import Java.util.BitSet;

import org.openscience.cdk.DefaultChemObjectBuilder;
import org.openscience.cdk.exception.CDKException;
import org.openscience.cdk.exception.InvalidSmilesException;
import org.openscience.cdk.fingerprint.ExtendedFingerprinter;
import org.openscience.cdk.smiles.SmilesParser;

public class FingerprinterTest {

 /**
  * @param args
  * @throws CDKException
  * @throws InvalidSmilesException
  */
 public static void main(String[] args) throws InvalidSmilesException, CDKException {
  ExtendedFingerprinter fingerprinter = new ExtendedFingerprinter();
  SmilesParser sp = new SmilesParser(DefaultChemObjectBuilder.getInstance());
  BitSet bt = fingerprinter.getFingerprint(sp.parseSmiles("c2ccc1ccccc1c2"));
 }

}

import javax.servlet.http.HttpServletResponse;
import javax.vecmath.Point2d;

import org.apache.log4j.Logger;
import org.openscience.cdk.Molecule;
import org.openscience.cdk.interfaces.IAtom;
import org.openscience.cdk.interfaces.IMolecule;
import org.openscience.cdk.io.MDLReader;
import org.openscience.cdk.layout.StructureDiagramGenerator;
import org.openscience.cdk.renderer.Renderer2DModel;
import org.openscience.cdk.renderer.SimpleRenderer2D;

public class ImageTypeExporterUtil {
 private static final Logger logger = Logger.getLogger(ImageTypeExporterUtil.class);
 
 /**
  * show molecule structure to image type (png, jpeg)
  *
  * @param mol String molecule stucture
  * @param length width and height
  * @param response HttpServletResponse object
  * @throws Exception
  *             if occurred exception ,then throw Exception
  */
 public static void showAsImage(String stucture, Integer length, HttpServletResponse response) throws Exception {
  logger.debug("ImageTypeExporterUtil.showAsImage..");
  
  StringReader mdl = new StringReader(stucture);
  MDLReader cdkMDL = new MDLReader(mdl);
  Molecule mol = new Molecule();
  cdkMDL.read(mol);
  // null coordinates
  Iterator<IAtom> itatoms = mol.atoms();
  while (itatoms.hasNext()) {
   IAtom atom = itatoms.next();
   atom.setPoint2d(null);
   atom.setPoint3d(null);
  }
  // generate 2D coordinates
  StructureDiagramGenerator sdg = new StructureDiagramGenerator();
  sdg.setMolecule(mol);
  try {
   sdg.generateCoordinates();
  } catch (Exception ex) {
   ex.printStackTrace();
  }
  IMolecule layedOutMol = sdg.getMolecule();
  // scale molecule
  final double UNDEF_POS = 100000;
  double minX = UNDEF_POS, minY = UNDEF_POS, maxX = UNDEF_POS, maxY = UNDEF_POS;
  itatoms = layedOutMol.atoms();
  while (itatoms.hasNext()) {
   IAtom atom = itatoms.next();
   Point2d point2d = atom.getPoint2d();
   if (minX == UNDEF_POS || minX > point2d.x)
    minX = point2d.x;
   if (minY == UNDEF_POS || minY > point2d.y)
    minY = point2d.y;
   if (maxX == UNDEF_POS || maxX < point2d.x)
    maxX = point2d.x;
   if (maxY == UNDEF_POS || maxY < point2d.y)
    maxY = point2d.y;
  }
  double scaleX = length / (maxX - minX + 1);
  double scaleY = length / (maxY - minY + 1);
  double scale = scaleX > scaleY ? scaleY : scaleX;
  double centreX = scale * (maxX + minX) / 2.;
  double centreY = scale * (maxY + minY) / 2.;
  double offsetX = length / 2. - centreX;
  double offsetY = length / 2. - centreY;
  itatoms = layedOutMol.atoms();
  while (itatoms.hasNext()) {
   IAtom atom = itatoms.next();
   Point2d a = atom.getPoint2d();
   Point2d b = new Point2d();
   b.x = a.x * scale + offsetX;
   b.y = a.y * scale + offsetY;
   atom.setPoint2d(b);
  }
  // set rendering properties
  Renderer2DModel r2dm = new Renderer2DModel();
  r2dm.setDrawNumbers(false);
  r2dm.setUseAntiAliasing(true);
  r2dm.setColorAtomsByType(true);
  r2dm.setShowAtomTypeNames(false);
  r2dm.setShowAromaticity(true);
  r2dm.setShowImplicitHydrogens(false);
  r2dm.setShowReactionBoxes(false);
  r2dm.setKekuleStructure(false);
  Dimension dim = new Dimension();
  dim.setSize(length, length);
  r2dm.setBackgroundDimension(dim);
  r2dm.setBackColor(java.awt.Color.WHITE);
  // render the image
  SimpleRenderer2D renderer = new SimpleRenderer2D();
  renderer.setRenderer2DModel(r2dm);
  BufferedImage bufferedImage = new BufferedImage(length, length,
    BufferedImage.TYPE_INT_RGB);
  Graphics2D graphics = bufferedImage.createGraphics();
  graphics.setPaint(java.awt.Color.WHITE);
  Rectangle2D.Float rectangle = new Rectangle2D.Float(0, 0, length, length);
  graphics.fill(rectangle);
  renderer.paintMolecule(layedOutMol, graphics);
  // write the image to response
  response.setContentType("image/png");
  OutputStream out = response.getOutputStream();
  try {
   javax.imageio.ImageIO.write(bufferedImage, "png", out);
  } finally {
   out.close();
  }
 }
}

package com.founder.cdk;

import Java.io.File;
import Java.io.FileNotFoundException;
import Java.io.FileReader;
import Java.util.ArrayList;
import Java.util.List;

import org.openscience.cdk.ChemFile;
import org.openscience.cdk.ChemObject;
import org.openscience.cdk.exception.CDKException;
import org.openscience.cdk.interfaces.IAtomContainer;
import org.openscience.cdk.io.MDLV2000Reader;
import org.openscience.cdk.smiles.smarts.SMARTSQueryTool;
import org.openscience.cdk.tools.manipulator.ChemFileManipulator;

public class SMARTSQueryToolTest {

 static SMARTSQueryTool sqt;static {
        try {
            sqt = new SMARTSQueryTool("c2ccc1ccccc1c2");
        } catch (CDKException e) {           
        }
    }

 /**
  * @param args
  */
 public static void main(String[] args) {
  String filename = "H:\\molecules.sdf";
  try {
            MDLV2000Reader reader = new MDLV2000Reader(new FileReader(new File(filename)));
            ChemFile chemFile = (ChemFile) reader.read((ChemObject) new ChemFile());
            List<IAtomContainer> containersList = ChemFileManipulator.getAllAtomContainers(chemFile);
           
            List<IAtomContainer> substructureList = new ArrayList<IAtomContainer>();
            
            sqt.setSmarts("c1ccc3c(c1)ccc4c2ccccc2ccc34");  //重新设置匹配的smiles值
            boolean matched = false;
            for (IAtomContainer molecule : containersList) {
                matched = sqt.matches(molecule);
                if (matched){
                 substructureList.add(molecule);
                }
            }
            System.out.println(substructureList.size());
           
            for (IAtomContainer molecule : substructureList) {
                 System.out.println(molecule.getProperty("ID"));
            }
           
        } catch (CDKException e) {
            e.printStackTrace();
        } catch (FileNotFoundException e) {
           e.printStackTrace();
        }

 }

}

通过测试, matches方法速度很慢, 一般一个结构需要200ms-1000ms左右.

import Java.io.File;
import Java.io.FileNotFoundException;
import Java.io.FileReader;
import Java.util.List;

import org.openscience.cdk.ChemFile;
import org.openscience.cdk.ChemObject;
import org.openscience.cdk.Molecule;
import org.openscience.cdk.exception.CDKException;
import org.openscience.cdk.interfaces.IAtomContainer;
import org.openscience.cdk.io.MDLReader;
import org.openscience.cdk.io.MDLV2000Reader;
import org.openscience.cdk.tools.manipulator.ChemFileManipulator;

public class ReadSDFTest {

 /**
  * @param args
  * @throws CDKException
  * @throws FileNotFoundException
  */
 public static void main(String[] args) throws CDKException, FileNotFoundException {
  String filename = "H:\\molecules.sdf";
       
//  InputStream ins = ReadSDFTest.class.getClassLoader().getResourceAsStream(filename);
//  MDLReader reader = new MDLReader(ins);

   //alternatively, you can specify a file directly
   MDLV2000Reader reader = new MDLV2000Reader(new FileReader(new File(filename)));

  ChemFile chemFile = (ChemFile)reader.read((ChemObject)new ChemFile());
  
  List<IAtomContainer> containersList = ChemFileManipulator.getAllAtomContainers(chemFile);
  
  Molecule molecule = null;
  for (IAtomContainer mol : containersList) {
   molecule = (Molecule) mol;
   System.out.println(molecule.getProperties());
   System.out.println(molecule.getProperty("CD_MOLWEIGHT"));
//   Fingerprinter fp = new Fingerprinter();
//   BitSet bt = fp.getFingerprint(molecule);
//   System.out.println(bt);
  }
 }

}

import Java.io.StringReader;
import Java.sql.Connection;
import Java.sql.ResultSet;
import Java.sql.SQLException;
import Java.util.ArrayList;
import Java.util.BitSet;
import Java.util.List;

import org.openscience.cdk.Molecule;
import org.openscience.cdk.exception.CDKException;
import org.openscience.cdk.fingerprint.Fingerprinter;
import org.openscience.cdk.io.MDLReader;
import org.openscience.cdk.similarity.Tanimoto;

public class CDKTest {

 /**
  * @param args
  */
 public static void main(String[] args) {
  
  // MySQL
  long t1 = System.currentTimeMillis();
  try {
   Class.forName("com.mysql.jdbc.Driver").newInstance();
   Connection con = Java.sql.DriverManager
     .getConnection(
       "jdbc:mysql://localhost/coocoo?useUnicode=true&characterEncoding=utf-8&zeroDateTimeBehavior=convertToNull",
       "root", "root");
   

   ResultSet results = null;
   String querySQL = "select id, structure from structure ";
   
   results = con.createStatement().executeQuery(querySQL);
 
   // dump out the results

   List<Molecule> list = new ArrayList<Molecule>();
   Fingerprinter fp = new Fingerprinter();
   BitSet bt = null;
   while (results.next()) {
    Long id = results.getLong("id");
    
    //根据结构数据生成分子对象
    StringReader mdl = new StringReader(results.getString("structure"));
    MDLReader cdkMDL = new MDLReader(mdl);
    Molecule molecule = new Molecule();
    cdkMDL.read(molecule);
    if (id == 1220) {
     bt = fp.getFingerprint(molecule);
    }
    list.add(molecule);
    
   } 
   System.out.println("size:=" + list.size());
   
   List<Molecule> resultList = new ArrayList<Molecule>();
        
         long t2 = System.currentTimeMillis();
         System.out.println("Thread: collection data in " + (t2 - t1) + " ms.");
         for (Molecule molecule : list) {
             try {
                 float coefficient = Tanimoto.calculate(fp.getFingerprint(molecule), bt);  //计算相似度
                 if (coefficient > 0.9) {
                  resultList.add(molecule);
                 }
             } catch (CDKException e) {

             }
         }
         long t3 = System.currentTimeMillis();
        
         System.out.println(resultList.size());
         System.out.println("Thread: Search in " + (t3 - t2) + " ms.");
        
   con.close();
  } catch (InstantiationException e) {
   e.printStackTrace();
  } catch (IllegalAccessException e) {
   e.printStackTrace();
  } catch (ClassNotFoundException e) {
   e.printStackTrace();
  } catch (SQLException e) {
   e.printStackTrace();
  } catch (CDKException e) {
   e.printStackTrace();
  }
  long t4 = System.currentTimeMillis();
        System.out.println("Thread: all in " + (t4 - t1) + " ms.");
 }

}

Rich Apodaca wrote a great serious posts named Fast Substructure Search Using Open Source Tools providing details on substructure search with MySQL. But, however, poor binary data operation functions of MySQL limited the implementation of similar structure search which typically depends on the calculation of Tanimato coefficient. We are going to use Java & CDK to add this feature.

As default output of CDK fingerprint, java.util.BitSet with Serializable interface is perfect data format of fingerprint data storage. Java itself provides several collections such as ArrayList, LinkedList, Vector class in package Java.util. To provide web access to the search engine, thread unsafe ArrayList and LinkedList have to be kicked out. How about Vector? Once all the fingerprint data is well prepared, the collection  function we need to do similarity search is just iteration. No add, no delete. So, a light weight array is enough.

Most of the molecule information is stored in MySQL database, so we are going to map fingerprint to corresponding row in data table. Here is the MolDFData class, we use a long variable to store corresponding primary key in data table.

public class MolDFData implements Serializable {

    private long id;

   private BitSet fingerprint;

    public MolDFData(long id, BitSet fingerprint) {

        this.id = id;

        this.fingerprint = fingerprint;

    }

    public long getId() {

        return id;

    }

    public void setId(long id) {

        this.id = id;

    }

    public BitSet getFingerprint() {

        return fingerprint;

    }

    public void setFingerprint(BitSet fingerprint) {

        this.fingerprint = fingerprint;

    }

}

This is how we storage our fingerprints.

private MolFPData[] arrayData;

No big deal with similarity search. Just calculate the Tanimoto coefficient, if it’s bigger than minimal  similarity you set, add this one into result.

    public List searchTanimoto(BitSet bt, float minSimlarity) {

        List resultList = new LinkedList();
        int i;
        for (i = 0; i < arrayData.length; i++) {
            MolDFData aListData = arrayData[i];
            try {
                float coefficient = Tanimoto.calculate(aListData.getFingerprint(), bt);
                if (coefficient > minSimlarity) {
                    resultList.add(new SearchResultData(aListData.getId(), coefficient));
                }
            } catch (CDKException e) {
            }
            Collections.sort(resultList);
        }
        return resultList;
    }
Pretty ugly code?  Maybe. But it really works, at a acceptable speed.

Tests were done using the code blow on a macbook(Intel Core Due 1.83 GHz, 2G RAM).

long t3 = System.currentTimeMillis();
List<SearchResultData> listResult = se.searchTanimoto(bs, 0.8f);
long t4 = System.currentTimeMillis();
System.out.println("Thread: Search done in " + (t4 - t3) + " ms.");

In my database of 87364 commercial compounds, it takes 335 ms.

import org.openscience.cdk.annotations.TestClass;
import org.openscience.cdk.annotations.TestMethod;
import org.openscience.cdk.exception.CDKException;

import Java.util.BitSet;

/**
 *  Calculates the Tanimoto coefficient for a given pair of two
 *  fingerprint bitsets or real valued feature vectors.
 *
 *  The Tanimoto coefficient is one way to
 *  quantitatively measure the "distance" or similarity of
 *  two chemical structures.
 *
 *  <p>You can use the FingerPrinter class to retrieve two fingerprint bitsets.
 *  We assume that you have two structures stored in cdk.Molecule objects.
 *  A tanimoto coefficient can then be calculated like:
 *  <pre>
 *   BitSet fingerprint1 = Fingerprinter.getFingerprint(molecule1);
 *   BitSet fingerprint2 = Fingerprinter.getFingerprint(molecule2);
 *   float tanimoto_coefficient = Tanimoto.calculate(fingerprint1, fingerprint2);
 *  </pre>
 *
 *  <p>The FingerPrinter assumes that hydrogens are explicitely given, if this
 *  is desired!
 *  <p>Note that the continuous Tanimoto coefficient does not lead to a metric space
 *
 *@author         steinbeck
 * @cdk.githash
 *@cdk.created    2005-10-19
 *@cdk.keyword    jaccard
 *@cdk.keyword    similarity, tanimoto
 * @cdk.module fingerprint
 */
@TestClass("org.openscience.cdk.similarity.TanimotoTest")
public class Tanimoto
{

    /**
     * Evaluates Tanimoto coefficient for two bit sets.
     *
     * @param bitset1 A bitset (such as a fingerprint) for the first molecule
     * @param bitset2 A bitset (such as a fingerprint) for the second molecule
     * @return The Tanimoto coefficient
     * @throws org.openscience.cdk.exception.CDKException  if bitsets are not of the same length
     */
    @TestMethod("testTanimoto1,testTanimoto2")
    public static float calculate(BitSet bitset1, BitSet bitset2) throws CDKException
    {
        float _bitset1_cardinality = bitset1.cardinality();
        float _bitset2_cardinality = bitset2.cardinality();
        if (bitset1.size() != bitset2.size()) {
            throw new CDKException("Bisets must have the same bit length");
        }
        BitSet one_and_two = (BitSet)bitset1.clone();
        one_and_two.and(bitset2);
        float _common_bit_count = one_and_two.cardinality();
        return _common_bit_count/(_bitset1_cardinality + _bitset2_cardinality - _common_bit_count);
    }
   
    /**
     * Evaluates the continuous Tanimoto coefficient for two real valued vectors.
     *
     * @param features1 The first feature vector
     * @param features2 The second feature vector
     * @return The continuous Tanimoto coefficient
     * @throws org.openscience.cdk.exception.CDKException  if the features are not of the same length
     */
    @TestMethod("testTanimoto3")
    public static float calculate(double[] features1, double[] features2) throws CDKException {

        if (features1.length != features2.length) {
            throw new CDKException("Features vectors must be of the same length");
        }

        int n = features1.length;
        double ab = 0.0;
        double a2 = 0.0;
        double b2 = 0.0;

        for (int i = 0; i < n; i++) {
            ab += features1[i] * features2[i];
            a2 += features1[i]*features1[i];
            b2 += features2[i]*features2[i];
        }
        return (float)ab/(float)(a2+b2-ab);
    }
}

通过源码可以看出calculate(BitSet bitset1, BitSet bitset2)方法,是通过比较两个分子的fingerprint的位,来计算相似度.通过BitSet的and操作得到共同的个数,然后在除以总共为true的个数,这样就得到相似值.

JChemPaint was started by Christoph Steinbeck in the late 1990's to be the complementary structure editor to Jmol. It was then co-developed by Egon Willighagen and others. Jmol again is a visualisation and analysis tool for 3D molecular structures, started by Dan Gezelter at Notre Dame University, initiator of the Open Science Project and, like JChemPaint, developed by an international team of opensource programmers.

In at least three aspects JChemPaint is different from other 2D editors:

• JChemPaint is open source and free software. We believe that scientific software, especially when its development was publicly funded, should be free. As the GNU people put it: «`Free software´ is a matter of liberty, not price. To understand the concept, you should think of `free speech´, not `free beer´». Everyone can participate in the development of the program. Everyone can download and change the source code, provided that they make the changes publicly available again, according to the GNU Lesser General Public License, LGPL. This ensures that the community can take advantage of any bugfix or enhancement made to the system. It also ensures that a scientist, who needs a standard piece of software like a structure editor as a helper application in his/her new program, does not have to reinvent the wheel over and over again because all the structure editors that have been written before are now proprietary software. If there is a free structure editor, he/she can focus on the real science.
• JChemPaint is in constant development and you can help (see below).
• Since JChemPaint is written in Java, it runs on any computing platform and operating system for which a Java Virtual Machine (of version >= 1.3 up to JCP 2.4 and version >= 1.5 for JCP > 2.4) has been implemented (like Linux, Windows, Solaris, AIX and others).
• JChemPaint is available free of charge.
• JChemPaint is translated into several languages: Dutch, French, German, Polish, Portuguese and Spanish.