feat_df, meta_df, idx, gene_list = select_strains([candidate_gene],
CONTROL_STRAIN,
feat_df=feat,
meta_df=meta)
# filter features
feat_df, meta_df, featsets = filter_features(feat_df,
meta_df)
# strain_numbers.append(meta_df.groupby('worm_strain')['file_id_prestim'].describe()['count'])
strain_lut, stim_lut, feat_lut = make_colormaps(gene_list,
featlist=featsets['all'],
idx=idx,
candidate_gene=[candidate_gene],
)
# colorbars to map colors to strains
plot_colormap(strain_lut)
plt.savefig(saveto / 'strain_cmap.png')
plot_cmap_text(strain_lut)
plt.savefig(saveto / 'strain_cmap_text.png')
plot_colormap(stim_lut, orientation='horizontal')
plt.savefig(saveto / 'stim_cmap.png')
plot_cmap_text(stim_lut)
plt.savefig(saveto / 'stim_cmap_text.png')
plt.close('all')
genes = [g for g in genes if 'myo' not in g and 'unc-54' not in g]
#%%
for g in genes:
print('making ts stimuli plots for {}'.format(g))
candidate_gene = g
timeseries_fname = RAW_DATA_DIR / 'Results' / '{}_timeseries.hdf5'.format(
candidate_gene)
(saveto / candidate_gene).mkdir(exist_ok=True)
# only select strains of interest
meta, idx, gene_list = select_strains(candidate_gene,
CONTROL_STRAIN,
meta_df=meta_ts)
strain_lut, stim_lut = make_colormaps(gene_list, [], idx,
candidate_gene)
if g == 'figo-1':
strain_lut['C43B7.2'] = strain_lut['figo-1']
timeseries_df = pd.read_hdf(timeseries_fname, 'timeseries_df')
# hires_df = pd.read_hdf(timeseries_fname, 'hires_df')
# test_feats = [f for f in timeseries_df.columns if 'radial_velocity' in f]
# t = ['d_relative_to_neck_radial_velocity_head_tip',
# 'relative_to_neck_radial_velocity_head_tip']
for k in strain_lut.keys():
_ts_df = timeseries_df.query('@k in worm_gene')
subset=['worm_gene'],
inplace=True)
# remove data from dates to exclude
good_date = meta.query('@DATES_TO_DROP not in imaging_date_yyyymmdd').index
# bad wells
good_wells_from_gui = meta.query('is_bad_well == False').index
meta = meta.loc[good_wells_from_gui & good_date,:]
# only select strains of interest
meta, idx, gene_list = select_strains(CANDIDATE_GENE,
CONTROL_STRAIN,
meta_df=meta,
feat_df=None)
strain_lut, stim_lut = make_colormaps(gene_list,
idx,
CANDIDATE_GENE,
CONTROL_STRAIN,
featlist=[])
#strain to gene dictionary
strain_dict = strain_gene_dict(meta)
gene_dict = {v:k for k,v in strain_dict.items()}
meta = align_bluelight_meta(meta)
if is_reload_timeseries_from_results:
# this uses tierpytools under the hood
timeseries_df, hires_df = load_bluelight_timeseries_from_results(
meta,
RAW_DATA_DIR / 'Results')
# save to disk
timeseries_df.to_hdf(timeseries_fname, 'timeseries_df', format='table')
for f in list(zip(window_feat_files, window_fname_files)))
genes = [g for g in meta.worm_gene.unique() if g != CONTROL_STRAIN]
# genes = [g.replace('C43B7.2', 'figo-1') for g in genes]
genes = [g for g in genes if 'myo' not in g and 'unc-54' not in g]
#%%
feat_df, meta_df, idx, gene_list = select_strains(STRAINS,
CONTROL_STRAIN,
feat_df=feat,
meta_df=meta)
# filter features
feat_df, meta_df, featsets = filter_features(feat_df, meta_df)
strain_lut, stim_lut, feat_lut = make_colormaps(gene_list,
featlist=featsets['all'],
idx=idx,
candidate_gene=STRAINS)
# colorbars to map colors to strains
plot_colormap(strain_lut)
plt.savefig(saveto / 'strain_cmap.png')
plot_cmap_text(strain_lut)
plt.savefig(saveto / 'strain_cmap_text.png')
plot_colormap(stim_lut, orientation='horizontal')
plt.savefig(saveto / 'stim_cmap.png')
plot_cmap_text(stim_lut)
plt.savefig(saveto / 'stim_cmap_text.png')
plt.close('all')
# strain_numbers = []
#%%
for count,g in enumerate(genes):
print('Analysing {} {}/{}'.format(g, count+1, len(genes)))
candidate_gene = g
(saveto / candidate_gene).mkdir(exist_ok=True)
meta_df, idx, gene_list = select_strains(candidate_gene,
CONTROL_STRAIN,
meta_df=meta)
strain_lut, stim_lut = make_colormaps(gene_list,
featlist=[],
idx=idx,
candidate_gene=candidate_gene,
)
subsample = meta_df.groupby('worm_gene').sample(n_subsample,
random_state=seed)
for k,v in strain_lut.items():
_meta = subsample.query('@k in `worm_gene`')
for i,r in _meta.iterrows():
_imgstore = RAW_DATA_DIR / 'Results' / r['imgstore_name'] / 'metadata_featuresN.hdf5'
with pd.HDFStore(_imgstore) as fid:
ts = fid['/timeseries_data'].query("`well_name` == @r.well_name")