ADMEWORKS MODEL BUILDER STANDARD EDITION (Latest version) WORK FLOW

UPLOAD SDF FILE WITH PARAMETER OF INTEREST

SELECT CONCERNED

DESCRIPTOR

CALCULATE DESCRIPTOR

CREATE ADMEWORK MODEL

APPLY THE ABOVE MODEL TO QUERY DATA

QUERY PARAMETER PREDICTED

ADMEWORK

CALCULATIONS REPORT TO PREDICT THE TOXICITY OF BELOW MOLECULE

COOCH 2CH3

H 3C

N H

SCH3

ethyl 6-methyl-2-(methylthio)-4-phenyl-1,4-dihydropyrimidine-5-carboxylate

(QUERY DATA)

The above molecule was drawn by ChemBioDraw (ChemBridgeSoft product) & saved in the SDF file format. Further molecule will be used for toxicity prediction based on descriptor selected during model building process.
BEFORE PERFORMING THE CALCULATIONS, POINTS TAKEN INTO THE CONSIDERATIONS.

1)

PREPARE THE CHEMICAL FILE IN SDF FILE FORMAT THAT WILL BE USED FOR CREATING PREDICTIVE MODEL FOR EXAMPLE:- IN OUR CASE WE WANT TO CREATE A MODEL WHICH WILL PREDICT THE TOXICITY OF ANOTHER COMPOUND WITH UNKOWN TOXICITY. SO THE FIRST THING WE NEED A TRAINING SAMPLE FILE CREATED IN SDF FILE FORMAT WITH THE DETAIL OF PARAMETER WE ARE INTERESTED TO PREDICT. SDF File(jurs_3D.sdf, in our case) contain following details:A) Chemical molecule list in one column. B) Parameter with known values (in our case toxicity) in another column.

2) SELECT THE IMPORTANT FEATURES (DESCRIPTORS) ON THE BASIS OF WHICH WE WILL PREDICT
THAT PARTICULAR PARAMETERS (TOXICITY),IN OUR CASE WE ARE USING DESCRIPTORS LIKE:a) ATOM COUNT b) LEADLIKENESS c) FQLOGP d) FQLOGS e)TPSA etc.

NOTE:- Feature(Descriptor) extraction, whose purpose is to reduce the data by measuring certain features or properties. These features (or, more precisely, the values of these features) are then passed to a prediction model that evaluates the evidence presented and makes a final decision as to the query chemical. The models are based on values of physicochemical, topological, geometrical, and electronic parameters derived from the molecular structure - called descriptors. ADME MODEL BUILDER has list of descriptor out of which select the important descriptor which will be used to predict that particular chemical property of query chemical. FURTHER, CALCULATE DESCRIPTORS USING CALCULATE TAB.

Prediction of Toxicity Rational Approach OPEN ADME WORK MODEL BUILDER

LOAD SDF FILE IN ADMEWORK MODEL BUILDER Step 1 Loading the data and classifying the molecules 1) Go to File | Import | SDF and locate the jurs_3D.sdf file

SELECT YOUR SDF FILE AND OPEN IT (for ex:-jurs_3D.sdf)

THE SELECTED SDF FILE SUCCESSFULLY LOADED INTO THE ADMEWORK MODELBUILDER:There should be 322 molecules loaded. F

Select the Sample | Select Numerical Target menu option. Perform operations as prompted on the SELECT TOXICITY & PRESS OK BUTTON.

Double click on the Toxicity column to sort the samples in ascending order. In this exercise we will start with a hypothesis that the toxicity of the samples is correlated to their quantum properties, the solubility in water and the presence of some fragments that may have toxic effects. a) Locate the ATOM_COUNT branch on the All descriptors tab and choose this set of descriptors. b) In the same tab locate LEADLIKENESS, OPERABLE RULE, TPSA, FQLOGS, FQLOGS & MLOGP branch and select it. c) Go to the Substructure descriptors by descriptors tab and choose SC. Next click OK. 351 descriptors are loaded in to the worksheet. d) Go to Descriptors| Calculate and wait until all descriptors are calculated.

ADD DESCRIPTORS:-

DESCRIPTOR SELECTED:-

CALCULATE DESCRIPTOR:Go to Descriptors| Calculate and wait until all descriptors are calculated

DESCRIPTOR GENERATED WE CAN SEE HERE ADDITONAL COLUMNS IN MODEL BUILDER WORK SHEET. Feature Selection 1) Go to Feature Selection | Basic | Zero test, set the % non zero values parameter to 1%. Next click OK. The change in the% non zero values parameter is caused by the fact that some substructures may be very rare in the training set, but there influence may be important.

Go to Feature Selection | Basic | Automated correlation test. Leave the default settings and click OK.

SCAT TE-R PLOT Absolute of Residues Plot

Displays statistical data for current temporary model

From the Model | Create | Regression menu choose Interactive LMR. Use this model to analyze the influence of the descriptors and the sample outliers.

R2 Hi Pl sto ot ry

Model Parameters

FINALLY CLICK ON OK BUTTON IN ABOVE MODEL.

The plots (Scatter Plot and Absolute of Residues Plot) show how this model works when applied to your sample set. Position the cursor over the points and lines on the plots (do not click!). The structures of molecules corresponding to each point or bar on the plots will be displayed. History plot will help us in retrieving the best model.

PREDICTION COLUMN GENERATED WITH PREDICTED VALUES.

RIGHT CLICK ON PREDICTION COLUMN & SELECT CROSS VALIDATION METHOD.

The attempt to access cross validation (or cross validation error) will result in displaying a dialog asking for the nvalue. Leave N-out : N-out parameter for Cross Validation. This is the number of samples excluded from the training set while building each of the validation models. For example: for a 50 sample set setting N to 5 will cause 10 models to be created during the validation calculation. Split N-fold : N-fold parameter for Cross Validation. This is the number of parts that the training set will be divided into. In fact, it is also the number of models to be created during validation.

A NEW CROSS VALIDATION COLUMN GENERATED

If you want to see the model equation's parameters and statistical data, go to the Model | Parameters menu option. The following dialog will be displayed:

Saving and Later Use of Created Model :-

IMPORTING MODEL DATA FROM AN .XML FILE.

2) THE MODEL WILL APPEAR IN A BLANK NEW WORKSHEET. NOW LOAD YOUR QWERY MOLECULE IN OUR CASE:-

COOCH 2CH3

H 3C

N H

SCH3

ethyl 6-methyl-2-(methylthio)-4-phenyl-1,4-dihydropyrimidine-5-carboxylate
(QUERY MOLECULE) THE ABOVE MOLECULE IS SAVED IN "SDF" stands for structure-data file FORMAT. LOAD THIS MOLECULE IN ABOVE MODEL AND CLICK ON APPLY ALL.

Load some SDF data (e.g. 6ethyl-6-methyl-2-(methyltjio)4-phenyl-1,4-dihydropyrimidine-5-carboxylate.sdf), and estimate the toxicity of the molecules by activating the Model | Apply All menu option.

THE PREDICTION COLUMN IS FILLED WITH PREDICTED VALUES FOR THE QUERY MOLECULE SET.

CONCLUSION:-

In the above calculations, we tried to demonstrate how to calculate the toxicity of new chemical on the basis of model generated in ADMEWORK Model Builder using :1) ATOM COUNT 2) LEADLIKENESS 3) TPSA 4) FQLOGP etc as a descriptor. The important point to remember while predicting any chemical properties of new molecule is dependent on the type of descriptor selected and is chemical relation with queried sample properties.

For Further Details & Queries, Contact: SCUBE Scientific Software Solutions (P) Ltd, 613, (6th Floor), Hemkunt Chambers, 89, Nehru Place, New Delhi - 110019. Phone: 011-41618828/29 Email: info@scubeindia.com Telefax: 011-41618828