def make_deseq2_count(unsel,
sel,
data_deseq2_annot,
data_lbl_col,
prj_dh,
type_form='aas'):
"""
Makes DESeq2 count file
:param unsel: unselected condition
:param sel: selected condtition
:param prj_dh: path to the project directory
:param data_lbl_col: column of the data_lbl pandas table to be used
:param data_deseq2_annot: pandas table with annotation information
"""
data_deseq2_count_dh = '%s/data_fit/%s_all' % (prj_dh, type_form)
data_deseq2_count_fh = '%s/%s_WRT_%s.deseq2_count.csv' % (
data_deseq2_count_dh, sel, unsel)
if not exists(data_deseq2_count_fh):
data_lbl_fhs = [
"%s/data_lbl/%s/%s" % (prj_dh, type_form, s.split('.')[0])
for s in data_deseq2_annot.index
]
col_sep = "."
data_lbl_all = fhs2data_combo(data_lbl_fhs,
cols=[data_lbl_col],
index='mutids',
col_sep=col_sep)
data_lbl_all = debad(data_lbl_all, axis=0, condi='any', bad='nan')
data_lbl_all.to_csv(data_deseq2_count_fh)
else:
data_lbl_all = pd.read_csv(data_deseq2_count_fh).set_index('mutids')
return data_lbl_all, data_deseq2_count_fh
def transform_data_lbl(
prj_dh,
transform_type,
type_form='aas',
data_lbl_col='NiA_norm',
):
"""
Transforamtion of counts of mutants in data_lbl table
:param prj_dh: path to the project directory
:param transform_type: type of transformation log, plog, glog etc
:returns data_lbl: data_lbl with transformed counts
"""
data_lbl_fhs = glob("%s/data_lbl/aas/*" % prj_dh)
if len(data_lbl_fhs) > 0:
col_sep = "."
data_lbl_all = fhs2data_combo(data_lbl_fhs,
cols=[data_lbl_col],
index='mutids',
col_sep=col_sep)
data_lbl_all_dh = '%s/data_lbl/%s_all' % (prj_dh, type_form)
if not exists(data_lbl_all_dh):
makedirs(data_lbl_all_dh)
data_lbl_all_fh = '%s/%s.csv' % (data_lbl_all_dh, data_lbl_col)
data_lbl_all.to_csv(data_lbl_all_fh)
if (transform_type == 'log2') or (transform_type == 'log'):
data_lbl_all = data_lbl_all.apply(np.log2)
elif transform_type == 'plog':
data_lbl_all = data_lbl_all.apply(plog)
else:
logging.error("trnaform_type not valid: %s" % transform_type)
sys.exist()
data_lbl_col = 'NiA_tran'
data_lbl_all_fh = '%s/%s.csv' % (data_lbl_all_dh, data_lbl_col)
data_lbl_all.to_csv(data_lbl_all_fh)
for col in data_lbl_all:
data_lbl_fn, tmp = col.split('.')
data_lbl_fh = '%s/data_lbl/%s/%s' % (prj_dh, type_form,
data_lbl_fn)
data_lbl = pd.read_csv(data_lbl_fh).set_index('mutids')
if not data_lbl_col in data_lbl:
data_lbl_cols = data_lbl.columns.tolist()
data_lbl = pd.concat([data_lbl, data_lbl_all.loc[:, col]],
axis=1)
data_lbl.columns = data_lbl_cols + [data_lbl_col]
data_lbl.index.name = 'mutids'
data_lbl.to_csv(data_lbl_fh)
def transform_data_lbl(prj_dh,transform_type,
type_form='aas',data_lbl_col='NiA_norm',):
"""
Transforamtion of counts of mutants in data_lbl table
:param prj_dh: path to the project directory
:param transform_type: type of transformation log, plog, glog etc
:returns data_lbl: data_lbl with transformed counts
"""
data_lbl_fhs=glob("%s/data_lbl/aas/*" % prj_dh)
if len(data_lbl_fhs)>0:
col_sep="."
data_lbl_all=fhs2data_combo(data_lbl_fhs,cols=[data_lbl_col],index='mutids',col_sep=col_sep)
data_lbl_all_dh='%s/data_lbl/%s_all' % (prj_dh,type_form)
if not exists(data_lbl_all_dh):
makedirs(data_lbl_all_dh)
data_lbl_all_fh='%s/%s.csv' % (data_lbl_all_dh,data_lbl_col)
data_lbl_all.to_csv(data_lbl_all_fh)
if (transform_type=='log2') or (transform_type=='log'):
data_lbl_all=data_lbl_all.apply(np.log2)
elif transform_type=='plog':
data_lbl_all=data_lbl_all.apply(plog)
else:
logging.error("trnaform_type not valid: %s" % transform_type)
sys.exist()
data_lbl_col='NiA_tran'
data_lbl_all_fh='%s/%s.csv' % (data_lbl_all_dh,data_lbl_col)
data_lbl_all.to_csv(data_lbl_all_fh)
for col in data_lbl_all:
data_lbl_fn,tmp=col.split('.')
data_lbl_fh='%s/data_lbl/%s/%s' % (prj_dh,type_form,data_lbl_fn)
data_lbl=pd.read_csv(data_lbl_fh).set_index('mutids')
if not data_lbl_col in data_lbl:
data_lbl_cols=data_lbl.columns.tolist()
data_lbl=pd.concat([data_lbl,
data_lbl_all.loc[:,col]],axis=1)
data_lbl.columns=data_lbl_cols+[data_lbl_col]
data_lbl.index.name='mutids'
data_lbl.to_csv(data_lbl_fh)
def make_deseq2_count(unsel,sel,data_deseq2_annot,data_lbl_col,prj_dh,type_form='aas'):
"""
Makes DESeq2 count file
:param unsel: unselected condition
:param sel: selected condtition
:param prj_dh: path to the project directory
:param data_lbl_col: column of the data_lbl pandas table to be used
:param data_deseq2_annot: pandas table with annotation information
"""
data_deseq2_count_dh='%s/data_fit/%s_all' % (prj_dh,type_form)
data_deseq2_count_fh='%s/%s_WRT_%s.deseq2_count.csv' % (data_deseq2_count_dh,sel,unsel)
if not exists(data_deseq2_count_fh):
data_lbl_fhs=["%s/data_lbl/%s/%s" % (prj_dh,type_form,s.split('.')[0]) for s in data_deseq2_annot.index]
col_sep="."
data_lbl_all=fhs2data_combo(data_lbl_fhs,cols=[data_lbl_col],index='mutids',col_sep=col_sep)
data_lbl_all=debad(data_lbl_all,axis=0,condi='any',bad='nan')
data_lbl_all.to_csv(data_deseq2_count_fh)
else:
data_lbl_all=pd.read_csv(data_deseq2_count_fh).set_index('mutids')
return data_lbl_all,data_deseq2_count_fh
def data_lbl2data_fit_lite(fits_pairs,prj_dh,data_lbl_dh,data_fit_dh,force=False):
"""
Short wrapper for conversion of mutation counts to fold changes
:param fits_pair: list with pair of selected and reference condition
:param pj_dh: path to the protject directory
:param data_lbl_dh: path to the diorectory containing data_lbl csv tables
:param data_fit_dh: path to the directory of data_fit csvs
"""
data_fit_fh='%s/%s/aas/%s_WRT_%s' % (prj_dh,data_fit_dh,fits_pairs[1],fits_pairs[0])
if not exists(data_fit_fh) or force:
data_lbl_fhs=['%s/%s/aas/%s' % (prj_dh,data_lbl_dh,s) for s in fits_pairs]
data_fit=fhs2data_combo(data_lbl_fhs,cols=['NiA_tran'],
# labels=['ref','sel'],
index='mutids',col_sep='.')
data_fit.columns=['NiA_tran.ref','NiA_tran.sel']
data_fit.loc[:,'FCA']=data_fit.loc[:,'NiA_tran.sel']-data_fit.loc[:,'NiA_tran.ref']
data_fit.loc[:,'FCA_norm']=data_fit.loc[:,'NiA_tran.sel']-data_fit.loc[:,'NiA_tran.ref']
if not exists(dirname(data_fit_fh)):
makedirs(dirname(data_fit_fh))
data_fit.to_csv(data_fit_fh)
def corrplot(info):
"""
Plots a correlation matrix heatmap between range of features and fold change values
:param info: dict, with the information of the experiment
"""
from dms2dfe.lib.io_dfs import fhs2data_combo
from glob import glob
from dms2dfe.lib.plot_mut_data_heatmaps import clustermap
from dms2dfe.lib.io_ml_data import make_dXy
ml_input = info.ml_input
prj_dh = info.prj_dh
data_fit_fhs = glob('%s/data_fit/aas/*' % prj_dh)
data_feats_all_fh = '%s/data_feats/aas/data_feats_all' % prj_dh
data_feats_all = pd.read_csv(data_feats_all_fh).set_index('mutids')
data_fit_all = fhs2data_combo(data_fit_fhs, ['%sA' % ml_input], 'mutids')
data_fit_all.columns = [c.split(': ')[0] for c in data_fit_all]
for c in data_fit_all:
plot_fh = '%s/plots/aas/%s.corr.pdf' % (prj_dh, c)
if not exists(plot_fh):
if not exists(dirname(plot_fh)):
makedirs(dirname(plot_fh))
dXy = data_feats_all.join(data_fit_all[c])
dXy, Xcols, ycol = make_dXy(dXy,
ycol=c,
if_rescalecols=False,
unique_quantile=0.25)
dXy, Xcols, ycol = feats_sel_corr(dXy, ycol, range_coef=[0.9, 0.8])
g, ax = clustermap(
dXy.corr(method='spearman'),
highlight_col=c,
vlim=[-0.5, 0.5],
figsize=[10, 10],
plot_fh=plot_fh,
)
def data_lbl2data_fit_lite(fits_pairs,
prj_dh,
data_lbl_dh,
data_fit_dh,
force=False):
"""
Short wrapper for conversion of mutation counts to fold changes
:param fits_pair: list with pair of selected and reference condition
:param pj_dh: path to the protject directory
:param data_lbl_dh: path to the diorectory containing data_lbl csv tables
:param data_fit_dh: path to the directory of data_fit csvs
"""
data_fit_fh = '%s/%s/aas/%s_WRT_%s' % (prj_dh, data_fit_dh, fits_pairs[1],
fits_pairs[0])
if not exists(data_fit_fh) or force:
data_lbl_fhs = [
'%s/%s/aas/%s' % (prj_dh, data_lbl_dh, s) for s in fits_pairs
]
data_fit = fhs2data_combo(
data_lbl_fhs,
cols=['NiA_tran'],
# labels=['ref','sel'],
index='mutids',
col_sep='.')
data_fit.columns = ['NiA_tran.ref', 'NiA_tran.sel']
data_fit.loc[:,
'FCA'] = data_fit.loc[:,
'NiA_tran.sel'] - data_fit.loc[:,
'NiA_tran.ref']
data_fit.loc[:,
'FCA_norm'] = data_fit.loc[:,
'NiA_tran.sel'] - data_fit.loc[:,
'NiA_tran.ref']
if not exists(dirname(data_fit_fh)):
makedirs(dirname(data_fit_fh))
data_fit.to_csv(data_fit_fh)
def plot_data_comparison_multiviolin(
prj_dh,
data_fits,
col,
data_fiti_ctrl=0,
aasORcds="aas",
ns=True,
numeric=False,
color_test=(0.6, 0.8, 1), #'mediumpurple'#"lime"
color_ctrl="lightgray",
data_fits_labels=None,
pval=True,
stars=True,
violin_width=0.9,
color_xticks=None,
force=False,
ylims=None,
col_hue='Conditions',
label_test='Test',
label_ctrl='Control',
ylabel=None,
figsize=[4, 3],
plot_fh=None):
"""
Plotting distributions of comparison of fold change data
:param prj_dh: path to project directory
:param data_fits: pandas dataframe with fold change data
:param col: column with the fold change data
"""
data_fit_fhs = [
"%s/data_fit/%s/%s" % (prj_dh, aasORcds, s) for s in data_fits
]
if plot_fh != None:
data_comparison_ctrl_fh = plot_fh + "data_comparison_ctrl.csv"
data_comparison_test_fh = plot_fh + "data_comparison_test.csv"
else:
data_comparison_ctrl_fh = "data_comparison_ctrl.csv"
data_comparison_test_fh = "data_comparison_test.csv"
if (not exists(data_comparison_ctrl_fh)) or\
(not exists(data_comparison_test_fh)) or force:
data_fit_test_fhs = [fh for i, fh in enumerate(data_fit_fhs) if i != 0]
data_fit_ctrl_fhs = [fh for i, fh in enumerate(data_fit_fhs) if i == 0]
data_comparison_test = fhs2data_combo(data_fit_test_fhs, [col],
index='mutids')
data_comparison_ctrl = fhs2data_combo(data_fit_ctrl_fhs, [col],
index='mutids')
data_comparison_test.columns = data_fits[1:]
data_comparison_ctrl.columns = data_fits[1:]
#data_comparison_test.to_csv(data_comparison_test_fh)
#data_comparison_ctrl.to_csv(data_comparison_ctrl_fh)
else:
data_comparison_test = pd.read_csv(data_comparison_test_fh).set_index(
'mutids')
data_comparison_ctrl = pd.read_csv(data_comparison_ctrl_fh).set_index(
'mutids')
data_all = pd.concat([
data_comparison_ctrl.loc[:,
data_comparison_ctrl.columns.tolist()[0]],
data_comparison_test
],
axis=1)
data_all.columns = data_fits
data_comparison_test = data_comparison_test.unstack().reset_index()
data_comparison_ctrl = data_comparison_ctrl.unstack().reset_index()
data_comparison_test.columns = ["condi", "mutids", col]
data_comparison_ctrl.columns = ["condi", "mutids", col]
data_comparison_test.loc[:, col_hue] = label_test
data_comparison_ctrl.loc[:, col_hue] = label_ctrl
data_comparison = data_comparison_test.append(data_comparison_ctrl)
y_max = data_comparison.loc[:, col].max()
y_min = data_comparison.loc[:, col].min()
data_fit_ctrl = data_fits[data_fiti_ctrl]
data_fit_tests = [
s for s in data_fits if data_fits.index(s) != data_fiti_ctrl
]
# fig=
plt.figure(figsize=figsize, dpi=300)
ax = plt.subplot(111)
import seaborn as sns
from scipy.stats import mannwhitneyu, wilcoxon
# plt.style.use('seaborn-whitegrid')
# plt.style.use('seaborn-white')
sns.violinplot(
x="condi",
y=col,
hue=col_hue,
split=True,
data=data_comparison,
# color='m',
palette={
label_test: color_test,
label_ctrl: color_ctrl
},
inner="quartile",
width=violin_width,
cut=0,
# bw=0.1,
scale="width",
ax=ax)
plt.legend(
title=col_hue,
loc='upper center',
bbox_to_anchor=(0.5, 1.1), #, 1., .102),
ncol=2,
borderaxespad=0.1,
frameon=True,
)
ax.set_xlabel("")
if not ylabel is None:
ax.set_ylabel(ylabel)
# ax.grid(b=True)
ax.yaxis.grid(True)
if data_fits_labels != None:
ax.set_xticklabels(data_fits_labels[1:])
if ylabel != None:
ax.set_ylabel(ylabel)
plt.tight_layout()
if ylims is None:
ax.set_ylim([y_min, y_max])
else:
ax.set_ylim([ylims[0], ylims[1]])
# return data_comparison,data_all
if pval:
col_ctrl = data_all.columns.tolist()[0]
col_tests = data_all.columns.tolist()[1:]
for col_testi in range(len(col_tests)):
col_test = col_tests[col_testi]
pval, side = get_wilcoxon(data_all,
col_ctrl,
col_test,
side='one',
denan=False)
side = '' #stitch
# print pval
if stars:
pval = pval2stars(pval, ns=ns, numeric=False)
if ns and (pval > 0.05):
side = ''
pval = ''
result = "%s\n%s" % (side, pval)
else:
result = "%s\n%s" % (side, pval)
# print result
# print ax.get_ylim()[0]+(ax.get_ylim()[1]-ax.get_ylim()[0])*0.05
ax.text(
col_testi,
ax.get_ylim()[0] +
(ax.get_ylim()[1] - ax.get_ylim()[0]) * 0.05,
result,
ha='center',
color='b',
bbox=dict(
facecolor='w',
edgecolor='none',
# boxstyle='round',
alpha=0.6,
))
# data_all.to_csv('test1.csv')
if plot_fh != None:
#plt.tight_layout()
plt.savefig(plot_fh, format='pdf')
plt.clf()
plt.close()
return ax, data_comparison, data_all
def transform_data_lbl_deseq(prj_dh,transform_type,rscript_fh,type_form='aas'):
"""
Transforamtion of counts of mutants in data_lbl table using DESeq2
:param prj_dh: path to the project directory
:param transform_type: type of transformation rlog or VST
:returns data_lbl: data_lbl with transformed counts
"""
data_lbl_fhs=glob("%s/data_lbl/aas/*" % prj_dh)
data_lbl_col='NiA_norm'
col_sep="."
data_lbl_all=fhs2data_combo(data_lbl_fhs,cols=[data_lbl_col],index='mutids',col_sep=col_sep)
data_lbl_all_dh='%s/data_lbl/%s_all' % (prj_dh,type_form)
if not exists(data_lbl_all_dh):
makedirs(data_lbl_all_dh)
data_lbl_all_fh='%s/%s.csv' % (data_lbl_all_dh,data_lbl_col)
# data_lbl_all=debad(data_lbl_all,axis=0,condi='any',bad='nan')
# #psudocount to avoid all zero error
# data_lbl_all=data_lbl_all.fillna(0)
# data_lbl_all=data_lbl_all+0.5
data_lbl_all.to_csv(data_lbl_all_fh)
data_lbl_tran_col='NiA_tran'
data_lbl_all_tran_fh='%s/%s.csv' % (data_lbl_all_dh,data_lbl_tran_col)
if not exists(data_lbl_all_tran_fh):
data_lbl_all_tran=pd.DataFrame(index=data_lbl_all.index)
data_lbl_all_tran.index.name='mutids'
repli=pd.read_csv('%s/cfg/repli' % prj_dh).set_index('varname')
for avg in repli.index:
avg_col="%s%s%s" % (avg,col_sep,data_lbl_col)
if transform_type=='rlog':
data_lbl_tran_fh="%s/%s.deseq2_annot.csv.deseq2_rld.csv" % (data_lbl_all_dh,avg_col)
elif transform_type=='vst':
data_lbl_tran_fh="%s/%s.deseq2_annot.csv.deseq2_vsd.csv" % (data_lbl_all_dh,avg_col)
if not exists(data_lbl_tran_fh):
data_deseq2_annot=pd.DataFrame(columns=['condition','type','number of lanes','total number of reads','exon counts'])
data_deseq2_annot.index.name='file'
data_deseq2_annot_fh='%s/%s.deseq2_annot.csv' % (data_lbl_all_dh,avg_col)
data_deseq2_count=pd.DataFrame()
data_deseq2_count_fh='%s/%s.deseq2_count.csv' % (data_lbl_all_dh,avg_col)
for rep in repli.loc[avg,:]:
if not pd.isnull(rep):
rep_col="%s%s%s" % (rep,col_sep,data_lbl_col)
data_deseq2_annot.loc[rep_col,'condition']=rep
if len(data_deseq2_count)==0:
data_deseq2_count=pd.DataFrame(data_lbl_all.loc[:,rep_col].copy())
else:
data_deseq2_count.loc[:,rep_col]=data_lbl_all.loc[:,rep_col]
# data_deseq2_annot_index_all=["%s%s%s" % (basename(fh),col_sep,data_lbl_col) \
# for fh in data_lbl_fhs if (basename(fh) not in repli.index)]
# data_deseq2_annot_index_no_reps=[i for i in data_deseq2_annot_index_all if (i not in data_deseq2_annot.index)]
# for idx in data_deseq2_annot_index_no_reps:
# data_deseq2_annot.loc[idx,'condition']=idx
data_deseq2_annot.to_csv(data_deseq2_annot_fh)
data_deseq2_count=debad(data_deseq2_count,axis=0,condi='any',bad='nan')
data_deseq2_count.to_csv(data_deseq2_count_fh)
deseq_fh="%s/deseq2.R" % (abspath(dirname(__file__)))
log_fh="%s.log" % data_deseq2_annot_fh
with open(log_fh,'a') as log_f:
deseq2_com='%s %s %s %s 1' % (rscript_fh,deseq_fh,data_deseq2_count_fh,data_deseq2_annot_fh)
# print deseq2_com
subprocess.call(deseq2_com,shell=True,stdout=log_f, stderr=subprocess.STDOUT)
data_lbl_tran=pd.read_csv(data_lbl_tran_fh).set_index('Unnamed: 0')
data_lbl_tran.index.name='mutids'
if len(data_lbl_all_tran)==0:
data_lbl_all_tran=data_lbl_tran.copy()
else:
data_lbl_all_tran=data_lbl_all_tran.join(data_lbl_tran)
# print data_lbl_all_tran.columns.tolist()
# print len(data_lbl_all_tran)
if len(data_lbl_all_tran.columns.tolist())>0:
data_lbl_all_tran.to_csv(data_lbl_all_tran_fh)
else:
logging.error('transform_type can not be %s: no replicates found' % transform_type)
sys.exit()
else:
data_lbl_all_tran=pd.read_csv(data_lbl_all_tran_fh).set_index('mutids')
data_lbl_tran_col='NiA_tran'
for col in data_lbl_all_tran:
data_lbl_fn,tmp=col.split('.')
data_lbl_fh='%s/data_lbl/%s/%s' % (prj_dh,type_form,data_lbl_fn)
data_lbl=pd.read_csv(data_lbl_fh).set_index('mutids')
if not data_lbl_tran_col in data_lbl:
data_lbl_cols=data_lbl.columns.tolist()
data_lbl=data_lbl.join(data_lbl_all_tran.loc[:,col])
data_lbl.columns=data_lbl_cols+[data_lbl_tran_col]
data_lbl.index.name='mutids'
data_lbl.to_csv(data_lbl_fh)
def transform_data_lbl_deseq(prj_dh,
transform_type,
rscript_fh,
type_form='aas'):
"""
Transforamtion of counts of mutants in data_lbl table using DESeq2
:param prj_dh: path to the project directory
:param transform_type: type of transformation rlog or VST
:returns data_lbl: data_lbl with transformed counts
"""
data_lbl_fhs = glob("%s/data_lbl/aas/*" % prj_dh)
data_lbl_col = 'NiA_norm'
col_sep = "."
data_lbl_all = fhs2data_combo(data_lbl_fhs,
cols=[data_lbl_col],
index='mutids',
col_sep=col_sep)
data_lbl_all_dh = '%s/data_lbl/%s_all' % (prj_dh, type_form)
if not exists(data_lbl_all_dh):
makedirs(data_lbl_all_dh)
data_lbl_all_fh = '%s/%s.csv' % (data_lbl_all_dh, data_lbl_col)
# data_lbl_all=debad(data_lbl_all,axis=0,condi='any',bad='nan')
# #psudocount to avoid all zero error
# data_lbl_all=data_lbl_all.fillna(0)
# data_lbl_all=data_lbl_all+0.5
data_lbl_all.to_csv(data_lbl_all_fh)
data_lbl_tran_col = 'NiA_tran'
data_lbl_all_tran_fh = '%s/%s.csv' % (data_lbl_all_dh, data_lbl_tran_col)
if not exists(data_lbl_all_tran_fh):
data_lbl_all_tran = pd.DataFrame(index=data_lbl_all.index)
data_lbl_all_tran.index.name = 'mutids'
repli = pd.read_csv('%s/cfg/repli' % prj_dh).set_index('varname')
for avg in repli.index:
avg_col = "%s%s%s" % (avg, col_sep, data_lbl_col)
if transform_type == 'rlog':
data_lbl_tran_fh = "%s/%s.deseq2_annot.csv.deseq2_rld.csv" % (
data_lbl_all_dh, avg_col)
elif transform_type == 'vst':
data_lbl_tran_fh = "%s/%s.deseq2_annot.csv.deseq2_vsd.csv" % (
data_lbl_all_dh, avg_col)
if not exists(data_lbl_tran_fh):
data_deseq2_annot = pd.DataFrame(columns=[
'condition', 'type', 'number of lanes',
'total number of reads', 'exon counts'
])
data_deseq2_annot.index.name = 'file'
data_deseq2_annot_fh = '%s/%s.deseq2_annot.csv' % (
data_lbl_all_dh, avg_col)
data_deseq2_count = pd.DataFrame()
data_deseq2_count_fh = '%s/%s.deseq2_count.csv' % (
data_lbl_all_dh, avg_col)
for rep in repli.loc[avg, :]:
if not pd.isnull(rep):
rep_col = "%s%s%s" % (rep, col_sep, data_lbl_col)
data_deseq2_annot.loc[rep_col, 'condition'] = rep
if len(data_deseq2_count) == 0:
data_deseq2_count = pd.DataFrame(
data_lbl_all.loc[:, rep_col].copy())
else:
data_deseq2_count.loc[:,
rep_col] = data_lbl_all.loc[:,
rep_col]
# data_deseq2_annot_index_all=["%s%s%s" % (basename(fh),col_sep,data_lbl_col) \
# for fh in data_lbl_fhs if (basename(fh) not in repli.index)]
# data_deseq2_annot_index_no_reps=[i for i in data_deseq2_annot_index_all if (i not in data_deseq2_annot.index)]
# for idx in data_deseq2_annot_index_no_reps:
# data_deseq2_annot.loc[idx,'condition']=idx
data_deseq2_annot.to_csv(data_deseq2_annot_fh)
data_deseq2_count = debad(data_deseq2_count,
axis=0,
condi='any',
bad='nan')
data_deseq2_count.to_csv(data_deseq2_count_fh)
deseq_fh = "%s/deseq2.R" % (abspath(dirname(__file__)))
log_fh = "%s.log" % data_deseq2_annot_fh
with open(log_fh, 'a') as log_f:
deseq2_com = '%s %s %s %s 1' % (rscript_fh, deseq_fh,
data_deseq2_count_fh,
data_deseq2_annot_fh)
# print deseq2_com
subprocess.call(deseq2_com,
shell=True,
stdout=log_f,
stderr=subprocess.STDOUT)
data_lbl_tran = pd.read_csv(data_lbl_tran_fh).set_index(
'Unnamed: 0')
data_lbl_tran.index.name = 'mutids'
if len(data_lbl_all_tran) == 0:
data_lbl_all_tran = data_lbl_tran.copy()
else:
data_lbl_all_tran = data_lbl_all_tran.join(data_lbl_tran)
# print data_lbl_all_tran.columns.tolist()
# print len(data_lbl_all_tran)
if len(data_lbl_all_tran.columns.tolist()) > 0:
data_lbl_all_tran.to_csv(data_lbl_all_tran_fh)
else:
logging.error('transform_type can not be %s: no replicates found' %
transform_type)
sys.exit()
else:
data_lbl_all_tran = pd.read_csv(data_lbl_all_tran_fh).set_index(
'mutids')
data_lbl_tran_col = 'NiA_tran'
for col in data_lbl_all_tran:
data_lbl_fn, tmp = col.split('.')
data_lbl_fh = '%s/data_lbl/%s/%s' % (prj_dh, type_form, data_lbl_fn)
data_lbl = pd.read_csv(data_lbl_fh).set_index('mutids')
if not data_lbl_tran_col in data_lbl:
data_lbl_cols = data_lbl.columns.tolist()
data_lbl = data_lbl.join(data_lbl_all_tran.loc[:, col])
data_lbl.columns = data_lbl_cols + [data_lbl_tran_col]
data_lbl.index.name = 'mutids'
data_lbl.to_csv(data_lbl_fh)
def plot_data_comparison_multiviolin(prj_dh,data_fits,col,
data_fiti_ctrl=0,
aasORcds="aas",
ns=True,numeric=False,
color_test=(0.6, 0.8, 1),#'mediumpurple'#"lime"
color_ctrl="lightgray",
data_fits_labels=None,
pval=True,
stars=True,
violin_width=0.9,
color_xticks=None,
force=False,
ylims=None,
col_hue='Conditions',
label_test='Test',
label_ctrl='Control',
ylabel=None,
figsize=[4,3],
plot_fh=None):
"""
Plotting distributions of comparison of fold change data
:param prj_dh: path to project directory
:param data_fits: pandas dataframe with fold change data
:param col: column with the fold change data
"""
data_fit_fhs=["%s/data_fit/%s/%s" % (prj_dh,aasORcds,s) for s in data_fits]
if plot_fh!=None:
data_comparison_ctrl_fh=plot_fh+"data_comparison_ctrl.csv"
data_comparison_test_fh=plot_fh+"data_comparison_test.csv"
else:
data_comparison_ctrl_fh="data_comparison_ctrl.csv"
data_comparison_test_fh="data_comparison_test.csv"
if (not exists(data_comparison_ctrl_fh)) or\
(not exists(data_comparison_test_fh)) or force:
data_fit_test_fhs=[fh for i,fh in enumerate(data_fit_fhs) if i!=0]
data_fit_ctrl_fhs=[fh for i,fh in enumerate(data_fit_fhs) if i==0]
data_comparison_test=fhs2data_combo(data_fit_test_fhs,[col],index='mutids')
data_comparison_ctrl=fhs2data_combo(data_fit_ctrl_fhs,[col],index='mutids')
data_comparison_test.columns=data_fits[1:]
data_comparison_ctrl.columns=data_fits[1:]
#data_comparison_test.to_csv(data_comparison_test_fh)
#data_comparison_ctrl.to_csv(data_comparison_ctrl_fh)
else:
data_comparison_test=pd.read_csv(data_comparison_test_fh).set_index('mutids')
data_comparison_ctrl=pd.read_csv(data_comparison_ctrl_fh).set_index('mutids')
data_all=pd.concat([data_comparison_ctrl.loc[:,data_comparison_ctrl.columns.tolist()[0]],
data_comparison_test],axis=1)
data_all.columns=data_fits
data_comparison_test=data_comparison_test.unstack().reset_index()
data_comparison_ctrl=data_comparison_ctrl.unstack().reset_index()
data_comparison_test.columns=["condi","mutids",col]
data_comparison_ctrl.columns=["condi","mutids",col]
data_comparison_test.loc[:,col_hue]=label_test
data_comparison_ctrl.loc[:,col_hue]=label_ctrl
data_comparison=data_comparison_test.append(data_comparison_ctrl)
y_max=data_comparison.loc[:,col].max()
y_min=data_comparison.loc[:,col].min()
data_fit_ctrl=data_fits[data_fiti_ctrl]
data_fit_tests=[s for s in data_fits if data_fits.index(s)!=data_fiti_ctrl]
# fig=
plt.figure(figsize=figsize,dpi=300)
ax=plt.subplot(111)
import seaborn as sns
from scipy.stats import mannwhitneyu,wilcoxon
# plt.style.use('seaborn-whitegrid')
# plt.style.use('seaborn-white')
sns.violinplot(x="condi", y=col,
hue=col_hue,
split=True,
data=data_comparison,
# color='m',
palette={label_test: color_test,
label_ctrl: color_ctrl},
inner="quartile",
width=violin_width,
cut=0,
# bw=0.1,
scale="width",ax=ax)
plt.legend(title=col_hue,loc='upper center',
bbox_to_anchor=(0.5, 1.1),#, 1., .102),
ncol=2, borderaxespad=0.1,frameon=True,
)
ax.set_xlabel("")
if not ylabel is None:
ax.set_ylabel(ylabel)
# ax.grid(b=True)
ax.yaxis.grid(True)
if data_fits_labels!=None:
ax.set_xticklabels(data_fits_labels[1:])
if ylabel!=None:
ax.set_ylabel(ylabel)
plt.tight_layout()
if ylims is None:
ax.set_ylim([y_min,y_max])
else:
ax.set_ylim([ylims[0],ylims[1]])
# return data_comparison,data_all
if pval:
col_ctrl=data_all.columns.tolist()[0]
col_tests=data_all.columns.tolist()[1:]
for col_testi in range(len(col_tests)):
col_test=col_tests[col_testi]
pval,side=get_wilcoxon(data_all,col_ctrl,col_test,side='one',denan=False)
side='' #stitch
# print pval
if stars:
pval=pval2stars(pval,ns=ns,numeric=False)
if ns and (pval>0.05):
side=''
pval=''
result="%s\n%s" % (side,pval)
else:
result="%s\n%s" % (side,pval)
# print result
# print ax.get_ylim()[0]+(ax.get_ylim()[1]-ax.get_ylim()[0])*0.05
ax.text(col_testi,ax.get_ylim()[0]+(ax.get_ylim()[1]-ax.get_ylim()[0])*0.05,
result,ha='center',color='b',
bbox=dict(facecolor='w', edgecolor='none',
# boxstyle='round',
alpha=0.6,)
)
# data_all.to_csv('test1.csv')
if plot_fh!=None:
#plt.tight_layout()
plt.savefig(plot_fh,format='pdf')
plt.clf();plt.close()
return ax,data_comparison,data_all