from itertools import izip, permutations

import json

import pprint

import random

import re

import math

from scipy.stats.stats import pearsonr, spearmanr

from utils import *

TWIN_AGES = sorted(twin_age.values())

TWINS = sorted(twin_age, key = lambda x: twin_age[x])

TWIN_PAIR_AGES = [twin_age[t1] for t1, t2 in twinpairs]

#getvf = lambda data, s: [data[t][s] for t in TWINS]

getvf = lambda data, s: [round (data[t][s],22) for t in TWINS]

getvd = lambda data, s: [round(abs(data[t1][s] - data[t2][s]), 22) for t1, t2 in twinpairs]

#getvd = lambda data, s: [abs(data[t1][s] - data[t2][s]) for t1, t2 in twinpairs]

def gen_random_ages(ages = None):

return random_ages

def _pearsonr(values, mean_values, ages, mean_ages):

val_diffs = [v - mean_values for v in values]

age_diffs = [age - mean_ages for age in ages]

return float(sum(vald*aged for vald, aged in izip(val_diffs, age_diffs)))/math.sqrt(sum(vald*vald for vald in val_diffs)*sum(aged*aged for aged in age_diffs))

def _mono(values):

return 0

def mono(sites, data, ages, deltas = False):

return res

def cc(sites, data, ages, deltas = False):

return res

def spearmanc(sites, data, ages, deltas = False):

return res

def eucl(vector):

return math.sqrt(sum(v**2 for v in vector))

def _cosined(values, eucl_values, ages, eucl_ages):

return sum(v*a for v, a in izip(values, ages))/(eucl_values * eucl_ages)

def cosined(sites, data, ages, deltas = False):

return res

def mean(array):

return float(sum(array))/len(array)

def _dotprod(values, mean_values, ages, mean_ages):

return sum((val - mean_values) * (age - mean_ages) for val, age in izip(values, ages))

def dotprod(sites, data, ages, deltas = False):

return res

def FDR(res, alpha = 0.05):

return pval

if __name__ == '__main__':

DATA_DIR = os.path.join(os.getcwd(), 'data')

anno = {}

for l in open(os.path.join(ANNO_DIR,'RRBS_mapable_regions.info.annotated')):

_, regId, _, _, regAnno = l.split('\t')

anno[int(regId)] = regAnno.strip()

elapsed('annotation')

sites = None

data = {}

for twin_id in datafiles:

fname = reg_fname(twin_id)

cdata = {}

print fname

current_sites = set()

for l in open(fname):

chrNo, regId, start, end, regSites, methLevel = filter(None,re.split(r'\s+',l))

regId = int(regId)

current_sites.add(regId)

cdata[regId] = float(methLevel)

sites = current_sites if sites is None else sites & current_sites

data[twin_id] = cdata

print len(current_sites), len(sites)

elapsed('reading data')

sites = sorted(sites)

# res = pearsonr_on_individual_fragment_deltas(sites, data)

# method = dotprod_on_individual_fragments

method = cc

deltas = True

alpha = 0.15

# method = pearsonr_on_individual_fragments

# method = pearsonr_on_individual_fragments_minus_means

res = method(sites, data, TWIN_PAIR_AGES if deltas else TWIN_AGES, deltas = deltas)

# fdr = FDR(res, alpha=0.37)

fdr = FDR(res, alpha = alpha)

fdr = 1

print fdr

elapsed('calculating scores')

print 'Accepted:', len([s for s in sorted(res, key = lambda k: res[k]['score'], reverse = True) if res[s].get('pval', -1) <= fdr])

out = open(outd('%s_%s_%.2lf.out' % ('deltas' if deltas else 'fragments', method.func_name, alpha)), 'w')

out.write("#fragment\tscore\tp-value\trscore\tdata\tannotation\n")

for s in sorted(res, key = lambda k: res[k]['score'], reverse = True):

if res[s].get('pval', -1) <= fdr:

out.write("%d\t%f\t%f\t%s\t%s\t%s\t%s\n" % (s, res[s]['score'], res[s].get('pval', -1), res[s]['rscore'],res[s].get('rpvals',''), res[s]['data'], anno[s]))

out.close()

elapsed('real data')

# rtimes = 1

# raccepted = 0

# rage = list(reversed(TWIN_PAIR_AGES if deltas else TWIN_AGES))

#

# for rs in xrange(rtimes):

## random.shuffle(rage)

# res = method(sites, data, rage, deltas = deltas)

# fdr = 1

#

#

# out = open(os.path.join(DATA_DIR, '%s_%s_%.2lf.random_out' % ('deltas' if deltas else 'fragments', method.func_name, alpha)), 'w')

# out.write("#fragment\tscore\tp-value\tdata\tannotation\n")

#

# for s in sorted(res, key = lambda k: res[k]['score'], reverse = True):

# if res[s].get('pval', -1) <= fdr:

# out.write("%d\t%f\t%f\t%s\t%s\n" % (s, res[s]['score'], res[s].get('pval', -1), res[s]['data'], anno[s]))

# out.close()

#

# fdr = FDR(res, alpha = alpha)

# print pprint.pformat([s for s in sorted(res, key = lambda k: res[k]['score'], reverse = True) if res[s].get('pval', -1) <= fdr][:10])

# print 'random fdr:', fdr

# _raccepted = len([s for s in sorted(res, key = lambda k: res[k]['score'], reverse = True) if res[s].get('pval', -1) <= fdr])

# print '_raccepted:', _raccepted, '\n\n'

# raccepted += _raccepted

# print "Random: ", float(raccepted)/rtimes

elapsed('%s_%s.out' % (method.func_name, 'deltas' if deltas else 'fragments'))