def submit(self,gene_list,gene_file=None,**kwargs) :
if gene_list == 'example' :
example_fn = resource_filename('adipo_sight','data/p65_bound_gene_list.txt')
gene_list = open(example_fn).read()
self.parse_kwargs(kwargs)
log.debug(self.args.get('gene_file'))
cherrypy.log('gene_file:%s'%str(gene_file))
if (gene_file is not None and
gene_file.file is not None) :
gene_list_str = gene_file.file.read()
else :
gene_list_str = gene_list
gene_list_strip = re.sub(r'\s+',' ',gene_list_str)
cherrypy.log('user submitted genes: %s'%gene_list_strip)
gene_list = gene_list_strip.split()
# make sure the genes are unique
gene_list = list(set(gene_list))
orig_gene_list = gene_list
# make all genes lower case as they are in the db
gene_list = [g.lower() for g in gene_list]
# sort both the original and lower()ed gene lists in parallel,
# put them back
both_gene_lists = zip(gene_list,orig_gene_list)
both_gene_lists.sort()
gene_list, orig_gene_list = zip(*both_gene_lists)
self.gene_name_map = dict(zip(gene_list,orig_gene_list)+zip(orig_gene_list,gene_list))
hid = str(hash(''.join(gene_list)))
session_d = self.sessions.get(hid)
session_d['hid'] = hid
session_d['gene_list'] = gene_list
session_d['orig_gene_list'] = orig_gene_list
session_d['gene_name_map'] = self.gene_name_map
self.gene_list = gene_list
self.orig_gene_list = orig_gene_list
self.sessions.save(hid,session_d)
# now that we've set everything, redirect to the processing page
raise cherrypy.HTTPRedirect("http://fraenkel.mit.edu/adipo_sight/processing?hid=%s"%hid,status=303)
def run(self) :
session_d = self.session_d
if session_d is None :
log.error('hid passed but no session found, aborting')
return 'Error occurred, please try submitting again'
gene_list = session_d['gene_list']
d = {}
# get motif scores
cherrypy.log('getting motif scores')
# sqlite has a limitation of 999 SQL variables, need to split up
# this query
batch_size = 500
gene_list_batches = [gene_list[i:i+500] for i in xrange(0,len(gene_list),500)]
cherrypy.log('gene_list_batches len()s: %s'%str([len(b) for b in gene_list_batches]))
scores = []
for batch in gene_list_batches :
score_q = (self.db_session.query(db.Region,db.SeqData)
.join(db.RegionMembership)
.join(db.RegionSet)
.join(db.SeqData)
.filter(db.SeqData.seq_type.has(db.SeqType.name=='motif scores'))
.filter(db.RegionSet.name.in_(batch))
.filter(db.RegionMembership.dist_to_feature.between(
-int(self.args['upstream']),
int(self.args['downstream'])
))
)
scores.extend(score_q.all())
score_mat = []
gene_names = set()
condition_scores = defaultdict(list)
cherrypy.log('loading motif scores')
for region, seqdata in scores :
for region_membership in region.membership :
region_set = region_membership.region_set
gene_names.add(region_set.name)
condition = seqdata.condition.name
motif_scores = cPickle.loads(seqdata.value)
condition_scores[condition].append(motif_scores)
score_mat.append(motif_scores)
cherrypy.log('found conditions: %s'%condition_scores.keys())
score_mat = np.array(score_mat)
session_d['found'] = gene_names
session_d['missing'] = [g for g in gene_list if g not in gene_names]
# compare motif scores of requested genes to all hypersensitive regions
# in the dataset
cherrypy.log('loading motif background')
# pick n random background sequences, but seed so the same indices are picked for every
# different # of input DHS regions
random.seed("jo mama")
# walk through the conditions and compute scores
sig_scores = defaultdict(dict)
hs_regions = {}
enriched_motifs = {}
for c, scores in condition_scores.items() :
score_mat = np.array(scores)
hs_regions[c] = len(scores)
# get the background out for this condition
motif_bg_fn = resource_filename('adipo_sight','data/%s_hypersensitive_peaks_bg_motif_scores.npy'%c)
all_scores = np.load(motif_bg_fn).T
bg_inds = random.sample(xrange(all_scores.shape[0]),min(score_mat.shape[0],all_scores.shape[0]))
all_scores = all_scores[bg_inds,:]
cherrypy.log('score_mat.shape = %s'%str(score_mat.shape))
cherrypy.log('all_scores.shape = %s'%str(all_scores.shape))
cherrypy.log('done loading motif background')
# calculate MWW
pvals = mww_multiprocess(score_mat.T,all_scores.T,True)
log.debug('motif scores for condition: %s'%c)
log.debug('pvals: %s'%str(pvals.shape))
motif_name_fn = resource_filename('adipo_sight','data/motif_names.txt')
motif_names = np.array(open(motif_name_fn).readlines())
motif_cluster_fn = resource_filename('adipo_sight','data/motif_clusters.txt')
motif_cluster_map = dict((i,int(m)) for i,m in enumerate(open(motif_cluster_fn)))
thresh = pvals < self.args.get('diff_hyp_pval')
thresh_inds = np.where(thresh)[0]
thresh_names, thresh_pvals = motif_names[thresh], pvals[thresh]
thresh_imgs = np.array(['images/motif_logos/%03d_motif.png'%i for i in thresh_inds])
cluster_set = set()
for i,n,p in zip(thresh_inds,thresh_names,thresh_pvals) :
cluster_i = motif_cluster_map[i]
sig_scores[cluster_i].setdefault('name',set()).add(n.strip())
sig_scores[cluster_i][c] = min(sig_scores[cluster_i].get(c,1.),p)
cluster_set.add(cluster_i)
enriched_motifs[c] = len(cluster_set)
d['motifs'] = dict(sig_scores)
d['hs_regions'] = hs_regions
d['enriched_motifs'] = enriched_motifs
session_d['motif_enrich'] = d
