laydi/workflows/affy_workflow.py

import gtk
from system import dataset, logger, plots, workflow
from scipy import randn, array, transpose, zeros
import cPickle


class AffyWorkflow (workflow.Workflow):

    name = 'Affy Workflow'
    ident = 'affy'
    description = 'Affymetrics Workflow. Analysis of Affy-data.'
    def __init__(self, app):
        workflow.Workflow.__init__(self, app)

        load = workflow.Stage('load', 'Load Data')
        load.add_function(CelFileImportFunction())
        load.add_function(PhenotypeImportFunction())
        load.add_function(TestDataFunction())
        load.add_function(DatasetLoadFunction())
        self.add_stage(load)

        explore = workflow.Stage('explore', 'Explorative analysis')
        explore.add_function(PCAFunction(self))
        explore.add_function(PrintFunction())
        self.add_stage(explore)

        save = workflow.Stage('save', 'Save Data')
        save.add_function(DatasetSaveFunction())
        self.add_stage(save)

class PrintFunction(workflow.Function):
    def __init__(self):
        workflow.Function.__init__(self, 'printer', 'Print Stuff')

    def run(self, data):
        dim1, dim2 = data.get_dim_names()
        print dim1, dim2
        print "\t", "\t".join(data.get_identifiers(dim2))
        for row in zip(data.get_identifiers(dim1), data.asarray().tolist()):
            print "\t".join(map(str, row))
        

class TestDataFunction(workflow.Function):  
    def __init__(self):
        workflow.Function.__init__(self, 'test_data', 'Generate Test Data')

    def run(self):
        logger.log('notice', 'Injecting foo test data')
        x = randn(20,30)
        X = dataset.Dataset(x)
        return [X, plots.LinePlot(X)]


class DatasetLoadFunction(workflow.Function):
    """Loader for previously pickled Datasets."""
    def __init__(self):
        workflow.Function.__init__(self, 'load_data', 'Load Pickled Dataset')

    def run(self):
        chooser = gtk.FileChooserDialog(title="Select cel files...", parent=None,
                                        action=gtk.FILE_CHOOSER_ACTION_OPEN,
                                        buttons=(gtk.STOCK_CANCEL, gtk.RESPONSE_CANCEL,
                                                 gtk.STOCK_OPEN, gtk.RESPONSE_OK))
        pkl_filter = gtk.FileFilter()
        pkl_filter.set_name("Python pickled data files (*.pkl)")
        pkl_filter.add_pattern("*.[pP][kK][lL]")
        all_filter = gtk.FileFilter()
        all_filter.set_name("All Files (*.*)")
        all_filter.add_pattern("*")
        chooser.add_filter(pkl_filter)
        chooser.add_filter(all_filter)

        try:
            if chooser.run() == gtk.RESPONSE_OK:
                return [cPickle.load(open(chooser.get_filename()))]
        finally:
            chooser.destroy()


class DatasetSaveFunction(workflow.Function):
    """QND way to save data to file for later import to this program."""
    def __init__(self):
        workflow.Function.__init__(self, 'save_data', 'Save Pickled Dataset')

    def run(self):
        if not data:
            logger.log("notice", "No data to save.")
            return
        else:
            data = data[0]

        chooser = gtk.FileChooserDialog(title="Save pickled data...", parent=None,
                                        action=gtk.FILE_CHOOSER_ACTION_SAVE,
                                        buttons=(gtk.STOCK_CANCEL, gtk.RESPONSE_CANCEL,
                                                 gtk.STOCK_SAVE, gtk.RESPONSE_OK))
        pkl_filter = gtk.FileFilter()
        pkl_filter.set_name("Python pickled data files (*.pkl)")
        pkl_filter.add_pattern("*.[pP][kK][lL]")
        all_filter = gtk.FileFilter()
        all_filter.set_name("All Files (*.*)")
        all_filter.add_pattern("*")
        chooser.add_filter(pkl_filter)
        chooser.add_filter(all_filter)
        chooser.set_current_name(data.get_name() + ".pkl")

        try:
            if chooser.run() == gtk.RESPONSE_OK:
                cPickle.dump(data, open(chooser.get_filename(), "w"), protocol=2)
                logger.log("notice", "Saved data to %r." % chooser.get_filename())
        finally:
            chooser.destroy()
                

class CelFileImportFunction(workflow.Function):
    """Loads Affymetrics .CEL-files into matrix."""
    def __init__(self):
        workflow.Function.__init__(self, 'cel_import', 'Import Affy')

    def run(self,data):
        import rpy
        chooser = gtk.FileChooserDialog(title="Select cel files...", parent=None,
                                        action=gtk.FILE_CHOOSER_ACTION_OPEN,
                                        buttons=(gtk.STOCK_CANCEL, gtk.RESPONSE_CANCEL,
                                                 gtk.STOCK_OPEN, gtk.RESPONSE_OK))
        chooser.set_select_multiple(True)
        cel_filter = gtk.FileFilter()
        cel_filter.set_name("Cel Files (*.cel)")
        cel_filter.add_pattern("*.[cC][eE][lL]")
        all_filter = gtk.FileFilter()
        all_filter.set_name("All Files (*.*)")
        all_filter.add_pattern("*")
        chooser.add_filter(cel_filter)
        chooser.add_filter(all_filter)

        try:
            if chooser.run() == gtk.RESPONSE_OK:
                rpy.r.library("affy")
    
                silent_eval = rpy.with_mode(rpy.NO_CONVERSION, rpy.r)
                silent_eval('E <- ReadAffy(filenames=c("%s"))' % '", "'.join(chooser.get_filenames()))
                silent_eval('E <- rma(E)')

                m = rpy.r('m <- E@exprs')
    
                vector_eval = rpy.with_mode(rpy.VECTOR_CONVERSION, rpy.r)
                rownames = vector_eval('rownames(m)') 
                colnames = vector_eval('colnames(m)') 

                # We should be nice and clean up after ourselves
                rpy.r.rm(["E", "m"])
                
                if m:
                    data = dataset.Dataset(m, (('ids', rownames), ('filename', colnames)), name="Affymetrics Data")
                    plot = plots.LinePlot(data, "Gene profiles")
                    return [data, plot]
                else:
                    logger.log("notice", "No data loaded from importer.")
        finally:
            chooser.destroy()


class PhenotypeImportFunction(workflow.Function):
    def __init__(self):
        workflow.Function.__init__(self, 'import_phenotype', 'Import Phenotypes')

    def run(self):
        chooser = gtk.FileChooserDialog(title="Select cel files...", parent=None,
                                        action=gtk.FILE_CHOOSER_ACTION_OPEN,
                                        buttons=(gtk.STOCK_CANCEL, gtk.RESPONSE_CANCEL,
                                                 gtk.STOCK_OPEN, gtk.RESPONSE_OK))
        all_filter = gtk.FileFilter()
        all_filter.set_name("Tab separated file (*.*)")
        all_filter.add_pattern("*")
        chooser.add_filter(all_filter)

        try:
            if chooser.run() == gtk.RESPONSE_OK:
                text = open(chooser.get_filename()).read()
                data = PhenotypeDataset(text)
                return [data]
                
        finally:
            chooser.destroy()
        

class PCAFunction(workflow.Function):
    """Generic PCA function."""
    def __init__(self, wf):
        workflow.Function.__init__(self, 'pca', 'PCA')
        self._workflow = wf

    def run(self,data):
        import rpy
        
        dim_2, dim_1 = data.get_dim_names()
    
        silent_eval = rpy.with_mode(rpy.NO_CONVERSION, rpy.r)
        rpy.with_mode(rpy.NO_CONVERSION, rpy.r.assign)("m", data.get_matrix())
        silent_eval("t = prcomp(t(m))")

        # we make a unique name for component dimension
        c = 0
        component_dim = prefix = "component"
        while component_dim in data.get_all_dims():
            component_dim = prefix + "_" + str(c)
            c += 1

        T_ids = map(str, range(1, rpy.r("dim(t$x)")[1]+1))
        T = dataset.Dataset(rpy.r("t$x"), [(dim_1, data.get_identifiers(dim_1)),
                                           (component_dim, T_ids)],
                            all_dims = data.get_all_dims(), name="T")
        P = dataset.Dataset(rpy.r("t$rotation"), [(dim_2, data.get_identifiers(dim_2)),
                                                  (component_dim, T_ids)],
                            all_dims = data.get_all_dims(), name="P")
        # cleanup
        rpy.r.rm(["t", "m"])

        loading_plot = plots.ScatterPlot(P, dim_2, component_dim, '1', '2',
                                         "Loadings")
        score_plot = plots.ScatterPlot(T, dim_1,component_dim, '1', '2',
                                       "Scores")
        
        return [T, P, loading_plot, score_plot]


class PhenotypeDataset(dataset.Dataset):
    def __init__(self, string):
        self._table = rows = [line.split("\t") for line in string.splitlines()]
        columns = zip(*rows[1:])
        cel_names = columns[0]
        col_names = rows[0][1:]
        phenotypes = []
        categories = {}

        for col_name, column in zip(col_names, columns[1:]):
            try:
                categories[col_name] = map(float, column)
                phenotypes.append(col_name)
            except ValueError:
                # category-data
                keys = []
                entries = {}
                for i, entry in enumerate(column):
                    if entry not in entries:
                        keys.append(entry)
                        entries[entry] = []

                    entries[entry].append(i)
                    

                for key in keys:
                    z = zeros(len(column))
                    for i in entries[key]:
                        z[i] = 1
                    key = "%s-%s" % (col_name, key)
                    phenotypes.append(key)
                    categories[key] = z

        matrix_data = []
        for key in phenotypes:
            matrix_data.append(categories[key])

        if matrix_data:
            a = transpose(array(matrix_data))
        else:
            a = None

        dataset.Dataset.__init__(self, a, identifiers=[('CEL', cel_names),
                                                       ('phenotypes', phenotypes)],
                                 shape=(len(cel_names),len(phenotypes)), name="Phenotype Data")
            

    def get_phenotype_table(self):
        return self._table