test_study.interactive_pca() does not display graphs. Displays errors.

[eliah-o]

A fresh copy of flotilla was installed. %matplotlib inline, import flotilla, and test_study = flotilla.embark('http://sauron.ucsd.edu/flotilla_projects/neural_diff_chr22/datapackage.json') were all run. Then test_study.interactive_pca() and the following stack trace is printed:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/Users/eliah/anaconda/envs/flotilla_env/lib/python2.7/site-packages/flotilla/visualize/ipython_interact.pyc in do_interact(data_type, sample_subset, feature_subset, featurewise, list_link, x_pc, y_pc, show_point_labels, savefile)
    131                                 x_pc=x_pc, y_pc=y_pc,
    132                                 show_point_labels=show_point_labels,
--> 133                                 feature_subset=feature_subset)
    134             if savefile != '':
    135                 # Make the directory if it's not already there

/Users/eliah/anaconda/envs/flotilla_env/lib/python2.7/site-packages/flotilla/data_model/study.pyc in plot_pca(self, data_type, x_pc, y_pc, sample_subset, feature_subset, title, featurewise, show_point_labels, reduce_kwargs, **kwargs)
    810                 order=order, color=color,
    811                 featurewise=featurewise, show_point_labels=show_point_labels,
--> 812                 title=title, reduce_kwargs=reduce_kwargs, **kwargs)
    813         elif "splicing".startswith(data_type):
    814             reducer = self.splicing.plot_pca(

/Users/eliah/anaconda/envs/flotilla_env/lib/python2.7/site-packages/flotilla/data_model/base.pyc in plot_pca(self, **kwargs)
    377     def plot_pca(self, **kwargs):
    378         return self.plot_dimensionality_reduction(reducer=DataFramePCA,
--> 379                                                   **kwargs)
    380 
    381     @property

/Users/eliah/anaconda/envs/flotilla_env/lib/python2.7/site-packages/flotilla/data_model/base.pyc in plot_dimensionality_reduction(self, x_pc, y_pc, sample_ids, feature_ids, featurewise, reducer, label_to_color, label_to_marker, groupby, order, color, reduce_kwargs, title, **plotting_kwargs)
    360         reduced = self.reduce(sample_ids, feature_ids,
    361                               featurewise=featurewise,
--> 362                               reducer=reducer, **reduce_kwargs)
    363         visualized = DecompositionViz(reduced.reduced_space,
    364                                       reduced.components_,

/Users/eliah/anaconda/envs/flotilla_env/lib/python2.7/site-packages/flotilla/util.pyc in memoizer(*args, **kwargs)
     94         key = str(args) + str(kwargs)
     95         if key not in cache:
---> 96             cache[key] = obj(*args, **kwargs)
     97         return cache[key]
     98 

/Users/eliah/anaconda/envs/flotilla_env/lib/python2.7/site-packages/flotilla/data_model/base.pyc in reduce(self, sample_ids, feature_ids, featurewise, reducer, standardize, reducer_kwargs, bins)
    595             subset = subset.T
    596 
--> 597         reducer_object = reducer(subset, **reducer_kwargs)
    598         reducer_object.means = means
    599         return reducer_object

/Users/eliah/anaconda/envs/flotilla_env/lib/python2.7/site-packages/flotilla/compute/decomposition.pyc in __init__(self, df, n_components, **decomposer_kwargs)
     18         if df.shape[1] <= 3:
     19             raise ValueError(
---> 20                 "Too few features (n={}) to reduce".format(df.shape[1]))
     21         super(DataFrameReducerBase, self).__init__(n_components=n_components,
     22                                                    **decomposer_kwargs)

ValueError: Too few features (n=2) to reduce

[olgabot]

Try using "all genes" in the feature_subset dropdown. The default is "variant", which is too small for this tiny data subset

[olgabot]

Also can you reformat your comment using github flavored markdown? It makes the code much easier to read.

[eliah-o]

Using "all genes" works. Thanks!