def matrix(data,freq):
# get total and ICmax
total = len(data)
IC_max = hpo_helper.IC(1,total)
result = defaultdict(float)
method_cache = {}
n = total
for d in data:
print(n)
n -= 1
hpos = d['hpo']
# apparently each hpo co-occur with itself
for h in hpos:
key = '-'.join([h,h])
method_cache[h] = method_cache.get(h,hpo_helper.IC(freq[h],total)/(IC_max))
result[key] = result.get(key, method_cache[h])
# check each combination of the hpos
for h in itertools.combinations(hpos,2):
key = '-'.join(sorted([h[0],h[1]]))
# using normaliser conveniently make the result consistently smaller than if h0 is a subclass of h1 if later using IC(h0)*IC(h1)/max_IC**2 for getting weights.
normaliser = 2*min(freq[h[0]],hpo_freq[h[1]])
result[key] += (method_cache[h[0]] + method_cache[h[1]]) / normaliser
return result
def asym_WAM(dbs, df, ic_df, freq, ancestors):
print('build hpo_asym_WAM')
weight_df = pd.DataFrame(index=ic_df.index)
max_ic = hpo_helper.IC(1, freq['HP:0000001'])
buff = {}
for i, h in enumerate(ic_df.index):
this = ic_df.reset_index().apply(beta,
axis=1,
dbs=dbs,
ic_df=ic_df,
freq=freq,
max_ic=max_ic,
h2=h,
buff=buff,
mode='asym')
weight_df[h] = this.values
return (df.multiply(weight_df))
def IC_wrapper(a):
return hpo_helper.IC(a, t)
basename = os.path.basename(input)
num = basename.split('.')[0].split('_')[1]
outfile = os.path.join(outfolder, 'hpofreq_' + num + '.json')
return outfile
'''
main
'''
if __name__ == "__main__":
usage = "usage: %prog [options] arg1 arg2"
parser = OptionParser(usage=usage)
parser.add_option("--input", dest="input", help="input sim file?")
(options, args) = parser.parse_args()
# get dbs
dbs = phenopolis_utils.get_mongo_collections()
# get input data
input_data = hpo_helper.get_json(options.input)
# get total and ICmax
total = len(input_data)
IC_max = hpo_helper.IC(1, total)
# expand hpos
hpo_helper.expand_hpo(input_data)
# get hpo_freq
hpo_freq = hpo_helper.get_hpo_freq(input_data)
# get output file
outfile = get_outfile(options.input)
hpo_helper.write_json(hpo_freq, outfile)
print('done')