package pp2.prediction.knn; import java.io.BufferedReader; import java.io.BufferedWriter; import java.io.IOException; import pp2.go.Relations; import pp2.tools.Tools; /** * performs all operations necessary to predict a list of targets * * @author Thomas Hopf * */ public class KNNKernel { // holds all relations in the GO ontology tree private Relations relations; public KNNKernel(String closureFile, String termFile, boolean noCellularComponent) { relations = new Relations(closureFile, termFile, noCellularComponent); } /** * fetches the next target with all BLAST hits from the reader * @param reader * @return */ private PredictionTarget getNextTarget(BufferedReader reader, boolean eliminateDuplicateHits) { PredictionTarget nextTarget = new PredictionTarget(); try { String line = null; while((line = reader.readLine()) != null) { // return current target when we hit the "end-of-target"-line "//" if(line.equals("//")) return nextTarget; // get ID of current target if(line.startsWith(">")) nextTarget.setTargetID(line.substring(1)); // if line starts with "#", this is one of the BLAST hits in the database if(line.startsWith("#")) nextTarget.addBlastHit(line.substring(1), eliminateDuplicateHits); } } catch(Exception e) { e.printStackTrace(); } return null; } /** * carries out the prediction of all targets in file containing multiple targets * @param targetFile */ public void predictTargetList(String targetFile, String outFile, boolean eliminateDuplicateHits, boolean useEValueThreshold, double eValueTreshold, int k, String distributionFile) { try { BufferedReader in = Tools.openFile(targetFile); BufferedWriter naiveWriter = Tools.writeToFile(outFile + ".naive"); BufferedWriter leaveWriter = Tools.writeToFile(outFile + ".leaves"); // read the distribution of global quality scores double[] distribution = null; if(distributionFile != null) { distribution = new double[100]; int i = 0; BufferedReader distReader = Tools.openFile(distributionFile); String line = null; while((line = distReader.readLine()) != null) { distribution[i] = Double.parseDouble(line); i++; } if(i != 100) System.err.println("warning: distribution file does not contain exactly 100 quantile lines."); distReader.close(); } // the current target for which to perform a prediction PredictionTarget target = null; int numPredicted = 0; // iterate over all targets while((target = getNextTarget(in, eliminateDuplicateHits)) != null) { // skip double // error in blast output file if(target.getTargetID() == null) continue; //System.err.println(">" + target.getTargetID()); try { target.predict(relations, useEValueThreshold, eValueTreshold, k, naiveWriter, leaveWriter, distribution); } catch(Exception e) { System.err.println("an error occured while predicting target " + target.getTargetID()); e.printStackTrace(); } numPredicted++; } System.out.println(" predicted " + numPredicted + " sequences."); in.close(); naiveWriter.close(); leaveWriter.close(); } catch(IOException e) { e.printStackTrace(); } } }