def correlation(label, genome, method1, method2, n=1):
"""Correlate binders from 2 methods"""
n=int(n)
path1 = os.path.join(datapath, '%s/%s/%s' %(label,genome,method1))
path2 = os.path.join(datapath, '%s/%s/%s' %(label,genome,method2))
print path1, path2
if not os.path.exists(path1) or not os.path.exists(path2):
return 1
b1 = getBinders(path1,method1,n=n)
b2 = getBinders(path2,method2,n=n)
#m = pd.merge(b1, b2, on=['peptide','name','pos'])#, suffixes=['1','2']
P1 = base.getPredictor(method1)
P2 = base.getPredictor(method2)
g1 = b1.groupby('name').agg({P1.scorekey:[np.mean,np.size,np.max]}).sort()
g1.columns = g1.columns.get_level_values(1)
g2 = b2.groupby('name').agg({P2.scorekey:[np.mean,np.size,np.max]}).sort()
g2.columns = g2.columns.get_level_values(1)
res = pd.merge(g1, g2, left_index=True,right_index=True)
gfile = getGenome(genome)
gn = sequtils.genbank2Dataframe(gfile, cds=True)
res = res.merge(gn[['locus_tag','length']],
left_index=True,right_on='locus_tag')
res['perc_x'] = res['size_x']/res.length*100
res['perc_y'] = res['size_y']/res.length*100
#print res[:20]
return res
def getFeature(g,tag):
"""Get gene feature from stored genbank file"""
if g == 'other':
return None,None,None,None
from Bio.Seq import Seq
from Bio.SeqRecord import SeqRecord
fname = getGenome(g)
df = sequtils.genbank2Dataframe(fname, cds=True)
df = df.drop_duplicates('locus_tag')
df = df.set_index('locus_tag')
keys = df.index
if tag not in keys:
return
row = df.ix[tag]
seq = row.translation
prod = row['product']
rec = SeqRecord(Seq(seq),id=tag,description=prod)
fastafmt = rec.format("fasta")
#print fastafmt
feature = row.to_dict()
ind = keys.get_loc(tag)
previous = keys[ind-1]
if ind<len(keys)-1:
next = keys[ind+1]
else:
next=None
return feature, fastafmt, previous, next
def genomeAnalysis(label, gname, method, n=3, cutoff=0.96):
"""Analyse over genome for cluster densities and other metrics -
some of this should be moved into mhcpredict library"""
path = os.path.join(datapath, '%s/%s/%s' %(label,gname,method))
if not os.path.exists(path):
return dict(res=None)
b = getBinders(path,method,n,cutoff)
P = base.getPredictor(method)
if P.rankascending==1:
func = np.min
else:
func = np.max
res = b.groupby('name').agg({P.scorekey:[np.size,func]}).sort()
res.columns = res.columns.get_level_values(1)
res.rename(columns={'size': 'binders'}, inplace=True)
#get genome and combine data
gfile = getGenome(gname)
g = sequtils.genbank2Dataframe(gfile, cds=True)
res = res.merge(g[['locus_tag','length','gene','product','order']],
left_index=True,right_on='locus_tag')
res['perc_binders'] = res['binders']/res.length*100
res = res.sort('perc_binders',ascending=False)
cutoff = round(cutoff,2)
clusterfile = os.path.join(path,'clusters_%s_%s.csv' %(n,cutoff))
if os.path.exists(clusterfile):
cl = pd.read_csv(clusterfile)
else:
cl = analysis.findClusters(b, method, dist=None, genome=g)
if len(cl) > 0:
cl.set_index('name').to_csv(clusterfile)
if len(cl) > 0:
gc = cl.groupby('name').agg({'density':np.max,'name':np.size})
gc.rename(columns={'name': 'clusters'}, inplace=True)
res = res.merge(gc,left_on='locus_tag',right_index=True,how='left')
res = res.fillna('-')
#get binder metrics
top = b.groupby('peptide').agg({P.scorekey:func,'allele':np.max,
'name': lambda x: ';'.join(list(x))})
top = top.sort(P.scorekey,ascending=P.rankascending).reset_index()
print len(top)
#add protein urls to results table
res['locus_tag'] = res['locus_tag'].apply(
lambda x: str(A(x, _href=URL(r=request,f='protein',args=[label,gname,x],
extension=''))))
if len(cl)>0:
cl['name'] = cl['name'].apply(
lambda x: str(A(x, _href=URL(r=request,f='protein',args=[label,gname,x],
extension=''))))
fig=None
return b,res,top,cl,fig
def getTagbyGene(g,gene):
"""Get the locus_tag from the gene name"""
fname = getGenome(g)
df = sequtils.genbank2Dataframe(fname, cds=True)
df = df.drop_duplicates('locus_tag')
df = df.set_index('locus_tag')
#if gene name provided we try to override the locus_tag
t = df[df.gene==gene].index
if len(t)>0:
return t[0]
else:
return
def genomeview():
"""Summary page for genome"""
g = request.args[0]
if len(request.args) == 1:
gfile = getGenome(g)
data = sequtils.genbank2Dataframe(gfile)
summary = sequtils.genbankSummary(data)
data = data[data.type=='CDS']
data = data.drop(['type','pseudo'],1)
#data=data.set_index('locus_tag')
return dict(genome=g,data=data,summary=summary)
else:
return dict()
def doSearch(genome, gene, desc):
"""Search genbank frame by gene or descr"""
gfile = getGenome(genome)
g = sequtils.genbank2Dataframe(gfile, cds=True)
g = g.fillna('')
if gene == '' and desc == '':
df = g
else:
if gene == '': gene = '@#!x'
if desc == '': desc = '@#!x'
df = g[(g.gene.str.contains(gene, case=False)) |
(g['product'].str.contains(desc, case=False))]
df = df.drop(['type','pseudo','note','translation'],1)
df = df.set_index('locus_tag')
return df
def conservationAnalysis(label, genome, method, tag, identity, n=3,
equery=None, **kwargs):
"""Conservation analysis"""
n=int(n)
identity=int(identity)
res = {}
blastpath = os.path.join(request.folder, 'static/data/blast')
cachedfile = os.path.join(blastpath, '%s_%s.csv' %(genome,tag))
#get predictions
preds, bcell, cutoffs = getPredictions(label,genome,tag,q=0.96)
if not preds.has_key(method):
return 1
pred = preds[method]
length = pred.getLength()
pb = pred.getPromiscuousBinders(n=n)
if len(pb)==0:
return 1
if os.path.exists(cachedfile):
print 'using %s' %cachedfile
alnrows = pd.read_csv(cachedfile,index_col=0)
else:
if genome != 'other':
gfile = getGenome(genome)
g = sequtils.genbank2Dataframe(gfile, cds=True)
prot = g[g['locus_tag']==tag]
else:
ffile,desc = getFasta(genome)
g = sequtils.fasta2Dataframe(ffile)
prot = g[g['locus_tag']==tag]
if len(prot)==0:
return dict(res=None)
seq = prot.translation.head(1).squeeze()
print tag,seq
alnrows = analysis.getOrthologs(seq,hitlist_size=400,equery=equery)
if alnrows is None:
alnrows=None
#cache blast results for re-use
alnrows.to_csv(cachedfile)
alnrows.drop_duplicates(subset=['sequence'], inplace=True)
def getConserved(pb,alnrows):
#find conserved binders
f=[]
for i,a in alnrows.iterrows():
seq = a.sequence
found = [seq.find(j) for j in pb.peptide]
f.append(found)
s = pd.DataFrame(f,columns=pb.peptide,index=alnrows.accession)
s = s.replace(-1,np.nan)
res = s.count()
return res
def findConservedwithIdentity(alnrows,pb):
vals=[]
lowest = alnrows.perc_ident.min().round()
for i in np.arange(lowest,100,10):
x = alnrows[alnrows['perc_ident']>=i]
s = getConserved(pb,x)/len(x)
s.name=i
vals.append(s)
df = pd.DataFrame(vals)
return df
stats = findConservedwithIdentity(alnrows,pb)
#limit to current identity level
alnrows = alnrows[alnrows['perc_ident']>=identity]
if len(alnrows) == 0:
return 2
found = getConserved(pb,alnrows)
pb = pb.set_index('peptide')
pb['conserved'] = found
pb['perc_cons'] = pb.conserved/len(alnrows)
pb=pb.sort('conserved',ascending=False).drop(['core','name'],1)
summary = pb
#print pb[:10]
fig=plt.figure(figsize=(5,5))
ax=fig.add_subplot(111)
stats=stats.T.drop_duplicates().T
print stats
stats.plot(ax=ax,grid=False,lw=1.5,colormap='rainbow')
plt.legend(loc=2,prop={'size':9})
ax.set_ylabel('percentage cons.')
ax.set_title('conservation vs identity')
plt.tight_layout()
#fig = plotConserved(stats)
return res, alnrows, summary, fig
def runPredictors(label,genome='',newlabel='',names='',fasta='',methods='tepitope',length=11,
mhc1alleles=[], drballeles=[], dpqalleles=[], preset=[],
bcellmethod='Bepipred', iedbmethod='IEDB_recommended',
user=None, **kwargs):
"""Run predictors and save results"""
limit = 50 #limit alleles
applySettings()
if names != '':
names = names.split(',')
else:
names=None
if newlabel != '':
label = newlabel
if genome != '':
query = db.genomes(db.genomes.name==genome)
f,gfile = db.genomes.file.retrieve(query.file)
df = sequtils.genbank2Dataframe(gfile, cds=True)
elif fasta != '':
#if no genome given use the fasta sequences
query = db.sequences(db.sequences.name==fasta)
f,ffile = db.sequences.file.retrieve(query.file)
df = sequtils.fasta2Dataframe(ffile)
genome = 'other'
else:
return
if type(methods) is not types.ListType:
methods = [methods]
length=int(length)
for method in methods:
P = base.getPredictor(method)
if method in ['iedbmhc1']:
alleles = mhc1alleles
P.iedbmethod = iedbmethod
elif method == 'bcell':
P.iedbmethod = bcellmethod
alleles = None
else:
if type(drballeles) is types.StringType:
drballeles=[drballeles]
if type(dpqalleles) is types.StringType:
dpqalleles=[dpqalleles]
alleles = drballeles + dpqalleles
#if preset is chosen override allele lists with this
if preset != '':
alleles = getPresetAlleles(preset)
print alleles
if len(alleles)>limit:
alleles=alleles[:limit]
savepath = os.path.join(datapath, label, genome, method)
if not os.path.exists(savepath):
os.makedirs(savepath)
else:
removeFiles(savepath)
P.predictProteins(df,length=length,names=names,alleles=alleles,
label=label,save=True,path=savepath)
#also pre-calculate binders for n=3
b = analysis.getAllBinders(savepath,method=method,n=3)
if b is not None:
binderfile = os.path.join(savepath, 'binders_3.csv')
b.to_csv(binderfile)
addPredictionstoDB(label,os.path.join(datapath, label),user)
return dict(proteins=len(df))