Created on Oct 30, 2019
@author: Vincentius Martin, Farica Zhuang
Make some plots for analysis
'''
import pandas as pd
import os
os.chdir("../../..")
import chip2probe.modeler.plotlib as pl
from chip2probe.modeler.cooptrain import CoopTrain
if __name__ == "__main__":
trainingpath = "output/homotypic/training/training_pwm.csv"
df = pd.read_csv(trainingpath) # , sep="\t")
df['label'] = df['label'].replace({"additive": "independent"})
train = CoopTrain(df, corelen=4)
# make distace stacked bar
pl.plot_stacked_categories(train.df,
"orientation",
path="dist_stackedbar.png")
# get stacked bar of ratio between different distance
pl.plot_stacked_categories(train.df,
"distance",
path="distance_ets_ets.png",
ratio=True)
#dft["orientation"] = dft.apply(lambda x: "%s%s" % (str(x["ets_ori"]), str(x["runx_ori"])),axis=1)
dft["orientation"] = dft.apply(
lambda x: "%s/%s" %
(orimap[int(x["ets_ori"])], orimap[int(x["runx_ori"])]),
axis=1)
#print(dft)
dft.to_csv("training_pwm.tsv", sep="\t", index=False, float_format='%.3f')
dft.rename(columns={
'runx_score': 'Runx1 strength\n(cooperator TF)',
'ets_score': 'Ets1 strength\n(main TF)'
},
inplace=True)
#plot.plot_stacked_categories(dft, "distance", path="distance_bar.png", title="Distance distribution", ratio=True)
plot.plot_stacked_categories(
dft,
"orientation",
path="ori_bar.png",
title="Relative sites orientation distribution",
ratio=True)
plot.plot_box_categories(
dft,
incols=["Runx1 strength\n(cooperator TF)", "Ets1 strength\n(main TF)"],
alternative="smaller")
# dft["runx_score"] = predict_strength(dft, pwm_runx, "runx", flanklen)
# dft["ets_score"] = predict_strength(dft, pwm_ets, "ets", flanklen)
# dft = dft.drop(["fullseq"], axis=1)
# dft = dft[dft['ets_score'].notnull() & dft['runx_score'].notnull()]
# dft.to_csv("training.tsv",sep="\t",index=False)
# plot.plot_stacked_categories(dft, "distance")
# plot.plot_box_categories(dft, incols=["distance","runx_score", "ets_score"], alternative="smaller")
train = gen_training(df, pwm_ets, kompas_ets)
print(train["label"].value_counts())
train.to_csv("output/Ets1Ets1/training/train_ets1_ets1.tsv",
index=False,
sep="\t")
train.rename(columns={
'site_str_score': 'Binding strength of the stronger site',
'site_wk_score': 'Binding strength of the weaker site'
},
inplace=True)
pl.plot_stacked_categories(
train,
"distance",
path="output/Ets1Ets1/training/distance_bar.png",
title="Distance distribution",
ratio=True,
figsize=(17, 4))
pl.plot_stacked_categories(
train,
"orientation",
path="output/Ets1Ets1/training/ori_bar.png",
title="Relative sites orientation\ndistribution",
ratio=True,
figsize=(9, 5))
pl.plot_box_categories(train,
path="output/Ets1Ets1/training/boxplot.png",
incols=[
"Binding strength of the stronger site",
"Binding strength of the weaker site"
curdf_coop.shape[0], coopfrac))
print("Confusion matrix:")
print(cm)
minp = curdf.apply(lambda x: x["p_o1"]
if x["p_o1"] < x["p_o2"] else x["p_o2"],
axis=1)
corr, _ = pearsonr(minp, curdf["main_prob"])
print("R(model_probability,p_value): %.2f" % corr, "\n")
fprtype, tprtype, _ = metrics.roc_curve(np.array(curdf["label"]),
np.array(curdf["main_prob"]))
auctype = metrics.auc(fprtype, tprtype)
plt.plot(fprtype, tprtype, label='AUC %s = %.2f' % (mtype, auctype))
if mtype == "distance":
g = curdf.groupby(["distance", "label"])["label"].count()
pl.plot_stacked_categories(curdf,
"distance",
path="cust_stackedbar_dist.png",
ratio=True)
elif mtype == "affinity":
g = curdf.groupby("seqid")
ghead = g.head(1)
ghead = ghead[ghead["label"] == 0][["seqid", "id"]]
gtail = g.tail(1)
gtail = gtail[gtail["label"] == 1][["seqid", "id"]]
ght = ghead.merge(gtail, on="seqid")
print(ght)
# if needed, print
elif mtype == "orientation":
for ori in ["HH", "HT/TH", "TT"]:
oridf = curdf[(curdf["orientation"] == ori)]
orimatch = oridf[oridf["label"] == oridf["main_pred"]].shape[0]
print("Orientation %s, match: (%d/%d)" %
train.to_csv("%s/training/train_%s_%s.tsv" % (basepath, maintf, cooptf),
sep="\t",
index=False,
float_format='%.3f')
train.rename(columns={
'%s_score' % maintf:
'%s strength\n(main TF)' % maintf.capitalize(),
'%s_score' % cooptf:
'%s strength\n(cooperator TF)' % cooptf.capitalize()
},
inplace=True)
plot.plot_stacked_categories(train,
"distance",
path="%s/training/distance_bar.png" %
basepath,
title="Distance distribution",
ratio=True,
figsize=(17, 4),
color=color)
plot.plot_stacked_categories(
train,
"orientation",
path="%s/training/ori_bar.png" % basepath,
title="Relative sites orientation\ndistribution",
ratio=True,
figsize=(9, 5),
color=color)
plot.plot_box_categories(
train,
path="%s/training/boxplot.png" % basepath,
incols=[