def main(): """ confusion_matrix.py: ########################################################### true_pred_rel, false_pred_rel = identify_true_false_positives( TG_TF_pred, TG_TF_ref ) ########################################################### unpredicted_rel = count_false_negatives( TG_TF_ref, true_pred_rel, tt_genes ) ########################################################### TP, FP, FN, TN, recall, specif, precis = calculate_confusion( total_rel, true_pred_rel, false_pred_rel, unpredicted_rel ) ########################################################### print_results(dataset, cutoff, TP, FP, FN, TN, recall, specif, precis) ########################################################### """ #Get test data from AtRegNet.txt AtReg_data = read_data(fa.filename_atreg) AtRegNet_parse = parse_AtReg_data(AtReg_data) TF_TG_ref = TFloc_pairs_AtRegNet(AtRegNet_parse, ['all']) TG_TF_ref = [[info[1], info[0]] for info in TF_TG_ref] TG_list_ref = [info[0] for info in TG_TF_ref] TG_set_ref = set(TG_list_ref) sh_TG_list_ref = list(TG_set_ref) TF_list_ref = [info[1] for info in TG_TF_ref] TF_set_ref = set(TF_list_ref) sh_TF_list_ref = list(TF_set_ref) ########################################################### ########################################################### artificial_data = copy.deepcopy(TG_TF_ref) dataset = "artificial dataset" cutoff = 0 ########################################################### ########################################################### if true_P: tt_genes = [info[0] for info in artificial_data] total_rel = artificial_data ########################################################### true_pred_rel, false_pred_rel = identify_true_false_positives( artificial_data, TG_TF_ref ) ########################################################### unpredicted_rel = count_false_negatives( TG_TF_ref, true_pred_rel, tt_genes ) ########################################################### TP, FP, FN, TN, recall, specif, precis = calculate_confusion( total_rel, true_pred_rel, false_pred_rel, unpredicted_rel ) ########################################################### print_results(dataset, cutoff, TP, FP, FN, TN, recall, specif, precis) ########################################################### ########################################################### percentages = [20, 40, 50, 60] if not false_N and false_P: #Add some noise in the form of False Positives for perc in percentages: cutoff = perc noised_artificial_data = add_false_positives( artificial_data, sh_TG_list_ref, sh_TF_list_ref, perc ) ########################################################### tt_genes = [info[0] for info in artificial_data] total_rel = noised_artificial_data ########################################################### true_pred_rel, false_pred_rel = identify_true_false_positives( noised_artificial_data, TG_TF_ref ) ########################################################### unpredicted_rel = count_false_negatives( TG_TF_ref, true_pred_rel, tt_genes ) ########################################################### TP, FP, FN, TN, recall, specif, precis = calculate_confusion( total_rel, true_pred_rel, false_pred_rel, unpredicted_rel ) ########################################################### print_results(dataset, cutoff, TP, FP, FN, TN, recall, specif, precis) ########################################################### ########################################################### if false_N and not false_P: #Add some noise in the form of False Negatives for perc in percentages: cutoff = perc noised_artificial_data = add_false_negatives(artificial_data, TG_TF_ref, perc) ########################################################### tt_genes = [info[0] for info in artificial_data] total_rel = artificial_data ########################################################### true_pred_rel, false_pred_rel = identify_true_false_positives( noised_artificial_data, TG_TF_ref ) ########################################################### unpredicted_rel = count_false_negatives( TG_TF_ref, true_pred_rel, tt_genes ) ########################################################### TP, FP, FN, TN, recall, specif, precis = calculate_confusion( total_rel, true_pred_rel, false_pred_rel, unpredicted_rel ) ########################################################### print_results(dataset, cutoff, TP, FP, FN, TN, recall, specif, precis) ########################################################### ########################################################### if false_P and false_N: #Add some noise in the form of False Positives #And then add some False_Negatives for perc in percentages: cutoff = perc print "ad:",len(artificial_data) noised_artificial_data = add_false_positives( artificial_data, sh_TG_list_ref, sh_TF_list_ref, perc ) print "nad:",len(noised_artificial_data) more_noised_artificial_data = add_false_negatives( noised_artificial_data, TG_TF_ref, perc ) print "mnad:",len(more_noised_artificial_data) ########################################################### tt_genes = [info[0] for info in noised_artificial_data] total_rel = noised_artificial_data ########################################################### true_pred_rel, false_pred_rel = identify_true_false_positives( more_noised_artificial_data, TG_TF_ref ) ########################################################### unpredicted_rel = count_false_negatives( TG_TF_ref, true_pred_rel, tt_genes ) ########################################################### TP, FP, FN, TN, recall, specif, precis = calculate_confusion( total_rel, true_pred_rel, false_pred_rel, unpredicted_rel ) ########################################################### print_results(dataset, cutoff, TP, FP, FN, TN, recall, specif, precis)
def main():
"""
"""
# Get test data from AtRegNet.txt
AtReg_data = read_data(fa.filename_atreg)
AtRegNet_parse = parse_AtReg_data(AtReg_data)
TF_TG_ref = TFloc_pairs_AtRegNet(AtRegNet_parse, ["all"])
TG_TF_ref = [[info[1], info[0]] for info in TF_TG_ref]
TG_list_ref = [info[0] for info in TG_TF_ref]
TG_set_ref = set(TG_list_ref)
TF_list_ref = [info[1] for info in TG_TF_ref]
TF_set_ref = set(TF_list_ref)
# -----------------------------------------------------------
# exp_list = ['Ligterink_2014','Ligterink_2014_gxe','Keurentjes_2007','Snoek_2012']
exp_list = ["Ligterink_2014"]
cutoff_list = [3] # , 4.3, 6.7]
chromo = [1, 2, 3, 4, 5]
# -----------------------------------------------------------
for dataset in exp_list:
for cutoff in cutoff_list:
print "Analysing %s %s" % (dataset, cutoff)
###########################################################
# Extract the true TG-TF relations and the total possible
# relations from the stored datafiles
filelocation = "%s/%s/genelist_%s/genelist_%s_co%s.txt" % (
fa.mr_folder,
fa.gfolder,
dataset,
dataset,
cutoff,
)
true_rel, total_rel = get_TGTF_from_genelist(filelocation, TG_TF_ref, TG_list_ref, TF_list_ref)
###########################################################
# The TG in the true TG-TF relations are the true_traits (tt)
# in this case named tt_genes
tt_genes = list(set([info[0] for info in true_rel]))
########################################################################
# Get for each true_trait the number of eQTLs
enriched_fn = "%s/%s/enriched_%s/enriched_%s_co%s.txt" % (
fa.mr_folder,
fa.enriched_folder,
dataset,
dataset,
cutoff,
)
trait_eqtl_genelist = get_info(enriched_fn)
# Select true traits based on number of eQTLs
tt_trait_eqtl_genelist = [[t[0], t[1], t[2]] for t in trait_eqtl_genelist if t[0] in tt_genes and t[1] > 0]
trait_with_eqtl = [info[0] for info in tt_trait_eqtl_genelist]
########################################################################
###########################################################
TG_TF_pred = process_enrichment(tt_trait_eqtl_genelist, TF_set_ref)
###########################################################
true_pred_rel, false_pred_rel = identify_true_false_positives(TG_TF_pred, TG_TF_ref)
###########################################################
unpredicted_rel = count_false_negatives(TG_TF_ref, true_pred_rel, trait_with_eqtl)
###########################################################
TP, FP, FN, TN, recall, specif, precis = calculate_confusion(
total_rel, true_pred_rel, false_pred_rel, unpredicted_rel
)
###########################################################
print "true_traits: %s" % len(set(tt_genes))
print_results(dataset, cutoff, TP, FP, FN, TN, recall, specif, precis)
def main():
"""
"""
tic = time.clock()
#Get test data from AtRegNet.txt
AtReg_data = read_data(fa.filename_atreg)
AtRegNet_parse = parse_AtReg_data(AtReg_data)
TF_TG_ref = TFloc_pairs_AtRegNet(AtRegNet_parse, ['all'])
TG_TF_ref = [[info[1], info[0]] for info in TF_TG_ref]
TG_list_ref = [info[0] for info in TG_TF_ref]
TG_set_ref = set(TG_list_ref)
sh_TG_list_ref = list(TG_set_ref)
TF_list_ref = [info[1] for info in TG_TF_ref]
TF_set_ref = set(TF_list_ref)
sh_TF_list_ref = list(TF_set_ref)
#-----------------------------------------------------------
exp_list = ['Snoek_2012']
#exp_list = ['Ligterink_2014','Ligterink_2014_gxe','Snoek_2012','Keurentjes_2007']
cutoff_list = [3]#[4.3,6.7]
chromo = [1,2,3,4,5]
#-----------------------------------------------------------
eQTL_threshold_list = [0,1,2,3]
#-----------------------------------------------------------
#get the premade 1000 distinct seeds of 8 digits each
seedfile = "%s/%s/random_seeds.txt"%(fa.mr_folder, fa.numfolder)
#print "Retrieving randomized seeds from %s"%seedfile
seeds = read_seeds(seedfile)
data_dict = {}
write_summary = False
write_conf = False
print_conf = True
summary = []
for dataset in exp_list:
for cutoff in cutoff_list:
for eQTL_threshold in eQTL_threshold_list:
print "Initializing analysis for dataset %s with cutoff %s"%(dataset, cutoff)
F1 = None
ref_F1 = None
############################################################
####Retrieve original confusion matrix results
subfolder_F1 = "/eqtl_%s/valnum_%s"%(eQTL_threshold, dataset)
F1_fn = "%s/%s/%s/valnum_results_%s_co%s"%(
fa.mr_folder, fa.numfolder,
subfolder_F1, dataset, cutoff
)
try:
ref_recall, ref_precision, ref_F1 = read_predicted_confusion_data(F1_fn)
except:
ref_recall= ref_precision= ref_F1 = None
if ref_F1 != None:
############################################################
####Retrieve genelist
subfolder_genelist = "genelist_%s"%dataset
genelist_fn = "%s/%s/%s/genelist_%s_co%s.txt"%(
fa.mr_folder, fa.gfolder,
subfolder_genelist, dataset,
cutoff
)
trait_genelist_list = get_genelist(genelist_fn)
true_rel, total_rel = get_TGTF_from_genelist(
genelist_fn, TG_TF_ref,
TG_list_ref, TF_list_ref
)
tt_genes = list(set([info[0] for info in true_rel]))
############################################################
####Retrieve enriched list
subfolder_enriched = "enriched_%s"%dataset
enriched_fn = "%s/%s/%s/enriched_%s_co%s.txt"%(
fa.mr_folder, fa.enriched_folder,
subfolder_enriched, dataset,
cutoff
)
trait_eqtl_genelist, dict_trait_enriched = get_enriched(enriched_fn)
############################################################
####Retrieve True Traits
emr_traits_fn = "%s/%s/emr_traitlist_%s_co%s.txt"%(
fa.mr_folder, fa.trait_folder,
dataset, cutoff
)
############################################################
#get all traits that have more than X eQTLs, where X = eQTL_threshold
truetrait_eqtl_list = [[t[0], t[1]] for t in trait_eqtl_genelist if t[0] in tt_genes and t[1]>eQTL_threshold]
trait_with_eqtl = [info[0] for info in truetrait_eqtl_list]
print "traits with eQTL", len(trait_with_eqtl)
############################################################
higher_recall=lower_recall=higher_precision=lower_precision=0
higher_F1=lower_F1=0
permutated_confusion = []
#permutate!
#print "Commencing permutation of %s, standby..."%len(seeds)
#i = 0
for seedling in seeds:
#reset variables
TP=FP=FN=TN=recall=specif=precision=F1= 0
#print i
#i += 1
trait_randomsample = []
#create [trait - sample gene list]
for tr_ge in trait_genelist_list:
g_trait, g_genelist = tr_ge
#print g_trait
#print len(g_genelist)
if g_trait in trait_with_eqtl and g_trait in dict_trait_enriched:
sample_size = len(dict_trait_enriched[g_trait])
rsamp = select_random_sample(g_genelist, sample_size, seedling)
trait_randomsample.append([g_trait,0, rsamp])
#q = len(g_genelist)
#print "take %s from %s"%(sample_size, q)
#TG_TF_pred is summed over all traits in a (dataset, cutoff) combination
#TG_TF_pred = get_randomized_predictions(trait_randomsample, TF_set_ref)
TG_TF_pred = process_enrichment(trait_randomsample, TF_set_ref)
#proceed with the random sample to the confusion matrix
###########################################################
true_pred_rel, false_pred_rel = identify_true_false_positives(
TG_TF_pred,
TG_TF_ref
)
###########################################################
unpredicted_rel = count_false_negatives(
TG_TF_ref, true_pred_rel,
tt_genes
)
###########################################################
TP, FP, FN, TN, recall, specif, precision, F1 = calculate_confusion(
total_rel, true_pred_rel,
false_pred_rel, unpredicted_rel
)
permutated_confusion.append([TP, FP, FN, TN, recall, specif, precision, F1])
###########################################################
#print "true_traits: %s"%len(set(tt_genes))
#print_results(dataset, cutoff, TP, FP, FN, TN, recall, specif, precision)
###########################################################
if recall != None:
if recall < ref_recall:
lower_recall += 1
if recall >= ref_recall:
higher_recall += 1
#else:
#print "recall is None"
#pass
if precision != None:
if precision < ref_precision:
lower_precision += 1
if precision >= ref_precision:
higher_precision += 1
#else:
#print "precision is None"
#pass
if ref_F1 != None and F1 != None:
if F1 < ref_F1:
lower_F1 += 1
if F1 >= ref_F1:
higher_F1 += 1
summary.append([dataset, cutoff, lower_recall, higher_recall, lower_precision, higher_precision, lower_F1, higher_F1])
###########################################################
if write_conf:
substorage = "%s/%s/%s"%(fa.mr_folder, fa.numfolder, dataset)
if not os.path.exists(substorage):
os.mkdir(substorage)
resultsfolder_conf = "%s/permutate_eqtl_%s_%s_co%s.txt"%(
substorage, eQTL_threshold,
dataset, cutoff
)
try:
print "Writing to file %s"%resultsfolder_conf
with open(resultsfolder_conf, 'w') as fo:
fo.write("-------------------------")
fo.write("\n")
fo.write("dataset: %s"%dataset)
fo.write("\n")
fo.write("cutoff: %s"%cutoff)
fo.write("\n")
fo.write("-------------------------")
fo.write("\n")
fo.write("lower_F1: \t%s"%lower_F1)
fo.write("\n")
fo.write("higher_F1: \t%s"%higher_F1)
fo.write("\n")
fo.write("-------------------------")
fo.write("\n")
fo.write("lower_recall: %s"%lower_recall)
fo.write("\n")
fo.write("higher_recall: %s"%higher_recall)
fo.write("\n")
fo.write("lower_precision: %s"%lower_precision)
fo.write("\n")
fo.write("higher_precision: %s"%higher_precision)
fo.write("\n")
fo.write("-------------------------")
fo.write("\n")
for [TP, FP, FN, TN, recall, specif, precision, F1] in permutated_confusion:
fo.write("-------------------------\n")
fo.write("TP\t%s\tFN\t%s"%(TP, FN))
fo.write("\n")
fo.write("FP\t%s\tTN\t%s"%(FP, TN))
fo.write("\n")
fo.write("-------------------------\n")
fo.write("recall\t%s"%recall)
fo.write("\n")
fo.write("specificity\t%s"%specif)
fo.write("\n")
fo.write("precision\t%s"%precision)
fo.write("\n")
fo.write("F1\t%s"%F1)
fo.write("\n")
fo.write("-------------------------\n")
except:
pass
if print_conf:
try:
print "-------------------------"
print "TP\t%s\tFN\t%s"%(TP, FN)
print "FP\t%s\tTN\t%s"%(FP, TN)
print "-------------------------"
print "dataset: %s"%dataset
print "cutoff: %s"%cutoff
print "eQTL: %s"%eQTL_threshold
print "-------------------------"
print "lower_F1:\t%s"%lower_F1
print "higher_F1:\t%s"%higher_F1
print "-------------------------"
print "lower_recall: %s"%lower_recall
print "higher_recall: %s"%higher_recall
print "lower_precision: %s"%lower_precision
print "higher_precision: %s"%higher_precision
print "-------------------------"
except:
pass
if write_summary:
summfolder_conf = "%s/%s/permutate_summary_eqtl_%s.txt"%(
fa.mr_folder, fa.numfolder,
eQTL_threshold
)
try:
with open(summfolder_conf, 'w') as fo:
for dataset, cutoff, lower_recall, higher_recall, lower_precision, higher_precision, lower_F1, higher_F1 in summary:
fo.write("-------------------------")
fo.write("\n")
fo.write("dataset: %s"%dataset)
fo.write("\n")
fo.write("cutoff: %s"%cutoff)
fo.write("\n")
fo.write("-------------------------")
fo.write("\n")
fo.write("lower_F1: \t%s"%lower_F1)
fo.write("\n")
fo.write("higher_F1: \t%s"%higher_F1)
fo.write("\n")
fo.write("-------------------------")
fo.write("\n")
fo.write("recall:")
fo.write("\n")
fo.write("lower: %s"%lower_recall)
fo.write("\n")
fo.write("higher: %s"%higher_recall)
fo.write("\n")
fo.write("precision:")
fo.write("\n")
fo.write("lower: %s"%lower_precision)
fo.write("\n")
fo.write("higher: %s"%higher_precision)
fo.write("\n")
fo.write("-------------------------")
fo.write("\n")
except:
pass
)
###########################################################
tt_genes = list(set([info[0] for info in true_relations]))
###########################################################
TG_TF_pred = process_enrichment(
dataset, cutoff, chromosome,
TF_set_ref, tt_genes
)
###########################################################
true_pred_rel, false_pred_rel = identify_true_false_positives(
TG_TF_pred,
TG_TF_ref
)
###########################################################
unpredicted_rel = count_false_negatives(
TG_TF_ref, true_pred_rel,
tt_genes
)
#TFloc_list contains predicted datapoints
#TFloc_list = [[info[1],info[0]] for info in total_rel_major]
regus_list = sorted(list(set([info[1] for info in total_rel_major])))
locus_list = sorted(list(set([info[0] for info in total_rel_major])))
regus_dict, TF_chr_len_list = create_reg_dict(regus_list)
locus_dict, chr_len_list = get_loci_from_dataset(locus_list)
#Prepare data for scatter plots
moddata_array, labels_mod = link_data_in_array(total_rel_major, locus_dict, regus_dict, {})