def create_pathway_index_html(data, fitter, fits, basedir, gene_dir,
series_dir, use_correlations, b_unique):
dct_pathways = load_17_pathways_breakdown(b_unique)
n_ranks = 3 # actually we'll have ranks of 0 to n_ranks
flat_fits = {} # (gene,region) -> fit (may be None)
for g in data.gene_names:
for r in data.region_names:
flat_fits[(g, r)] = None
for dsfits in fits.itervalues():
for (g, r), fit in dsfits.iteritems():
if use_correlations:
score = fit.with_correlations.LOO_score
else:
score = fit.LOO_score
fit.score = score
fit.rank = int(np.ceil(n_ranks * score)) if score > 0 else 0
flat_fits[(g, r)] = fit
html = get_jinja_env().get_template('pathways.jinja').render(**locals())
str_unique = '-unique' if b_unique else ''
filename = 'pathway-fits{}.html'.format(str_unique)
with open(join(basedir, filename), 'w') as f:
f.write(html)
def save_fits_and_create_html(data, fitter, fits=None, basedir=None,
do_genes=True, do_series=True, do_hist=True, do_html=True, only_main_html=False,
k_of_n=None,
use_correlations=False, correlations=None,
show_change_distributions=False,
html_kw=None,
figure_kw=None):
if fits is None:
fits = get_all_fits(data,fitter,k_of_n)
if basedir is None:
basedir = join(results_dir(), fit_results_relative_path(data,fitter))
if use_correlations:
basedir = join(basedir,'with-correlations')
if html_kw is None:
html_kw = {}
if figure_kw is None:
figure_kw = {}
print 'Writing HTML under {}'.format(basedir)
ensure_dir(basedir)
gene_dir = 'gene-subplot'
series_dir = 'gene-region-fits'
correlations_dir = 'gene-correlations'
scores_dir = 'score_distributions'
if do_genes and not only_main_html: # relies on the sharding of the fits respecting gene boundaries
plot_and_save_all_genes(data, fitter, fits, join(basedir,gene_dir), show_change_distributions)
if do_series and not only_main_html:
plot_and_save_all_series(data, fitter, fits, join(basedir,series_dir), use_correlations, show_change_distributions, figure_kw)
if do_hist and k_of_n is None and not only_main_html:
create_score_distribution_html(fits, use_correlations, join(basedir,scores_dir))
if do_html and k_of_n is None:
link_to_correlation_plots = use_correlations and correlations is not None
if link_to_correlation_plots and not only_main_html:
plot_and_save_all_gene_correlations(data, correlations, join(basedir,correlations_dir))
dct_pathways = load_17_pathways_breakdown()
pathway_genes = set.union(*dct_pathways.values())
data_genes = set(data.gene_names)
missing = pathway_genes - data_genes
b_pathways = len(missing) < len(pathway_genes)/2 # simple heuristic to create pathways only if we have most of the genes (currently 61 genes are missing)
create_html(
data, fitter, fits, basedir, gene_dir, series_dir, scores_dir, correlations_dir=correlations_dir,
use_correlations=use_correlations, link_to_correlation_plots=link_to_correlation_plots,
b_pathways=b_pathways, **html_kw
)
return fig
cfg.verbosity = 1
age_scaler = LogScaler()
all_data = GeneData.load('both').scale_ages(age_scaler)
pathway = '17full'
data = GeneData.load('both').restrict_pathway(pathway).scale_ages(age_scaler)
shape = Sigmoid(priors='sigmoid_wide')
fitter = Fitter(shape, sigma_prior='normal')
fits = get_all_fits(data, fitter)
R2_threshold = 0.5
for b_unique in [False, True]:
dct_pathways = load_17_pathways_breakdown(b_unique)
dct_pathways['17 pathways'] = None
for name, genes in dct_pathways.iteritems():
fig = plot_onset_times(all_data, data, fitter, fits, {name: genes},
R2_threshold, b_unique)
str_dir = 'unique' if b_unique else 'overlapping'
str_unique = ' (unique)' if b_unique else ''
filename = 'RP/{}/change-distributions-{}{}.png'.format(
str_dir, name, str_unique)
save_figure(fig, filename, under_results=True)
# selected plots
lst_pathways = [
'17 pathways', 'Amphetamine addiction', 'Cholinergic synapse',
'Cocaine addiction', 'Glutamatergic synapse'
]
def save_fits_and_create_html(data,
fitter,
fits=None,
basedir=None,
do_genes=True,
do_series=True,
do_hist=True,
do_html=True,
only_main_html=False,
k_of_n=None,
use_correlations=False,
correlations=None,
show_change_distributions=False,
exons_layout=False,
html_kw=None,
figure_kw=None):
if fits is None:
fits = get_all_fits(data, fitter, k_of_n)
if basedir is None:
basedir = join(results_dir(), fit_results_relative_path(data, fitter))
if use_correlations:
basedir = join(basedir, 'with-correlations')
if html_kw is None:
html_kw = {}
if figure_kw is None:
figure_kw = {}
print 'Writing HTML under {}'.format(basedir)
ensure_dir(basedir)
gene_dir = 'gene-subplot'
exons_dir = 'exons_subplot_series' if cfg.exons_plots_from_series else 'exons_subplot'
series_dir = 'gene-region-fits'
correlations_dir = 'gene-correlations'
scores_dir = 'score_distributions'
if do_genes and not only_main_html: # relies on the sharding of the fits respecting gene boundaries
plot_and_save_all_genes(data, fitter, fits, join(basedir, gene_dir),
show_change_distributions)
if do_series and not only_main_html:
plot_and_save_all_series(data, fitter, fits, join(basedir, series_dir),
use_correlations, show_change_distributions,
exons_layout, figure_kw)
if exons_layout and not only_main_html:
if cfg.exons_plots_from_series:
plot_and_save_all_exons_from_series(fits, join(basedir, exons_dir),
join(basedir, series_dir))
else:
plot_and_save_all_exons(data, fitter, fits,
join(basedir, exons_dir))
if do_hist and k_of_n is None and not only_main_html:
create_score_distribution_html(fits, use_correlations,
join(basedir, scores_dir))
if do_html and k_of_n is None:
link_to_correlation_plots = use_correlations and correlations is not None
if link_to_correlation_plots and not only_main_html:
plot_and_save_all_gene_correlations(
data, correlations, join(basedir, correlations_dir))
dct_pathways = load_17_pathways_breakdown()
pathway_genes = set.union(*dct_pathways.values())
data_genes = set(data.gene_names)
missing = pathway_genes - data_genes
b_pathways = len(missing) < len(
pathway_genes
) / 2 # simple heuristic to create pathways only if we have most of the genes (currently 61 genes are missing)
create_html(data,
fitter,
fits,
basedir,
gene_dir,
exons_dir,
series_dir,
scores_dir,
correlations_dir=correlations_dir,
use_correlations=use_correlations,
link_to_correlation_plots=link_to_correlation_plots,
b_pathways=b_pathways,
exons_layout=exons_layout,
**html_kw)
def create_pathway_index_html(data, fitter, fits, basedir, gene_dir, series_dir, use_correlations, b_unique):
from jinja2 import Template
dct_pathways = load_17_pathways_breakdown(b_unique)
n_ranks = 3 # actually we'll have ranks of 0 to n_ranks
flat_fits = {} # (gene,region) -> fit (may be None)
for g in data.gene_names:
for r in data.region_names:
flat_fits[(g,r)] = None
for dsfits in fits.itervalues():
for (g,r),fit in dsfits.iteritems():
if use_correlations:
score = fit.with_correlations.LOO_score
else:
score = fit.LOO_score
fit.score = score
fit.rank = int(np.ceil(n_ranks * score)) if score > 0 else 0
flat_fits[(g,r)] = fit
html = Template("""
<html>
<head>
<link rel="stylesheet" type="text/css" href="fits.css">
</head>
<body>
<H1>Fits broken down by pathway {% if b_unique %} (unique genes only) {% endif %} </H1>
{% for pathway_name, pathway_genes in dct_pathways.iteritems() %}
<P>
<H2>{{pathway_name}}</H2>
<table>
<th>
{% for region_name in data.region_names %}
<td class="tableHeading">
<b>{{region_name}}</b>
</td>
{% endfor %}
</th>
{% for gene_name in pathway_genes %}
<tr>
<td>
<a href="{{gene_dir}}/{{gene_name}}.png"><b>{{gene_name}}</b></a>
</td>
{% for region_name in data.region_names %}
<td>
{% if flat_fits[(gene_name,region_name)] %}
<a href="{{series_dir}}/fit-{{gene_name}}-{{region_name}}.png">
{% if flat_fits[(gene_name,region_name)].score %}
<div class="score rank{{flat_fits[(gene_name,region_name)].rank}}">
{{flat_fits[(gene_name,region_name)].score | round(2)}}
</div>
{% else %}
No Score
{% endif %}
</a>
{% endif %}
</td>
{% endfor %}
</tr>
{% endfor %}
</table>
</P>
{% endfor %} {# dct_pathways #}
</body>
</html>
""").render(**locals())
str_unique = '-unique' if b_unique else ''
filename = 'pathway-fits{}.html'.format(str_unique)
with open(join(basedir,filename), 'w') as f:
f.write(html)
return fig
cfg.verbosity = 1
age_scaler = LogScaler()
all_data = GeneData.load('both').scale_ages(age_scaler)
pathway = '17full'
data = GeneData.load('both').restrict_pathway(pathway).scale_ages(age_scaler)
shape = Sigmoid(priors='sigmoid_wide')
fitter = Fitter(shape, sigma_prior='normal')
fits = get_all_fits(data, fitter)
R2_threshold = 0.5
for b_unique in [False,True]:
dct_pathways = load_17_pathways_breakdown(b_unique)
dct_pathways['17 pathways'] = None
for name,genes in dct_pathways.iteritems():
fig = plot_onset_times(all_data, data, fitter, fits, {name:genes}, R2_threshold, b_unique)
str_dir = 'unique' if b_unique else 'overlapping'
str_unique = ' (unique)' if b_unique else ''
filename = 'RP/{}/change-distributions-{}{}.png'.format(str_dir,name,str_unique)
save_figure(fig, filename, under_results=True)
# selected plots
lst_pathways = ['17 pathways', 'Amphetamine addiction', 'Cholinergic synapse', 'Cocaine addiction', 'Glutamatergic synapse']
dct_pathways = {k:dct_pathways[k] for k in lst_pathways}
fig = plot_onset_times(all_data, data, fitter, fits, dct_pathways, R2_threshold, b_unique)
str_dir = 'unique' if b_unique else 'overlapping'
str_unique = ' (unique)' if b_unique else ''
filename = 'RP/{}/selected-change-distributions{}.png'.format(str_dir,str_unique)
