QSAR

10       Performing Molecular Shape Analysis

Molecular shape analysis (MSA) is a formal approach to incorporating conformational flexibility and shape data into a QSAR. QSARs containing this type of data are commonly called 3D QSARs. The term molecular shape analysis applies to the process described by Hopfinger and Burke (1990)for generating 3D QSARs.

This chapter describes

This chapter provides an overview of MSA in the Drug Discovery Workbench:

Accessing molecular shape analysis (this page)

Overview of molecular shape analysis (page 178)

1. Generating conformations (page 182)

2. Hypothesizing an active conformer (page 186)

3. Identifying a shape reference compound (page 187)

4. Aligning molecules (page 190)

6. Determining other molecular features (page 193)

7. Generating a trial QSAR (page 193)

This part of the chapter directs you to appropriate chapters in the Cerius2 books where more detailed information about each activity can be found.

Before you begin

Before you start MSA, you must have a properly licensed copy of the appropriate Cerius2 software, including a copy of QSAR+, installed on your system. If you have any questions about your system setup or your software license, please talk to your system administrator.

You should be familiar with the Cerius2 interface and tools before you begin using MSA. For information on the Cerius2 environment, consult the manuals listed in the preface, How To Use This Book.

Additionally, you must be familiar with the information presented in Chapter 14, Using the Equation Viewer.

To start MSA

Go to the QSAR deck of cards and choose the SHAPE ANALYSIS (MSA) card from the deck.

Overview of molecular shape analysis

MSA is an iterative process where steps may be repeated until the molecular shape similarities and other descriptors are checked and adjusted to generate a QSAR equation with optimal statistical significance.

The goal of MSA is to generate a QSAR equation that incorporates spatial molecular similarity data. The process, as described by Hopfinger and Burke, involves seven tasks illustrated in the following figure:

Figure 6 . MSA process

The outcome of the MSA process is an optimized QSAR that can be used for activity estimation and ligand evaluation. The set of choices available for each task is employed to generate trial QSARs. The QSAR that corresponds to the best fit between observed activities and computed molecular descriptors defines the specific requirements for each MSA task.

MSA tasks

The tasks in the MSA process are:

1.   Generate conformers -- The purpose of this task is to generate and analyze conformers for each structure to be investigated, then to reduce the number of conformers to those that are likely to be relevant to biological activity. You can do the analysis automatically as structures are loaded into the study table or at any time after structures are loaded.

In Cerius2, conformation generation can be done in several ways, primarily depending on the size of the molecule and the set of rotatable torsion angles involved. Please see the section 1. Generating conformations on page 182, and Cerius2 Conformational Search and Analysis.

2.   Hypothesize an active conformer -- This part of the process generates a structure that corresponds to the structure present in the rate-limiting step controlling biological action. Typically, this step involves ligand-receptor binding, but it may involve metabolic activation or deactivation, membrane transport, or generation of a transition state.

MSA was developed to treat QSARs in which the geometry of the receptor is unknown. All information about the active conformer must be gleaned from observed biological activities and from corresponding intramolecular conformational propertiescomputed for the ligands. If X-ray binding data are available, they can be used to specify active conformations.

MSA provides a variety of methods for identifying possible active conformers. Please see the section 2. Hypothesizing an active conformer on page 186.

3.   Select a candidate shape reference compound -- The shape reference compound is the molecule that is used when shape descriptors are calculated for the study table. To select the reference compound, each compound in the data set is tested in one or more possible active conformations.

MSA compares all other molecules in the study table to the shape reference compound and provides information about each comparison. The criterion for selecting the shape reference compound is to optimize the statistical significance of the corresponding QSAR. Please see the section 3. Identifying a shape reference compound on page 187.

4.   Perform pair-wise molecular shape superpositions -- MSA requires that each compound in the range of active candidate conformations be aligned and compared with the shape reference compound. The fourth step in MSA is to perform pair- wise molecular superpositions to determine what and how atoms of data set compounds are equivalent to atoms in the shape reference compound.

MSA provides several methods to align entire structures, as well as to select certain atoms to be aligned. Please see the section 4. Aligning molecules on page 190.

5.   Measure molecular shape commonality -- Calculate shape descriptors to compare the properties that two molecules have in common and to measure molecular shape commonality.

Briefly, to select and calculate shape descriptors:

a.   Choose the Descriptors/Select... menu item in the study table to select descriptors with the Descriptors control panel.

b.   Set the Descriptor family popup to MSA and then click the button on the extreme left side. The MSA descriptors are displayed in the table.

c.   Select shape descriptors by clicking the appropriate row in the first column.

d.   Click the ADD button to add the descriptor to the study table.

For more detailed information on this step, see Chapter 7, Working with Descriptors.

6.   Determine other molecular features -- You can also add other molecular properties to the QSAR by calculating non-shape descriptors. Included in QSAR+ are a wide variety of spatial, electronic, and thermodynamic descriptors that constitute possible additional sources of features that govern biological activity. In this step, you select the specific descriptors that are appropriate for your study and add them to the study table. For information on descriptors, see Chapter 7, Working with Descriptors .

7.   Construct a trial QSAR -- The final step in MSA is to construct a trial QSAR. You select the data that you want to include in the QSAR equation by specifying the structures to be included and the conformation to use for each structure. Please see the section 7. Generating a trial QSAR on page 193.

To deal with conformations, the study table for an MSA QSAR has a third dimension. QSAR+ calculates and stores shape and non-shape descriptor values for each conformer generated from the structures in the study table. The information is stored in a separate Conformations table created for each structure. You determine the set of conformer data that you want to use by specifying the conformer to be used in the study table. Typically, the dataset is the set most similar to the shape reference.

1. Generating conformations

Many descriptors that QSAR+ calculates depend on the 3D structure of a molecule. Shape descriptors used in MSA also depend on conformation. Therefore, when you are constructing a 3D QSAR equation, you may want to generate conformations before calculating descriptors. In doing this, you assure that conformation is a factor in generating the equation.

This section describes

Setting conformation generation preferences (page 183)

Generating conformations (page 185)

Conformers should be generated using structures that have been subjected to energy minimization to generate a low-energy conformation for each structure. The simple way to be sure the structures in your study table are minimized is to check the Minimize Energy checkbox (Molecule Preferences control panel), so that all structures are minimized as they are loaded into the study table. Please see the Working with Molecules chapter.

Alternatively, you can request that all structures be minimized by checking the Minimize Structures before Generating Conformers checkbox on the QSAR Conformation Generation control panel.

Setting conformation generation preferences

Accessing the QSAR Conformation Generation control panel (this page).

Selecting a conformation generation method (page 183).

Applying an energy cutoff (page 184).

Specifying the number of conformers to be generated (page 185).

Accessing the QSAR Conformation Generation control panel

Alternatively, select the Preferences/Molecules... menu item, then click the Conformations button in the Molecule Preferences control panel.

Selecting a conformation generation method

Checking the This Panel checkbox allows you to use the QSAR Conformation Generation control panel to specify settings for generating conformers.

To select a method for conformational search

Check the Use Optimal Search Method checkbox or the Use checkbox to specify a conformational search method. If you select Use Optimal Search Method, MSA uses the best method to apply to the structures in the study table to generate the lowest-energy conformers.

If you want to select a particular method to generate conformers, check the Use checkbox. Available methods include Grid Scan, Random Sample, and Boltzmann Jump.

A brief description of each available conformational search method is provided below. For more information, see Cerius2 Conformational Search and Analysis.

Grid Scan -- This method is used to perform a simple systematic search in which each specified torsion angle is varied over a grid of equally spaced values.

Random Sample -- In this method, the starting conformation of a structure is perturbed by randomly altering values of all variable torsion angles. Each angle is assigned a value within a specified torsion angle window.

Boltzmann Jump -- In this method, torsion angles of a molecule are randomly altered within a specified angle window. After each random move, the Metropolis method is used to accept or reject the move.

As part of the conformer-generation process, conformers can be minimized. Select this option by checking the Minimize Conformers checkbox.

You can choose to save only unique conformers by checking the Retain Unique Conformers checkbox.

The conformers that are generated, minimized, and retained as unique are located at energy minima.

Applying an energy cutoff

• No Cutoff -- No maximum. All conformers are accepted.
• Relative -- A cutoff is specified relative to the lowest-energy conformer. A conformer is accepted if its energy value is less than the the cuttoff value plus the value of the lowest-energy conformer.
• Absolute -- An absolute cutoff value means that all conformers with an energy value below the cutoff are accepted.

Specifying the number of conformers

Generating conformations

To generate conformers when you have completed specifying conformer-generation settings, click the Generate Conformers button at the top of the QSAR Conformation Generation control panel. Choose from the popup to generate conformers for Current, Selected, or All molecules in the study table.

For an alternative way to generate conformers for structures already loaded in the study table, click Generate Conformers in the QSAR Conformation Generation control panel. To open the Conformation Generation control panel, click Generate Conformations on the MSA card in the QSAR deck.

To specify conformer generation as a default when structures are added to the study table, check the Generate Conformers checkbox in the Molecule Preferences control panel. To open the Molecule Preferences panel, select the Preferences/Molecules menu item in the study table.

For more information on the latter two options, see Chapter 6, Working with Molecules.

2. Hypothesizing an active conformer

This section describes

Accessing the Active Conformation control panel (this page)

Selecting the active conformer (page 186)

Displaying the active conformer (page 187)

Before you begin this step, you must do the following:

• Create a study table containing the structures you are investigating. Each structure must have its associated biological activity entered in the study table.
• Generate conformations for all structures in the study table (see Generating conformations on page 185.

Selecting the active conformer

Global Minimum of Most Active -- MSA looks at the global minimum of the most active compound in the study table (based on the value in the Activity column) and makes it the active conformer.

Selected Study Molecule -- Specify that the selectedconformer be the active conformer.

1.   Before you select a method for identifying an active conformer, you may need to override the default value that MSA uses to determine the most active molecule.

This value is set in the QSAR Preferences control panel with the Bigger Activity Values Indicate Greater Activity checkbox. The QSAR Preferences control panel, is opened by selecting the QSAR Preferences/General... menu item;

To specify that higher numbers indicate greater activity, check the Bigger Activity Values Indicate Greater Activity checkbo.

To specify that smaller numbers indicate greater activity, uncheck the Bigger Activity Values Indicate Greater Activity checkbox.

2.   After indicating how MSA should rank activities, select the method to identify an active conformer by checking the box next to the criterion in the Active Conformation control panel.

3.   Click the Select Active Conformer button.

Displaying the active conformer

To display the active conformer

Click the Display Active Conformer button on the Active Conformation control panel.

The active conformer is displayed in the model window.

3. Identifying a shape reference compound

This section describes

Accessing the Shape Reference control panel (this page)

Selecting a shape reference compound (page 188)

Displaying the selected shape reference compound (page 189)

Before you begin this step, you must identify an active conformer. For more information on this activity, see the section 2. Hypothesizing an active conformer on page 186.

Accessing the Shape Reference control panel

Selecting a shape reference compound

Active Conformation -- MSA selects a conformer of the most active Study table molecule. The conformer is the one most likely to result in the measured activity (that is, the active conformer).

Largest Molecule (Volume) -- MSA selects the molecule in the study table with the largest volume to be the shape reference compound.

Largest Molecule (Surface Area) -- MSA selects the molecule in the study table with the largest surface area to be the shape reference compound.

Global Minimum of Most Active -- MSA selects the most active molecule in the study table to be the shape reference compound and minimizes that structure. The conformer that is selected is the lowest-energy conformer. Before you use this criterion, you must override the default selection method for selecting the most active molecule.

Selected Study Molecule -- Manually specify a molecule from the study table to be your shape reference compound. For information on how to make selections, see Chapter 5, Working with the Study Table.

To choose a shape reference compound

1.   Before you select a criterion for identifying a shape reference compound, you may need to override the default value that MSA uses to determine the most active molecule. This value is set in the QSAR Preferences control panel, but it can be changed in the Shape Reference control panel. A change of the function on either panel is reflected in both panels. For more information, see Setting molecule-processing preferences on page 113.

To specify that larger numbers indicate greater activity, check the Bigger Activity Values Indicate Greater Activity checkbox at the bottom of the Shape Reference control panel.

To specify that smaller numbers indicate greater activity, uncheck the Bigger Activity Values Indicate Greater Activity checkbox at the bottom of the Shape Reference control panel

2.   Select the criterion to identify a shape reference compound by clicking the box next to the criterion in the Shape Reference control panel.

3.   Click the Select Shape Reference button.

Displaying the selected shape reference compound

To display the shape reference compound

Click the Display Shape Reference button on the Shape Reference control panel. The model window changes to border mode, and the selected shape reference compound is displayed in the main part of the window.

MSA highlights the rows in the study table and in the Conformers table that contain information about the shape reference compound.

4. Aligning molecules

This section describes

Accessing the Shape Reference control panel (this page)

Aligning models (page 190)

Removing alignment information (page 193)

If you do not select a shape reference compound before you perform alignment activities, MSA automatically identifies a shape reference compound using the default selection method identified in the top portion of the Shape Reference control panel. For more information on selecting a shape reference compound, see the section 3. Identifying a shape reference compound on page 187.

Accessing the Shape Reference control panel

Aligning models

The MCS method looks at molecules as points and lines and uses the techniques of graph theory to identify patterns. It finds the largest subset of atoms in the shape reference compound that is shared by all the structures in the study table and uses this subset for alignment.

The CSS method starts with defining a core model, which is a substructure to find and match in all your selected models. The core model itself is just a Cerius² model regarded as composed of core atoms and substitution sites. Core atoms are the atoms in your core model that exactly match a substructure in your align models, and substitution sites represent sites where the core model and the matched align models differ.

MCS and CSS compared

MCS is a general procedure for finding a common substructure between two models (and hence for defining atom matches), but the underlying algorithm is based on an exhaustive tree search and can take a significant amount of time for models that have a highly branched structure, e.g., fused ring systems.

CSS has the disadvantage that you have to specify a suitable core model that you know is a common substructure for all your align models. However, a CSS search is generally much faster than a MCS search and gives much more control over the resultant atom matches. In addition, because each of your align models is first matched to a single core model (these matches being stored internally), matching all align models to each other takes only a little longer than matching all align models to a single target model. (With MCS each pair of align models is matched independently.)

Aligning models by MCS method

1.   Determine what portion of the shape reference compound (target) you want to use to align the study table structures by setting the Align to Targets Using popup in the Shape Reference control panel to ALL or SELECTED.

2.   Determine what structures you want to align to the shape reference compound by setting the Align Target Molecule(s) popup to ALL structures, SELECTED structures, or only the CURRENT structure.

3.   Check the Overlay Aligned Molecules check box if you want molecules to be displayed in overlay mode.

4.   Click the ALIGN pushbutton.

If you specified overlay mode for the display, the results of the alignment are displayed in the model window. If you checked Recalculate Descriptors After Align, these calculations are completed automatically after the alignment is complete.

You can also align models by selecting Align Models from the ALIGN MOLECULES card in the DRUG DISCOVERY card deck. From the Align Models control panel, you can perform both MCS and CSS alignment.

Please see the online help for additional information.

Aligning models by CSS method

Clicking the DEFINE button specifies that the current model is the CSS core model. Clicking the Match Atoms using CSS Search button starts the process of atom matching for the selected models. The effect of the matching is the same as for MCS matching, i.e., a set of editable atom pairs is displayed in the Viewer window and the number of atom matches is displayed in the Align Models table.

Core atoms and substitution sites in your core model are identified according to the RMS Align setting in the Align Preferences control panel:

• Open valencies: All atoms in your core model are core atoms. If you added hydrogen atoms to your core model to fill all open valencies, then these atoms become substitution sites. Typically, defining your core model using this option is simply a matter of selecting the core substructure atoms in one of your align models, copying and pasting them to a create a new model, and clicking the DEFINE button in the Align Models control panel.
• R-Group atoms: Singularly attached atoms denoted as R-groups are identified as substitution sites on the core model, and the remaining atoms are core atoms. Hence, you simply create your core model using the Analog Builder module or load a previously built core model. CSS thus becomes a useful tool for aligning models built from core substituent libraries.

To delete alignment information

1.   Specify the molecules for which you want to remove alignment information. You can remove information for all, selected, or current molecules. Make your selection from the popup.

2.   Click Clear Alignment information from... Molecule(s).

5. Measure molecular shape commonality

6. Determining other molecular features

7. Generating a trial QSAR

After you complete the QSAR-generation process and examine the equation that is generated, you can repeat the entire process or any portion of it as described in Overview of molecular shape analysis on page 178. The outcome of the process is an optimized QSAR equation that corresponds to the best fit between observed activities and computed molecular descriptor data.

This section describes

Accessing the Select Conformers control panel (page 194)

Selecting conformers based on a specified set of descriptor data (page 194)

Generating a trial QSAR equation (page 195)

The information in this section assumes that you have completed Steps 1 through 6 (above) of the MSA process, including generating shape and nonshape descriptors for all molecules in the study table.

Accessing the Select Conformers control panel

Selecting conformers

The selection process examines all descriptor data for each conformer to determine the conformer with data that best match the shape reference data. The data are evaluated using the partial least squares regression method (pls). Since regression analysis is run for all specified conformers and descriptors, the conformer selection process can be lengthy.

To select conformers

1.   Determine which descriptors should be used by MSA to identify the significant conformation for each study table molecule. Indicate your choice by clicking the All 3D Descriptors radio box or the Shape Descriptors radio box.

2.   Determine what study table structures should be involved in the conformer-selection process by choosing All, Selected, or Current from the Select Conformers for popup.

3.   Decide if conformation-dependent properties should be recalculated before selecting significant conformations. Check the Regenerate Conformer Tables checkbox if you want recalculation performed.

4.   Decide if you want MSA to generate a QSAR when the conformer-selection process is complete. If so, check the Calculate QSAR When Done checkbox.

5.   Click Select to start the selection process.

Generating a trial QSAR equation

To generate a trial QSAR equation

You can generate a QSAR equation using one of several options:

• If you check the Calculate QSAR When Done checkbox on the Select Conformers control panel, MSA automatically generates a QSAR equation when it is finished selecting conformers to be used in the calculation. With this option, you can complete the MSA process with no additional intervention as soon as you make your selections on the Select Conformers control panel.

or:

• Click the RUN button on the QSAR tool bar at the top of the study table.