting wherein compounds are being virtually designed and optimized for good ADME properties. An additional dimension to optimize now could be avoiding undesirable interactions with specific biological targets or with other small molecules.
When new chemical entities (NCEs) are being designed and checked, during drug discovery, for various suitable ADME and Toxicity properties, the interactions of these NCEs with different targets is something that is not being looked at this point. However, checking this is important because there may be some undesirable interactions that are better avoided. Discovery of this aspect may or may not happen later in discovery pipeline, when experiments are run to screen for such undesirable interactions. Our invention presents an approach following which can help avoid undesirable interactions for compounds during the design stage itself The Application of instant invention is not possible to be envisaged by a person of average skill in the art. In fact, to devise this method and system it needs a cross-domain expertise from the areas of biology, chemistry, natural language processing and computer programming. Such high-end knowledge is very

difficult to acquire and deploy by a person or entity without considerable investment of efforts and resources. In its reduced form, the invention can simply be used as a search engine that combines Natural Language Processing with Substructure Search for efficient mining of scientific literature. The invention requires a pre-mapped 'interactions' database creation of which is complex and requiring timely updates in order to stay relevant. Perhaps a piece of technology that will allow creation of 'interactions map' or 'pathways' 'on-the-fly' will do away this requirement.
One may run experimental assays to determine if any undesirable interactions can happen between the NCE and given targets. However some limitations of this approach are:
a) Synthesis of assay costs in conducting these experiments
b) There is a high chance that what has not been explicity checked may be missed
Query Chem: a Google-powered web search combining text and chemical structures Justin Klekota, Frederick P. Roth and Stuart L. Schreiber Bioinformatics 2006 22(13): 1670-1673. In the aforementioned citation one needs to run a search engine like "QueryChem" with 'structure + keyword'. In addition, there is a need to predefine the keyword ~ results limited by what you define. The results displayed using this is not intuitive or user-Iriendly. Additional points on above citation are as follows:
a) In the aforementioned citation, the search is of a limited nature as defined by the keyword.
b) The display of results is not so intuitive and further refinement of search somewhat unwieldy.
c) Interaction results are defined and limited by the keyword set by the user. One has to run the search again with a different keyword to redefine the search. Thus, analysis may be limited by user's imagination.
d) Derives its interactions from internet in general which are liable not to be so accurate.
Therefore, the aforementioned flaws are addressed by the present invention.
Also, Run experimental assays to determine if any undesirable interactions may happen
between the NCE and given targets

a) Time and infrastructure costs in conducting these assays.
b) What has not been checked may be missed?
In addition, our invention provides further refinement or exploration on these results is unwieldy i. Cover as many biological interactions as currently available in literature
ii. Show results in a user-friendly and intuitive manner
iii. Allow further refinements in search and exploration
iv. Avoid time and infrastructure costs.
a) This invention presents an approach following which can help avoid undesirable interactions of compounds during the design stage itself The whole process is computational and hence avoids time and infrastructure costs.
b) Also, this approach is more comprehensive and refined in detecting undesirable interactions and hence superior to the experimental asssays.
c) The results are displayed in an intuitive and user-friendly format and allows for further refinement and redefinition of search in an easy manner.
d) This system shows 'all' interactions and allow filtering down based on various measures like 'relevance interactions' (binding, transcription, post-translational, small molecules, metabolism or transport regulation etc.), 'interaction networks' (shortest path network, network regulators, network targets etc.), 'advanced analysis' (relevance list, custom relevance interactions, custom interaction network etc.), 'enrichment analysis' (GO group enrichment, similar pathways etc.), 'numerical data analysis' etc.
e) Our 'interactions' database is built from Pubmed (http://www.ncbi.nlm.nih.gov/entrez/querv.fcgi?DB=pubmed) which is a scientifically valid and reputed resource wherein almost all biological research (abstract) gets reported and archived.
OBJECTS OF THE INVENTION:
The principal object of the present invention is to develop a method for in-silico optimization of drug like compounds.

Another object of the present invention is to develop a system for in-silico optimization of
drug like compounds.
Yet another object of the present invention is to map the pathway interactions of the target
compounds.
Still another object of the present invention is to identify undesirable interactions and
thereby obtaining optimized chemical compound.
STATEMENT OF THE PRESENT INVENTION:
Accordingly, the present invention provides A method for in-silico optimization of chemical compounds preferably drug-like compounds, wherein said method comprising acts of: developing chemical structure for a query compound; performing similarity search against the query compound in a predetermined pathway interaction database; defining searched target compounds found similar to the query compounds; and mapping pathway interactions of the target compounds to identify undesirable interactions as hit structures and thereby obtaining optimized chemical compound; and a system for in-silico optimization of chemical compounds preferably drug like compounds, said system comprises structure editor to draw or edit chemical structure; means for similarity search to identify undesirable interactions; and interaction database to store compound names, structure of the compounds and pre-mapped pathway interactions
BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS
Figure: 1 Snap shot of structure editor to draw and edit query structure
Figure: 2 Snap shot for a similarity search function
Figure: 3 snap shot showing number of matching structures for query compound
Figure: 4 Snap shot showing the matching structures
Figure: 5 Snap shot showing pathway map view for hit structures
Figure: 6 Snap shot showing list of interactions for each hit structure
Figure: 7 Snap shot showing each entity participating in the interactions
Figure: 8 Flow chart to understand the present invention and its linkage with external
sources
Figure: 9 Block diagram of system of instant invention

Figure: 10 matching structure of query compound
Figure: 11 Snap shot showing the detailed interaction network of Bupropion
Figure: 12 Norepinephrine uptake - Bupropion
DETAILED DESCRIPTION OF THE PRESENT INVENTION
The present invention is in relation to a method for in-silico optimization of chemical compounds preferably drug-like compounds, wherein said method comprising acts of:
a. developing chemical structure for a query compound;
b. performing similarity search against the query compound in a predetermined
pathway interaction database;
c. defining searched target compounds found similar to the query compounds; and
d. mapping pathway interactions of the target compounds to identify undesirable
interactions as hit structures and thereby obtaining optimized chemical
compound.
In another embodiment of the present invention said similarity search is performed between query compound of interest against the molecule interactions database which comprises common names, lUPAC names, structures of the compounds, pre-mapped pathway interactions and user defined pre-mapped chemical interactions.
In another embodiment of the present invention the method enables filtering of interactions based on various measures selected from a group comprising relevance interactions such as binding, transcription, post-translational, small molecules, metabolism or transport regulation, interaction networks such as shortest path network, network regulators, network targets , advanced analysis such as relevance list, custom relevance interactions, custom interaction network, enrichment analysis such as GO group enrichment, similar pathways and numerical data analysis and combinations thereof.
In another embodiment of the present invention the similarity search comprises method(s) to measure similarity of the query structure to the list of target structures.

In another embodiment of the present invention said method is selected from a group comprising a sub structure search method to identify target structures of which the query structure is a sub structure or vice versa; coefficient based threshold method on structural similarity and metric based threshold method on descriptor similarity.
In another embodiment of the present invention the interactions listed in the interactions database for hit structures are displayed in the pathway map view to enable exploration of information on the basis of interactions present in the interaction database with detailed information about the interactions for each of the hit structures and additional interactions for each entity participating in the interactions.
In another embodiment of the present invention the compounds are virtually designed and optimized for good absorption, distribution, metabolism excretion and to be free of toxic properties and the undesirable interactions are avoided by suitably redesigning the compound.
The present invention is in relation to a system for in-silico optimization of chemical compounds preferably drug like compounds, said system comprises
a. structure editor to draw or edit chemical structure;
b. means for similarity search to identify undesirable interactions; and
c. interaction database to store compound names, structure of the compounds and pre-
mapped pathway interactions.
In another embodiment of the present invention the compounds are virtually designed and optimized for good absorption, distribution, metabolism excretion and to be free of toxic properties.
The primary embodiment of the present invention is a method for in-silico optimization of chemical compounds preferably drug-like compounds, wherein said method comprising steps of: a) drawing chemical structure for a query compound;

b) performing similarity search against the query compound in an pathway interaction database;
c) defining searched target compounds found similar to the query compounds; and
d) mapping pathway interactions of the target compounds to identify undesirable interactions as hit structures and thereby obtaining optimized chemical compound.
In another embodiment of the present invention, wherein said similarity search is performed between query compound of interest against the molecule interactions database which comprises common names, lUPAC names, structures of the compounds, pre-mapped pathway interactions and user defined pre-mapped chemical interactions.
In yet another embodiment of the present invention, wherein the method enables filtering of interactions based on various measures selected from a group comprising relevance interactions such as binding, transcription, post-translational, small molecules, metabolism or transport regulation, interaction networks such as shortest path network, network regulators, network targets , advanced analysis such as relevance list, custom relevance interactions, custom interaction network, enrichment analysis such as GO group enrichment, similar pathways and numerical data analysis and combinations thereof
In still another embodiment of the present invention, wherein the similarity search comprises method(s) to measure similarity of the query structure to the list of target structures.
In still another embodiment of the present invention, wherein said method is selected from a group comprising a sub structure search method to identify target structures of which the query structure is a sub structure or vice versa; coefficient based threshold method on structural similarity and metric based threshold method on descriptor similarity.
In still another embodiment of the present invention, wherein the interactions listed in the interactions database for hit structures are displayed in the pathway map view to enable exploration of information on the basis of interactions present in the interaction database

with detailed information about the interactions for each of the hit structures and additional interactions for each entity participating in the interactions.
In still another embodiment of the present invention, wherein the compounds are virtually designed and optimized for good absorption, distribution, metabolism excretion and to be free of toxic properties and the undesirable interactions are avoided by suitably redesigning the compound.
The present invention is in relation to a system for in-silico optimization of chemical compounds preferably drug like compounds, said system comprises
a. structure editor to draw or edit chemical structure;
b. means for similarity search to identify undesirable interactions; and
c. interaction database to store compound names, structure of the compounds
and pre-mapped pathway interactions
In another embodiment of the present invention, wherein the compounds are virtually designed and optimized for good absorption, distribution, metabolism excretion and to be free of toxic properties.
The core idea of the invention is to 'draw' the structure of a compound ('query'), run a 'similarity' search against compounds in an 'interactions' database, define 'target' compounds 'similar' to the 'query' compound, and then check the interactions of these 'target' compounds.
This will now allow checking if the compound of interest potentially has any undesirable interactions (that should possibly be avoided by suitably redesigning the compound), i. Save time an infrastructure costs of running experimental assays ii. Check many more interactions compared to few checked in experimental assays iii. User-friendly and intuitive display of results iv. Allows further search and exploration in a dynamic manner
rhere is no technology that allows you to achieve what has been proposed here, at the proposed level of details, robustness and user-friendliness.

♦ In experiment-based assays, only few interactions get checked although results are of experimental nature and perhaps of more value. Our approach allows checking of potentially many more interactions though it is based on computational methods.
♦ In the software utilities mentioned above (see answer to question# 3), esp. QueryChem, the search is of a limited nature, the display of results not so intuitive and further refinement of search somewhat unwieldy. Our approach allows for an extensive search of interactions, throws up much more information, displays the results in an intuitive and user-friendly format, and allows for further refinement and redefinition of search in an easy manner.
*X* In the proposed invention, structures (provided by user) are used to search against structures (present in a database), whose interactions have already been kept pre-mapped in an interactions database. Once 'target' structures 'similar' to 'query' structures have been found, it is just a matter of displaying the interactions of these 'target' structures.
♦ In a utility like QueryChem, structures are first searched against structures available in public databases. The 'text' names of the 'hits' so obtained are then combined with user defined keywords and again used to search information from the internet.
♦ We show 'air interactions and allow filtering down based on various measures like 'relevance interactions' (binding, transcription, post-translational, small molecules, metabolism or transport regulation etc.), 'interaction networks' (shortest path network, network regulators, network targets etc.), 'advanced analysis' (relevance list, custom relevance interactions, custom interaction network etc.), 'enrichment analysis' (GO group enrichment, similar pathways etc.), 'numerical data analysis' etc.
In a utility like QueryChem, interaction results are defined and limited by the keyword set by the user. One has to run the search again with a different keyword to redefine the search. Thus, analysis may be limited by user's imagination.
♦ We have built our 'interactions' database from PubMed
(http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?DB=pubmed) which is a scientifically valid and reputed resource wherein almost all biological research (abstract) gets reported and archived.
10

♦ A utility like QueryChem, derives its interactions from internet in general which are
liable not to be so accurate. The highlights of the invention are provided below in point wise manner.
i. The ability of the invention to 'virtually' draw a compound, run a 'similarity' search against compounds in an 'interactions' database, check if the 'hit' compounds have any 'undesirable' interactions, which should perhaps be avoided by redesigning the compound (and going through this iteration cycle).
ii. Its significance in a drug discovery setting wherein compounds are being virtually designed and optimized for good ADME/Tox properties. An additional dimension to now optimize could be avoiding undesirable interactions, say, with specific biological targets or with other small molecules, iii. Conversion of molecule drawn in the editor to smiles, energy minimization using a force field, computation of structural descriptors and predictive models are run with these quantitative descriptors. Predictive model end points along with their confidence measures are updated, iv. A database ('interactions database') exists wherein you have:
1. Compound names (common names, lUPAC names etc.)
2. Structures for these compounds
3. Pre-mapped pathways where the interactions of these compounds are given, v. A 'structure editor' exists that allows you to draw a structure ('query structure').
vi. A 'similarity search function' exists that allows you to run a query against given structures ('target structures') present in a:
1. spreadsheet
2. an external database
3. 'backend' database
vii. The 'similarity search function' could use one or more of the following methods to measure similarity of the 'query structure' to the list of'target structures':
1. Sub-structure search method to identify 'target structures' of which the 'query structure' is a sub-structure (or fragment) of, or vice versa
2. Tanimoto (or other) coefficient based threshold on structural similarity
11

3. Euclidean (or other) metric based threshold on descriptor similarity
4. The 'target structures' found similar to the 'query structure' are then filtered ('hit structures') and displayed.
viii. The 'pathway map' view would allow exploration of the following information:
1. For each of the 'hit structures' a list of interactions present in the 'interactions database', with detailed information about these interactions;
2. For each entity participating in the above interactions, detailed information about it;
3. For each entity participating in the above interactions, an additional list of interactions with which this entity further interacts with.
The invention requires a premapped 'interactions' database.
The system will only be applicable for drug-like chemical molecules and not for proteins
and peptides, neither for biological entities.
The 'search' will be a structure based search, based on either 2D or 3D features.
'Similarity Search Function' is performed using query structure(s) on any dataset or (external) database. The i^earch can be performed on the basis of the 2D structural information of the query (Structural or fingerprint based similarity) or using some properties of the query which are elucidated as descriptors (Descriptor similarity). It essentially involves the following steps:
1) Draw query structure(s) using the structure editor available in the proposed system.
Alternately, open saved structure(s). This becomes the query dataset.
2) Fingerprint based similarity:
i) The basis for structure similarity are 2D MACCS fingerprints numbering 166
(reference cited below). These are represented as 'bits', ii) Choose the dataset/database against which the similarity search has to be
carried out. iii) Define the similarity threshold in terms of Tanimoto coefficient. Tanimoto
coefficient is a pair-wise similarity measure. It is a ratio of the 'bits' common
to a pair of compounds to the 'bits' which are different. (Reference 1). In
12

chemical world, compounds with Tanimoto coefficient of 0.85 or above are considered to be similar. A value of T indicates that the compounds are identical considering the 2D structural fingerprints. 3) Descriptor based similarity: i. Euclidean: This is the standard sum of squared distance (L2-norm) between two objects.
^^.-y,'
ii. Choose the dataset/database against which similarity search has to be carried out. iii. Define the threshold in terms of Euclidean distance (Reference 2). A value of 0.95
or more indicates good similarity. 3) The search runs and creates a subset containing only those compounds in the dataset/database which met the threshold criteria.
The substructure search involves
i. Draw a query structure using the structure editor or open a structure, ii. Choose a dataset or database on which the substructure search has to be carried out. iii. The search looks for the presence of the query in the target dataset / database. It
involves accepted heuristics and works using molecular graphs iv. The search finishes and creates a subset with 'Hit' compounds, which are
compounds which contain the 'query' structure.
a. list of interactions in the interaction database
The pathway interactions database is a storehouse with "read and write permissions" of interactions information between all entities in the database.Interactions depicted between any two or more entities in the map (either custom created by interpreting associated published literature or already derived using natural language processing from published literature) are also stored in the pathway interactions database. Information about any interaction can be retrieved by a single click from the database if a PathwayArchitechTM. User Interface is used view the map, in which it is created.
13

b. entity participation in the interactions
Any pathway map created using PathwayArchitectTM automatically writes all participating entities into the pathway database. Information about any entity in the map can be retrieved by a single click from the database if a PathwayArchitechTM User Interface is used view the map, in which it is created.
c. Additional list of interactions for entity participations as mentioned above.
There may be many molecular interactions for any entity from our map. If the participating interaction is relevant to our map it is depicted in our map. If the interaction is not relevant to our map, the information exists in the database, but is not included in the map. Using a Pathway Architect UI to view our map, for an given entity, one may be able to retrieve many more interacting entities using various algorithms in the tool.
f he technology of the instant Application is further elaborated with the help of following examples. However, the examples should not be construed to limit the scope of the invention. Example: 1
A database ('interactions database') exists wherein you have:
• Compound names (common names, lUPAC names etc.)
• Structures for these compounds
• Pre-mapped pathways where the interactions of these compounds are given.
The details involved in the database are provided in Table: 1 as below:
Tablel: Schema of the database that is a prerequisite for this study. (Source of interactions: www.rxlist.com)

Compound name

Structure

Pathway Interactions

Acetaminophen

~0-.^=",

Action on the hypothalamic heat regulating center

14


Acebutolol

H,C

a cardioselective, beta-adrenoreceptor blocking agent; antagonistic effects on peripheral vascular B2-receptors;

N , Nil
Bupropion
friamterene

inhibitor of the neuronal uptake of norepinephrine, serotonin, and dopamine, and does not inhibit monoamine oxidase.
inhibits the reabsorption of sodium ions in exchange for potassium and hydrogen ions at that segment of the distal tubule

Example: 2
A 'structure editor' exists that allows you to draw a structure ('query structure') is provided in Figure: I as a snap shot. In addition. Figure: 2 provide a snap shot for a 'similarity search function' that allows you to run a query against given structures ('target structures') present in a:
i. Spreadsheet; ii. an external database; and iii. 'backend' database. The 'similarity search function' could use one or more of the following methods to measure similarity of the 'query structure' to the list of'target structures':
i. Sub-structure search method to identify 'target structures' of which the 'query
structure' is a sub-structure (or fragment) of, or vice versa ii. Tanimoto (or other) coefficient based threshold on structural similarity iii. Euclidean (or other) metric based threshold on descriptor similarity The 'target structures' found similar to the 'query structure' are then filtered ('hit structures') and displayed which are as shown in Figures: 3 and 4.
15

The interactions listed in the 'interactions database' for the above 'hit structures' are then
depicted in the 'pathway map' view which is as shown in Figures: 5, 6 and 7.
The 'pathway map' view would allow exploration of the following information:
i. For each of the 'hit structures' a list of interactions present in the 'interactions
database', with detailed information about these interactions; ii. For each entity participating in the above interactions, detailed information about it; iii. For each entity participating in the above interactions, an additional list of interactions
with which this entity further interacts with.
P2xample; 3
To illustrate the novelty/inventiveness of instant method, the Applicant has used a query structure to search the database and get the 'hit' structure; subsequently, pulled out all the interactions of the hit structure from the interactions database. One of the hit was a drug by name 'Bupropion'.

BUPRIOPION
All its interactions are highlighted in the interactions database thus providing the help to eliminate 'undesirable interactions' of the original query structure. The details of the interactions are shown in figures: 10, 11 & 12.
We duplicated this experiment in 'Query chem' tool and found that it was not possible to get 'Bupropion' as a 'hit' structure using its databank. Hence the whole information was lost or rather 'not detected' at all in this method.
16

Details of the experiment: Structure used as a 'query':
Query Cheiiil
HOI.IE''oi(i|)kt>.tfILuJ;ui[lv
r^KunilaHciiaiMiuKkfaimd
Only resiits 1 -10 displayed on tks page
Si]bjtni[rieiiiCNriC)Ci=Q)clccccr!: N-methylcalhione j!ii);tiTjcr.ireB-iNC?iiriCj=0)cl[ccccliYnuraniMe SnbilnittirmCCCCiH^CrOickccsl. 53146 M;tniftrei!ii:CiHlCl-0:ckcciO]::cl: 23618 ;ijbitiijitoaiCCil^C(=0:clcc[|C)cel: 60487 £iiKtijctitfflCCiC);l-ICijCK')cli:i:ci:i:l:2-(liy[ko)!yaiMo)-2-iiielh^^ aib;tnict,ire m C[C!p]lH)C(=0|[lc:i.ccl: 2-amio-l-pheni^propan-l-one jit;tni:rie m CCC!]TiCi=0'![ici:ciCl)[:!. (2E]-2-araino-l-(4-clilorophenyl)bBtan-l-one S:!t-;&iifto8i'?N''!'g!HiJl)0-u!(liccal 2-(ffletli^amio)-l-pliaiylpfopaii-l-oiie SiA'itiUitic >L ¥.! CH-HC i CK' 'iilimC [(3R,4S)-4-bfomo-4>iiydfo-3H-pyrazol-3-yl]i
QnEEIDCOMPOM);CC(N)C(=0)clcccccl
' - '■■ ' '■CHaOCKilcktcccl 'kidi (I'lfiv uiliibitoi "N-iiiefliv!?;iflimoiie" OR "moiioiMtiiylpiopioir OR "inetliyhtiiiiioM" OR "iMiiKKtliiiioM" OR "ephsdiouf" OR "alplis-N-inetliylaiiiiiiopiopioplieiioM" OR ' alpliHa«li\iiiiiiiiopiopicpheuoiie" OR "alpliHiieWMimio-piopioplmioiie'MDR'i^ ■ ■^m^tllvla||^lllo4-l)kll^^plopall-l-ollf■■ OR'l-plopallOll^^metLylalnBll)■l-pllellyl'■ OR ISO-mett^ 1 ClOHlMl (l-8ill-Ill0ii: W4:v?---9 1IV8.UH,1-:H5 tS- inO si"
Oooik ^'liok uJukroi "N-ineWnikioiie" OR"inouoaffclpiopiou" OR"iMtW^luiioiie" OR"inetli(adimoiie" OR "fplipdiwie" OR"alpLs-H-mftlivlMnjiiopiopioplifnoM" OR " iilpliMnetliylaitimopiopiopheiiouf ■■ OR" ^lplia-m«li\1aiiuiio-piopioplinionf" OR "I-inetliyknmopiopioplieii(iiif" OR "2-iaetliykiimo-piopioplmioiie" OR ■■]-iMilivl;iMio-l-pheii\1piopa-l-oue" OR "l-piopDiioiie-l-infilivlainmo-l-plieuyl" OR "lSO-inffl\ylnirfflio-l-plieiiylpiop.iu-l-oiie" OR
i (mm) 'i-siij-:iioij:i9-6-4-v5---