HINT 2.30 Manual: Chapter 6S

LESSON 7: HINT Interaction Field Analysis (HIFA)

This lesson demonstrates a new 3D QSAR technique called HIFA (HINT Interaction Field Analysis) that is a unique method for analyzing a set of receptor/ligand complexes. HIFA is a hybrid of the CoMFA method, e.g., Lesson 6, and HINT intermolecular interaction maps (Lesson 5). The method was first described by S.F. Semus: Med. Chem. Res. 1999, 9, 535-547, in which the map creation and importation into the Sybyl Molecular Spreadsheet was performed manually. Here is described the use of an automated procedure for HIFA, but it still requires several semi-manual steps. The data set used is an adaptation of the M.K. Holloway et al. HIV-1 inhibitor data set: J. Med. Chem. 1995, 38, 305-17. These complexes have been optimized and energy minimized as described in D.T. Wei, J.C. Meadows and G.E. Kellogg, Med. Chem. Res. 1997, 7, 259-270.

It would be best for this lesson if there were no molecules or backgrounds from previous lessons currently active in SYBYL. If you are entering the HINT Tutorial at this point, follow the instructions in Step 1 of Lesson 1. It also should be noted that creating this spreadsheet is computationally intensive -- as described this run took 3-4 hours on a 360 MHz r12000 SGI Octane2. Because the HIFA demo data set is quite large it has been tarred and gzipped in the HINT demo directory (hifa.tgz). To unpack it, in your HINT demo directory, enter the command gzip -dc hifa.tgz | tar xvf - to create a hifa subdirectory in the hint demo directory that contains all of the data for the demo. You will want to cd to this hifa subdirectory.

  1. Create the SYBYL Molecular Spreadsheet from the HIV-1 inhibitors

    The general idea is to set each row of the spreadsheet to represent an enzyme/inhibitor complex. However, HINT expects the enzyme (receptor) and inhibitor (ligand) to be in separate molecular areas in order to complete an InterMolecular calculation. So, we have to trick Sybyl to make this work. What we do is load the ligands into the rows of the spreadsheet as in CoMFA or other 3D QSAR techniques, and then we place a reference to the corresponding receptor for each ligand in a special column of the MSS. This reference can either be to a molecular area or to a sybyl .mol2 file.

    First, create the Molecular Spreadsheet. From the File pulldown on the menubar select Molecular Spreadsheet and New.... The Data Source is Database, i.e., representing the ligand molecules. In the DATABASE_FILE dialog box, select inhibitors.mdb.

    Next, we need to add the biological activity data, which is in this case pIC50 measurements for these inhibitors. This step also will bypass a tedious part of the HIFA setup. The ASCII file pIC50.tripos has been created with two columns: 1) pIC50 and 2) receptor. The receptor column contains the filename for the .mol2 models corresponding to each ligand in the spreadsheet. This is a relatively convenient way to enter this information in the spreadsheet, but there are other options as will be described below. From the File menubar item on the Spreadsheet choose Import.... The Format: is Tripos and the File: is pIC50.tripos. Press Import. Two columns should be loaded into the speadsheet; in column 1 are the biological data and in column 2 are the "receptor" filenames. Note, for the filenames as input in pIC50.tripos to work, it is critical that you start Sybyl in the directory that contains these files.

  2. AutoFill with HIFA Column

    Making sure that none of the Table columns or rows are selected, press the Table AutoFill button, select Column and HIFA from the New column type menu. (If the HIFA column is not present, Cancel and type mss!reset_eslc to the Sybyl prompt. The HIFA column should now be in the AutoFill list.) This Add HIFA Columns dialog will calculate the HIFA maps for each ligand row of the MSS.

    There are a number of parameters to be set and/or confirmed on this dialog before the HIFA procedure can commence.

    First, choose the Field Values. For the first part of this lesson, use Interactions: Hydrophobic. Be sure the Smoothing is None or else the calculation will take approximately 8 times longer.

    For Receptor Partition we choose the methods and parameters to partition the HIV-1 in this case. Here the Partition Method should be Dictionary, the Hydrogen Treatment should be Essential and the Solvent Condition should be Inferred. (Note that in this case the protons, particularly around Asp25 and Asp125 of HIV-1 have been optimally positioned.) Direction Vectors should be off.

    Next, we need to set the Receptor data for HIFA. Although we have loaded column 2 with the .mol2 filenames for the calculation, HIFA does not know that, so, press Assign (Existing) Receptor Column... to activate the Receptor MolColumn Dialog. Choose Fill Column with: Sybyl .mol2 Files and enter 2 as the Receptor MolData Column. Press OK to return to the HIFA dialog. While we are on this topic, take a look at the Receptor MolData Dialog. This dialog lets you set the receptor data for each row as either Molecular Areas or Sybyl .mol2 Files. You also could choose to either Fill All Rows at Once or Cycle through Rows, in which case you would be prompted to enter the Molecular Area or .mol2 file for each row.

    Now we need to set the options for the Ligand Partition. Here the Partition Method should be Calculate, the Hydrogen Treatment should be Essential, and the Polar Proximity should be Via Bond. Direction Vectors should be off.

    The default Distance Function... should be fine. And the current Cutoffs; Cutoff Radius of 7.0 and Van der Waals Limit of 1.00 are appropriate.

    Lastly, we need to set the HIFA Region. As in CoMFA, the idea is to determine the union box the encloses all molecules with an adequate "margin". Select the Calculate Automatically... option to bring up the Calculate CoMFA Region Automatically dialog box which defines the region based on the union space spanned by the molecule set. Set the X, Y and Z Spacings to 1 Angstrom and the X, Y and Z Margins to 5 Angstrom. Choose a CoMFA Region Filename like hifa1A.rgn. Press OK.

    To start the HIFA column fill press OK on the HIFA dialog. When you are prompted for column name, enter hifa3_hp. This calculation will take some time, about 1.5-2.0 hours on a relatively fast SGI (360 MHz r12000 Octane2).

  3. Fill Second HIFA Column

    When the first HIFA column (hydrophobic) has finished you can add a second one (polar) or proceed to the PLS analysis in the next step.

    Because of a bug in Sybyl, you must first save the MSS, and completely exit Sybyl before adding a second "External" Field to the Spreadsheet. When you re-start Sybyl, choose File, Molecular Spreadsheet and Open from the menubar. Select inhibitors.tbl. To add the second column, first make sure that no columns or rows are selected in the MSS and press AutoFill. Again, choose the HIFA New Column Type. Use the same settings as above except set Interactions: Polar and you should set the Region to Use Pre-Existing and choose the HIFA Region File as hifa1A.rgn by pressing the ... button. To start the second HIFA column fill press OK on the HIFA dialog. When you are prompted for column name, enter hifa4_ab. This calculation will again take some time, about 1.5-2.0 hours on a relatively fast SGI (360 MHz r12000 Octane2).

  4. PLS ANalysis of HIFA Column(s)

    These columns are analyzed in precisely the same way as CoMFA columns. Choose QSAR, Partial Least Squares... on the MSS menubar. The 2nd column will autoamtically be dropped as it is a string field and not valid "data". First, choose columns 1 and 3 for analysis. Leave-One-Out and Use SAMPLS are appropriate for this first run. Press Do PLS. You should shortly see output like the following in the Sybyl text port:

    SAMPLS: SAMple-distance Partial Least Squares
            Copyright (c) 1993 Merck & Co., Inc.
                  All rights reserved.  
    Author: Bruce L. Bush, Merck & Co., Inc. 
             PO Box 2000, Rahway, NJ 07065  
       Subject:  TAILOR!PLS
    Option Name         Description                               Value
    -----------         -----------                               -----
    BOOTSTRAPPING       Number of bootstrap runs to make          0
    CENTERING           Force intercept thru 0.0?                 NO
    COMPONENTS          Number of components to use               6
    CROSSVALIDATION     Number of crossvalidation groups          0
    EPSILON             Convergence criterion                     1.e-4
    ITERATION           Maximum number of iterations              100
    SCALING_METHOD      Pre-analysis scaling (COMFA_STD for CoMF  COMFA_STD
    XLIST               Print info about xval groups?             NO
    SAMPLS has finished running. This is NOT a full PLS analysis.
    Standard Error of Prediction for 6 components:
    1.283 1.248 1.306 1.277 1.289 1.256
    Crossvalidated R2 for 6 components:
    0.265 0.327 0.287 0.342 0.354 0.409
    -- optimum is 0.409 at 6 components
      Please note: Running SAMPLS in Sybyl produces 
    ONLY cross-validated R2 and standard error values.  
       No graphical or file outputs are produced.
    If you also calculated the second HIFA column, you can calculate the PLS on this column (q2 = 0.255 at 5 components) and the combination of the two HIFA columns (q2 = 0.381 at 5 components).

    As described in Lesson 6 these fields can be extracted and plotted. The analysis of these "CoMFA" maps is complex, but the following is a preliminary discussion of the combination map for this case. The following were mapped: in the "Hydrophobic" (column 3) -0.005 (green), +0.01 (purple); in the "Polar" (column 4) -0.01 (blue), +0.01 (red). Green conours show where additional hydrophobic interactions would enhance activity; purple contours show where additional hydrophobic interactions compromise activity; blue contours show where enhancing the hydrogen bonds would improve activity; red contours show where trying to enhance hydrogen bonding may be fruitless. In the map for the HIV-1 inhibitors, the red region is wrapped around the blue region. This probably reflects the confusing nature of the hydrogen bonding between the two aspartates (25 and 125) and the ligand. The two large purple contours are representative of the fact that these regions of space need to be dominated by favorable polar interactions, and that when members of the ligand set placed hydrophobic groups in this region, the activity was decreased.

    HINT HIFA maps for HIV-1 inhibitors