net.maizegenetics.analysis.numericaltransform.ImputationPlugin Maven / Gradle / Ivy

Go to download
package net.maizegenetics.analysis.numericaltransform;
/*
 * Plugin for imputation methods in Numerical Transformations.
 * @author - Janu Verma
 */

import net.maizegenetics.dna.snp.GenotypeTable;
import net.maizegenetics.dna.snp.GenotypeTableBuilder;
import net.maizegenetics.dna.snp.score.ReferenceProbabilityBuilder;
import net.maizegenetics.phenotype.Phenotype;
import net.maizegenetics.phenotype.PhenotypeAttribute;
import net.maizegenetics.phenotype.NumericAttribute;
import net.maizegenetics.phenotype.PhenotypeBuilder;
import net.maizegenetics.dna.snp.GenotypeTableUtils;
import net.maizegenetics.plugindef.AbstractPlugin;
import net.maizegenetics.plugindef.DataSet;
import net.maizegenetics.plugindef.Datum;
import net.maizegenetics.plugindef.PluginParameter;
import net.maizegenetics.util.OpenBitSet;

import java.awt.Frame;
import java.util.ArrayList;
import java.util.List;
import org.apache.log4j.Logger;

import javax.swing.*;

public class ImputationPlugin extends AbstractPlugin {

    private static final Logger myLogger = Logger.getLogger(ImputationPlugin.class);

    private enum distanceChoice {

        Euclidean, Manhattan, Cosine
    }

    private PluginParameter byMean = new PluginParameter.Builder<>("ByMean", false, Boolean.class)
            .guiName("Imputation by mean")
            .description("If imputation is performed by computing mean of the respective column")
            .build();

    private PluginParameter nearestNeighbors = new PluginParameter.Builder<>("nearestNeighbors", 5, Integer.class)
            .guiName("Number of nearest neighbors to be evaluated")
            .description("Choice of k in k-nearest neighbors algorithm. Default is 5.")
            .build();

    // Use enum to set the distance metric.
    private PluginParameter distance = new PluginParameter.Builder<>("distance", distanceChoice.Euclidean, distanceChoice.class)
            .guiName("Choose Distance type")
            .description("Distance choice for computing nearest neighbors. Default choice is Euclidean distance.")
            .build();

    /**
     * Creates a new instance of the ImputationPlugin
     */
    public ImputationPlugin(Frame parentFrame, boolean isInteractive) {
        super(parentFrame, isInteractive);
    }

    @Override
    protected void preProcessParameters(DataSet input) {
        if ((input == null) || (input.getSize() != 1)) {
            throw new IllegalArgumentException("ImputationPlugin: preProcessParameters: Please select one Genotype Table or Phenotype.");
        }
        List data = input.getDataOfType(new Class[]{GenotypeTable.class, Phenotype.class});
        if (data.size() != 1) {
            throw new IllegalArgumentException("ImputationPlugin: preProcessParameters: Please select one Genotype Table or Phenotype.");
        }
    }

    @Override
    public DataSet processData(DataSet input) {
        List datumList = input.getDataOfType(GenotypeTable.class);

        //check size of datumList, throw error if not equal to one
        if (datumList.size() != 1) { //not a GenotypeTable
            datumList = input.getDataOfType(Phenotype.class);
            if (datumList.size() != 1) {
                throw new IllegalArgumentException("ImputationPlugin: Input must me a genotype table or phenotype.");
            }
            Phenotype myPhenotype = (Phenotype) datumList.get(0).getData();

            // Only restrict to the numerical(data) attributes.
            // Indices of the data attributes.
            int[] dataAttrIndices = myPhenotype.attributeIndicesOfType(Phenotype.ATTRIBUTE_TYPE.data);

            //dimensions of the data matrix.
            int dataAttributes = dataAttrIndices.length;
            int dataObservations = myPhenotype.numberOfObservations();

            //Initialize the data matrix. This will be of size observations * data_attaributes.
            double[][] data = new double[dataObservations][dataAttributes];

            for (int s = 0; s < dataObservations; s++) {
                for (int t = 0; t < dataAttributes; t++) {
                    if (!(myPhenotype.isMissing(s, dataAttrIndices[t]))) {
                        data[s][t] = (Float) myPhenotype.value(s, dataAttrIndices[t]);
                    } else {
                        data[s][t] = Double.NaN;
                    }
                }
            }

            double[][] result;

            // Convert the input file into a matrix, data.
            Boolean imputationByMean;
            imputationByMean = by_mean();
            if (imputationByMean) {
                result = ImputationByMean.impute(data);
            } else {
                boolean isManhattan = false;
                boolean isCosine = false;
                distanceChoice alpha = distance_choice();

                if (alpha == distanceChoice.Manhattan) {
                    isManhattan = true;
                }
                if (alpha == distanceChoice.Cosine) {
                    isCosine = true;
                }
                Integer Nbrs = nearest_neighbors();
                result = kNearestNeighbors.impute(data, Nbrs, isManhattan, isCosine);
            }

            List attributes = new ArrayList<>();
            List types = new ArrayList<>();
            attributes.add(myPhenotype.taxaAttribute());
            types.add(Phenotype.ATTRIBUTE_TYPE.taxa);

            attributes.addAll(myPhenotype.attributeListOfType(Phenotype.ATTRIBUTE_TYPE.factor));
            for (int i = 0; i < myPhenotype.numberOfAttributesOfType(Phenotype.ATTRIBUTE_TYPE.factor); i++) {
                types.add(Phenotype.ATTRIBUTE_TYPE.factor);
            }

            attributes.addAll(myPhenotype.attributeListOfType(Phenotype.ATTRIBUTE_TYPE.covariate));
            for (int i = 0; i < myPhenotype.numberOfAttributesOfType(Phenotype.ATTRIBUTE_TYPE.covariate); i++) {
                types.add(Phenotype.ATTRIBUTE_TYPE.covariate);
            }

            for (int i = 0; i < dataAttributes; i++) {
                float[] attrData = new float[dataObservations];
                for (int j = 0; j < dataObservations; j++) {
                    attrData[j] = (float) result[j][i];
                }
                PhenotypeAttribute oldAttribute = myPhenotype.attribute(dataAttrIndices[i]);
                NumericAttribute myAttribute = new NumericAttribute(oldAttribute.name(), attrData, new OpenBitSet(dataObservations));
                attributes.add(myAttribute);
                types.add(Phenotype.ATTRIBUTE_TYPE.data);
            }

            Phenotype imputedPhenotype = new PhenotypeBuilder().fromAttributeList(attributes, types).build().get(0);

            StringBuilder nameBuilder = new StringBuilder("Imputed_");
            nameBuilder.append(datumList.get(0).getName());
            StringBuilder commentBuilder = new StringBuilder(getMethodString());
            commentBuilder.append("\nfrom ").append(datumList.get(0).getName());
            Datum newDatum = new Datum(nameBuilder.toString(), imputedPhenotype, commentBuilder.toString());
            return new DataSet(newDatum, this);

        } else { //it is a GenotypeTable

            GenotypeTable myGenotype = (GenotypeTable) datumList.get(0).getData();

            int nsites = myGenotype.numberOfSites();
            int ntaxa = myGenotype.numberOfTaxa();

            //ReferenceProbability myProb;
            double[][] data;
            if (!myGenotype.hasReferenceProbablity()) {
                data = GenotypeTableUtils.convertGenotypeToDoubleProbability(myGenotype, true);
            } else {
                //myProb = myGenotype.referenceProbability();
                data = new double[nsites][ntaxa];
                for (int s = 0; s < nsites; s++) {
                    for (int t = 0; t < ntaxa; t++) {
                        data[s][t] = myGenotype.referenceProbability(t, s);
                    }
                }
            }

            double[][] result;

            // Convert the input file into a matrix, data.
            Boolean imputationByMean;
            imputationByMean = by_mean();
            if (imputationByMean) {
                result = ImputationByMean.impute(data);
            } else {
                boolean isManhattan = false;
                boolean isCosine = false;
                distanceChoice alpha = distance_choice();

                if (alpha == distanceChoice.Manhattan) {
                    isManhattan = true;
                }
                if (alpha == distanceChoice.Cosine) {
                    isCosine = true;
                }
                Integer Nbrs = nearest_neighbors();
                result = kNearestNeighbors.impute(data, Nbrs, isManhattan, isCosine);
            }

            //build new ReferenceProbability
            ReferenceProbabilityBuilder refBuilder = ReferenceProbabilityBuilder.getInstance(ntaxa, nsites, myGenotype.taxa());
            for (int t = 0; t < ntaxa; t++) {
                float[] values = new float[nsites];

                for (int s = 0; s < nsites; s++) {
                    values[s] = (float) result[s][t];
                }
                refBuilder.addTaxon(t, values);
            }

            //build new GenotypeTable
            GenotypeTable imputedGenotype = GenotypeTableBuilder.getInstance(myGenotype.genotypeMatrix(), myGenotype.positions(),
                    myGenotype.taxa(), myGenotype.depth(), myGenotype.alleleProbability(), refBuilder.build(), myGenotype.dosage(),
                    myGenotype.annotations());

            StringBuilder nameBuilder = new StringBuilder("Imputed_");
            nameBuilder.append(datumList.get(0).getName());
            StringBuilder commentBuilder = new StringBuilder(getMethodString());
            commentBuilder.append("\nfrom ").append(datumList.get(0).getName());
            Datum newDatum = new Datum(nameBuilder.toString(), imputedGenotype, commentBuilder.toString());
            return new DataSet(newDatum, this);
        }
    }

    private String getMethodString() {
    	if (byMean.value()) return "Missing values imputed as mean of trait.";
    	return String.format("Missing values imputed using K-nearest neigbor with %s distance.", distance.value().name());
    }
    
    @Override
    public String pluginDescription() {
        return "This plugin takes an input file (genotype/phenotype) with missing values"
                + "and imputes the missing values using one the chosen methods."
                + "It returns the imputed file.";
    }

    @Override
    public ImageIcon getIcon() {
        return null;
    }

    @Override
    public String getButtonName() {
        return "Numerical Impute";
    }

    @Override
    public String getToolTipText() {
        return "Numerical Impute";
    }

    public distanceChoice distance_choice() {
        return distance.value();
    }

    public ImputationPlugin distance_choice(distanceChoice value) {
        distance = new PluginParameter<>(distance, value);
        return this;
    }

    public Boolean by_mean() {
        return byMean.value();
    }

    public ImputationPlugin by_mean(Boolean value) {
        byMean = new PluginParameter<>(byMean, value);
        return this;
    }

    public Integer nearest_neighbors() {
        return nearestNeighbors.value();
    }

    public ImputationPlugin nearest_neighbors(Integer value) {
        nearestNeighbors = new PluginParameter<>(nearestNeighbors, value);
        return this;
    }

}