def make_bool_hist(n_true, n_false, hist_label):
""" fill a two-bin histogram with the fraction false in the first bin and the fraction true in the second """
hist = TH1D(hist_label, '', 2, -0.5, 1.5)
hist.Sumw2()
true_frac = float(n_true) / (n_true + n_false)
hist.SetBinContent(1, true_frac)
true_bounds = fraction_uncertainty.err(n_true,
n_true + n_false,
use_beta=True)
hist.SetBinError(
1,
max(abs(true_frac - true_bounds[0]),
abs(true_bounds[1] - true_bounds[1])))
false_frac = float(n_false) / (n_true + n_false)
hist.SetBinContent(2, false_frac)
false_bounds = fraction_uncertainty.err(n_false,
n_true + n_false,
use_beta=True)
hist.SetBinError(
2,
max(abs(false_frac - false_bounds[0]),
abs(false_bounds[1] - false_bounds[1])))
hist.GetXaxis().SetNdivisions(0)
hist.GetXaxis().SetBinLabel(1, 'right')
hist.GetXaxis().SetBinLabel(2, 'wrong')
hist.GetXaxis().SetLabelSize(0.1)
return hist
def finalize(self, calculate_uncertainty=True):
""" convert from counts to mut freqs """
assert not self.finalized
self.n_cached, self.n_not_cached = 0, 0
for gene in self.counts:
self.freqs[gene], self.plotting_info[gene] = {}, []
# NOTE <counts> hold the overall (not per-base) frequencies, while <freqs> holds the per-base frequencies
counts, freqs, plotting_info = self.counts[gene], self.freqs[
gene], self.plotting_info[gene]
sorted_positions = sorted(counts)
for position in sorted_positions:
freqs[position] = {}
plotting_info.append({})
plotting_info[-1]['name'] = utils.sanitize_name(
gene) + '_' + str(position)
plotting_info[-1]['nuke_freqs'] = {}
n_conserved, n_mutated = 0, 0
for nuke in utils.nukes:
nuke_freq = float(
counts[position][nuke]) / counts[position]['total']
freqs[position][nuke] = nuke_freq
plotting_info[-1]['nuke_freqs'][nuke] = nuke_freq
if calculate_uncertainty: # it's kinda slow
errs = fraction_uncertainty.err(
counts[position][nuke], counts[position]['total'])
if errs[2]:
self.n_cached += 1
else:
self.n_not_cached += 1
# print nuke_freq, errs[0], errs[1], '(', counts[position][nuke], ',', counts[position]['total'], ')'
assert errs[
0] <= nuke_freq # these checks are probably unnecessary. EDIT and totally saved my ass about ten minutes after writing the previous statement
assert nuke_freq <= errs[1]
freqs[position][nuke + '_lo_err'] = errs[0]
freqs[position][nuke + '_hi_err'] = errs[1]
if nuke == counts[position]['gl_nuke']:
n_conserved += counts[position][nuke]
else:
n_mutated += counts[position][nuke] # sum over A,C,G,T
# uncert = fraction_uncertainty.err(obs, total) # uncertainty for each nuke
counts[position]['freq'] = float(
n_mutated) / counts[position]['total']
mutated_fraction_err = (0.0, 0.0)
if calculate_uncertainty: # it's kinda slow
mutated_fraction_err = fraction_uncertainty.err(
n_mutated, counts[position]['total'])
if mutated_fraction_err[2]:
self.n_cached += 1
else:
self.n_not_cached += 1
counts[position]['freq_lo_err'] = mutated_fraction_err[0]
counts[position]['freq_hi_err'] = mutated_fraction_err[1]
self.mean_rates['all'].normalize()
for region in utils.regions:
self.mean_rates[region].normalize()
self.finalized = True
def finalize(self, calculate_uncertainty=True):
""" convert from counts to mut freqs """
assert not self.finalized
self.n_cached, self.n_not_cached = 0, 0
for gene in self.counts:
self.freqs[gene], self.plotting_info[gene] = {}, []
# NOTE <counts> hold the overall (not per-base) frequencies, while <freqs> holds the per-base frequencies
counts, freqs, plotting_info = self.counts[gene], self.freqs[gene], self.plotting_info[gene]
sorted_positions = sorted(counts)
for position in sorted_positions:
freqs[position] = {}
plotting_info.append({})
plotting_info[-1]['name'] = utils.sanitize_name(gene) + '_' + str(position)
plotting_info[-1]['nuke_freqs'] = {}
n_conserved, n_mutated = 0, 0
for nuke in utils.nukes:
nuke_freq = float(counts[position][nuke]) / counts[position]['total']
freqs[position][nuke] = nuke_freq
plotting_info[-1]['nuke_freqs'][nuke] = nuke_freq
if calculate_uncertainty: # it's kinda slow
errs = fraction_uncertainty.err(counts[position][nuke], counts[position]['total'])
if errs[2]:
self.n_cached += 1
else:
self.n_not_cached += 1
# print nuke_freq, errs[0], errs[1], '(', counts[position][nuke], ',', counts[position]['total'], ')'
assert errs[0] <= nuke_freq # these checks are probably unnecessary. EDIT and totally saved my ass about ten minutes after writing the previous statement
assert nuke_freq <= errs[1]
freqs[position][nuke + '_lo_err'] = errs[0]
freqs[position][nuke + '_hi_err'] = errs[1]
if nuke == counts[position]['gl_nuke']:
n_conserved += counts[position][nuke]
else:
n_mutated += counts[position][nuke] # sum over A,C,G,T
# uncert = fraction_uncertainty.err(obs, total) # uncertainty for each nuke
counts[position]['freq'] = float(n_mutated) / counts[position]['total']
mutated_fraction_err = (0.0, 0.0)
if calculate_uncertainty: # it's kinda slow
mutated_fraction_err = fraction_uncertainty.err(n_mutated, counts[position]['total'])
if mutated_fraction_err[2]:
self.n_cached += 1
else:
self.n_not_cached += 1
counts[position]['freq_lo_err'] = mutated_fraction_err[0]
counts[position]['freq_hi_err'] = mutated_fraction_err[1]
self.mean_rates['all'].normalize(overflow_warn=False) # we expect overflows in mute freq hists, so no need to warn us
for region in utils.regions:
self.mean_rates[region].normalize(overflow_warn=False)
# for gene in self.tmpcounts:
# for position in self.tmpcounts[gene]:
# self.tmpcounts[gene][position]['muted'].divide_by(self.tmpcounts[gene][position]['total'], debug=False)
self.finalized = True
def plot(self, plotdir):
utils.prep_dir(plotdir + '/plots', wildling=None, multilings=['*.csv', '*.svg', '*.root'])
for column in self.values:
if self.only_correct_gene_fractions and column not in bool_columns:
continue
if column in bool_columns:
right = self.values[column]['right']
wrong = self.values[column]['wrong']
errs = fraction_uncertainty.err(right, right+wrong)
print ' %s\n correct up to allele: %4d / %-4d = %4.4f (-%.3f, +%.3f)' % (column, right, right+wrong, float(right) / (right + wrong), errs[0], errs[1])
hist = plotting.make_bool_hist(right, wrong, self.name + '-' + column)
plotting.draw(hist, 'bool', plotname=column, plotdir=plotdir, write_csv=True)
else:
# TODO this is dumb... I should make the integer-valued ones histograms as well
hist = plotting.make_hist_from_dict_of_counts(self.values[column], 'int', self.name + '-' + column, normalize=True)
log = ''
if column.find('hamming_to_true_naive') >= 0:
hist.GetXaxis().SetTitle('hamming distance')
else:
hist.GetXaxis().SetTitle('inferred - true')
plotting.draw(hist, 'int', plotname=column, plotdir=plotdir, write_csv=True, log=log)
for column in self.hists:
hist = plotting.make_hist_from_my_hist_class(self.hists[column], column)
plotting.draw(hist, 'float', plotname=column, plotdir=plotdir, write_csv=True, log=log)
check_call(['./bin/makeHtml', plotdir, '3', 'null', 'svg'])
check_call(['./bin/permissify-www', plotdir]) # NOTE this should really permissify starting a few directories higher up
def plot(self, plotdir, only_csv=False):
utils.prep_dir(plotdir, wildling=None, multilings=['*.csv', '*.svg', '*.root'])
for column in self.values:
if self.only_correct_gene_fractions and column not in bool_columns:
continue
if column in bool_columns:
right = self.values[column]['right']
wrong = self.values[column]['wrong']
errs = fraction_uncertainty.err(right, right+wrong)
print ' %s\n correct up to allele: %4d / %-4d = %4.4f (-%.3f, +%.3f)' % (column, right, right+wrong, float(right) / (right + wrong), errs[0], errs[1])
hist = plotting.make_bool_hist(right, wrong, self.name + '-' + column)
plotting.draw_no_root(hist, plotname=column, plotdir=plotdir, write_csv=True, stats='0-bin', only_csv=only_csv)
else:
# TODO this is dumb... I should make the integer-valued ones histograms as well
hist = plotting.make_hist_from_dict_of_counts(self.values[column], 'int', self.name + '-' + column, normalize=True)
log = ''
if column.find('hamming_to_true_naive') >= 0: # TODO why doesn't this just use the config dicts in plotheaders or wherever?
hist.title = 'hamming distance'
else:
hist.title = 'inferred - true'
plotting.draw_no_root(hist, plotname=column, plotdir=plotdir, write_csv=True, log=log, only_csv=only_csv)
for column in self.hists:
plotting.draw_no_root(self.hists[column], plotname=column, plotdir=plotdir, write_csv=True, log=log, only_csv=only_csv)
if not only_csv:
plotting.make_html(plotdir)
def get_single_vals(pv):
yvals = [float(c) / t for c, t in zip(pv['ycounts'], pv['ytotals'])] # total shouldn't be able to be zero
tmphilos = [fraction_uncertainty.err(c, t) for c, t in zip(pv['ycounts'], pv['ytotals'])]
yerrs = [err[1] - err[0] for err in tmphilos]
print ' %s %s' % (xlabel, ylabel)
for iv in range(len(pv['xvals'])):
print ' %8.0f %5.0f / %-5.0f = %5.2f +/- %.3f' % (pv['xvals'][iv], pv['ycounts'][iv], pv['ytotals'][iv], yvals[iv], yerrs[iv])
return pv['xvals'], yvals, yerrs
def make_bool_hist(n_true, n_false, hist_label):
""" fill a two-bin histogram with the fraction false in the first bin and the fraction true in the second """
hist = TH1D(hist_label, '', 2, -0.5, 1.5)
hist.Sumw2()
true_frac = float(n_true) / (n_true + n_false)
hist.SetBinContent(1, true_frac)
true_bounds = fraction_uncertainty.err(n_true, n_true + n_false, use_beta=True)
hist.SetBinError(1, max(abs(true_frac - true_bounds[0]), abs(true_bounds[1] - true_bounds[1])))
false_frac = float(n_false) / (n_true + n_false)
hist.SetBinContent(2, false_frac)
false_bounds = fraction_uncertainty.err(n_false, n_true + n_false, use_beta=True)
hist.SetBinError(2, max(abs(false_frac - false_bounds[0]), abs(false_bounds[1] - false_bounds[1])))
hist.GetXaxis().SetNdivisions(0)
hist.GetXaxis().SetBinLabel(1, 'right')
hist.GetXaxis().SetBinLabel(2, 'wrong')
hist.GetXaxis().SetLabelSize(0.1)
return hist
def get_uncertainty(self, obs, total):
if self.calculate_uncertainty: # it's kinda slow
errs = fraction_uncertainty.err(obs, total)
if errs[2]:
self.n_cached += 1
else:
self.n_not_cached += 1
else:
errs = 0., 1.
return errs[0], errs[1]
def plot(self, plotdir, only_csv=False):
print ' plotting performance',
import fraction_uncertainty
import plotting
start = time.time()
for substr in self.subplotdirs:
utils.prep_dir(plotdir + '/' + substr, wildlings=('*.csv', '*.svg'))
for column in self.values:
if column in plotconfig.gene_usage_columns:
right = self.values[column]['right']
wrong = self.values[column]['wrong']
lo, hi = fraction_uncertainty.err(right, right + wrong)
hist = plotting.make_bool_hist(right, wrong, self.name + '-' + column)
plotting.draw_no_root(hist, plotname=column, plotdir=plotdir + '/gene-call', write_csv=True, stats='0-bin', only_csv=only_csv)
else:
hist = plotting.make_hist_from_dict_of_counts(self.values[column], 'int', self.name + '-' + column, normalize=False)
if 'hamming_to_true_naive' in column:
xtitle = 'hamming distance'
tmpplotdir = plotdir + '/mutation'
else:
xtitle = 'inferred - true'
if 'muted' in column:
tmpplotdir = plotdir + '/mutation'
else:
tmpplotdir = plotdir + '/boundaries'
plotting.draw_no_root(hist, plotname=column, plotdir=tmpplotdir, write_csv=True, only_csv=only_csv, xtitle=xtitle, shift_overflows=True)
for column in self.hists:
if '_vs_mute_freq' in column or '_vs_per_gene_support' in column: # only really care about the fraction, which we plot below
continue
plotting.draw_no_root(self.hists[column], plotname=column, plotdir=plotdir + '/mutation', write_csv=True, only_csv=only_csv, ytitle='counts', xtitle='inferred - true', shift_overflows=True)
# fraction correct vs mute freq
for region in utils.regions:
hright = self.hists[region + '_gene_right_vs_mute_freq']
hwrong = self.hists[region + '_gene_wrong_vs_mute_freq']
if hright.integral(include_overflows=True) == 0:
continue
plotting.make_fraction_plot(hright, hwrong, plotdir + '/gene-call', region + '_fraction_correct_vs_mute_freq', xlabel='mut freq', ylabel='fraction correct up to allele', xbounds=(0., 0.5), only_csv=only_csv, write_csv=True)
# per-gene support stuff
for region in utils.regions:
if self.hists[region + '_allele_right_vs_per_gene_support'].integral(include_overflows=True) == 0:
continue
hright = self.hists[region + '_allele_right_vs_per_gene_support']
hwrong = self.hists[region + '_allele_wrong_vs_per_gene_support']
plotting.make_fraction_plot(hright, hwrong, plotdir + '/gene-call', region + '_allele_fraction_correct_vs_per_gene_support', xlabel='support', ylabel='fraction with correct allele', xbounds=(-0.1, 1.1), only_csv=only_csv, write_csv=True)
if not only_csv: # write html file and fix permissiions
for substr in self.subplotdirs:
plotting.make_html(plotdir + '/' + substr, n_columns=4)
print '(%.1f sec)' % (time.time()-start)
def get_uncertainty(self, obs, total):
import fraction_uncertainty
if self.calculate_uncertainty: # it's kinda slow
errs = fraction_uncertainty.err(obs, total)
# if errs[2]:
# self.n_cached += 1
# else:
# self.n_not_cached += 1
else:
errs = 0., 1.
return errs[0], errs[1]
def plot(self, plotdir, only_csv=False):
print ' plotting performance',
start = time.time()
for substr in self.subplotdirs:
utils.prep_dir(plotdir + '/' + substr, wildlings=('*.csv', '*.svg'))
for column in self.values:
if column in bool_columns:
right = self.values[column]['right']
wrong = self.values[column]['wrong']
lo, hi, _ = fraction_uncertainty.err(right, right + wrong)
hist = plotting.make_bool_hist(right, wrong, self.name + '-' + column)
plotting.draw_no_root(hist, plotname=column, plotdir=plotdir + '/gene-call', write_csv=True, stats='0-bin', only_csv=only_csv)
else:
hist = plotting.make_hist_from_dict_of_counts(self.values[column], 'int', self.name + '-' + column, normalize=False)
if 'hamming_to_true_naive' in column:
xtitle = 'hamming distance'
tmpplotdir = plotdir + '/mutation'
else:
xtitle = 'inferred - true'
if 'muted' in column:
tmpplotdir = plotdir + '/mutation'
else:
tmpplotdir = plotdir + '/boundaries'
plotting.draw_no_root(hist, plotname=column, plotdir=tmpplotdir, write_csv=True, only_csv=only_csv, xtitle=xtitle, shift_overflows=True)
for column in self.hists:
if '_vs_mute_freq' in column or '_vs_per_gene_support' in column: # only really care about the fraction, which we plot below
continue
plotting.draw_no_root(self.hists[column], plotname=column, plotdir=plotdir + '/mutation', write_csv=True, only_csv=only_csv, ytitle='counts', xtitle='inferred - true', shift_overflows=True)
# fraction correct vs mute freq
for region in utils.regions:
hright = self.hists[region + '_gene_right_vs_mute_freq']
hwrong = self.hists[region + '_gene_wrong_vs_mute_freq']
if hright.integral(include_overflows=True) == 0:
continue
plotting.make_fraction_plot(hright, hwrong, plotdir + '/gene-call', region + '_fraction_correct_vs_mute_freq', xlabel='mut freq', ylabel='fraction correct up to allele', xbounds=(0., 0.5), only_csv=only_csv, write_csv=True)
# per-gene support stuff
for region in utils.regions:
if self.hists[region + '_allele_right_vs_per_gene_support'].integral(include_overflows=True) == 0:
continue
hright = self.hists[region + '_allele_right_vs_per_gene_support']
hwrong = self.hists[region + '_allele_wrong_vs_per_gene_support']
plotting.make_fraction_plot(hright, hwrong, plotdir + '/gene-call', region + '_allele_fraction_correct_vs_per_gene_support', xlabel='support', ylabel='fraction with correct allele', xbounds=(-0.1, 1.1), only_csv=only_csv, write_csv=True)
if not only_csv: # write html file and fix permissiions
for substr in self.subplotdirs:
plotting.make_html(plotdir + '/' + substr, n_columns=4)
print '(%.1f sec)' % (time.time()-start)
def make_hist_from_bin_entry_file(fname, hist_label='', log=''):
hist = Hist(fname=fname)
if '_gene' in fname and '_vs_' not in fname:
import re
tot = 30000
correct = int(tot * hist.bin_contents[1])
incorrect = int(tot * hist.bin_contents[2])
frac = float(correct) / (correct + incorrect)
# print correct, incorrect, tot
bounds = fraction_uncertainty.err(correct, correct + incorrect, for_paper=True)
print ' ', re.findall('[vdj]_gene', fname)[0], frac, 0.5*(bounds[1] - bounds[0])
# sys.exit()
roothist = make_hist_from_my_hist_class(hist, hist_label)
return roothist
def get_allele_finding_xyvals(self, gene, position):
gpcounts = self.counts[gene][position]
iterinfo = gpcounts.items()
obs = [d['muted'] for nm, d in iterinfo if nm < self.n_max_mutations_per_segment]
lohis = [fraction_uncertainty.err(d['muted'], d['total'], use_beta=True) for nm, d in iterinfo if nm < self.n_max_mutations_per_segment]
errs = [(hi - lo) / 2 for lo, hi, _ in lohis]
weights = [1./(e*e) for e in errs]
freqs = [float(d['muted']) / d['total'] if d['total'] > 0 else 0. for nm, d in iterinfo if nm < self.n_max_mutations_per_segment]
total = [d['total'] for nm, d in iterinfo if nm < self.n_max_mutations_per_segment]
n_mutelist = [nm for nm in gpcounts.keys() if nm < self.n_max_mutations_per_segment]
return {'obs' : obs, 'total' : total, 'n_mutelist' : n_mutelist, 'freqs' : freqs, 'errs' : errs, 'weights' : weights}
def add_gene_calls_vs_mute_freq_plots(args, hists, rebin=1.):
for idir in range(len(args.names)):
name = args.names[idir]
for region in utils.regions:
hright = hists[idir][region + '_gene_right_vs_mute_freq']
hwrong = hists[idir][region + '_gene_wrong_vs_mute_freq']
hdenom = TH1D(hright)
hdenom.Add(hwrong)
hfrac = TH1D(hright)
hfrac.Divide(hdenom)
hfrac.Scale(1. / rebin)
# print name, region
for ib in range(hfrac.GetNbinsX()+2):
lo, hi, cached = fraction_uncertainty.err(hright.GetBinContent(ib), hdenom.GetBinContent(ib), for_paper=True)
hfrac.SetBinError(ib, (hi - lo) / 2.)
# print '%5d %5d %.2f %.3f' % (hright.GetBinContent(ib), hdenom.GetBinContent(ib), hfrac.GetBinContent(ib), (hi - lo) / 2.)
hists[idir][region + '_gene_fraction_vs_mute_freq'] = hfrac
def make_hist_from_bin_entry_file(fname, hist_label='', log=''):
hist = Hist(fname=fname)
if '_gene' in fname and '_vs_' not in fname:
import re
tot = 30000
correct = int(tot * hist.bin_contents[1])
incorrect = int(tot * hist.bin_contents[2])
frac = float(correct) / (correct + incorrect)
# print correct, incorrect, tot
bounds = fraction_uncertainty.err(correct,
correct + incorrect,
for_paper=True)
print ' ', re.findall('[vdj]_gene',
fname)[0], frac, 0.5 * (bounds[1] - bounds[0])
# sys.exit()
roothist = make_hist_from_my_hist_class(hist, hist_label)
return roothist
def add_gene_calls_vs_mute_freq_plots(args, hists, rebin=1., debug=False):
print 'TODO what\'s up with rebin rescaling below?'
for idir in range(len(args.names)):
name = args.names[idir]
for region in utils.regions:
hright = hists[idir][region + '_gene_right_vs_mute_freq']
hwrong = hists[idir][region + '_gene_wrong_vs_mute_freq']
hdenom = copy.deepcopy(hright)
hdenom.add(hwrong)
hfrac = copy.deepcopy(hright)
hfrac.divide_by(hdenom)
# hfrac.Scale(1. / rebin)
if debug:
print name, region
for ib in range(hfrac.n_bins + 2):
lo, hi, cached = fraction_uncertainty.err(hright.bin_contents[ib], hdenom.bin_contents[ib], for_paper=True)
hfrac.errors[ib] = (hi - lo) / 2.
if debug:
print '%5d %5d %.2f %.3f' % (hright.bin_contents[ib], hdenom.bin_contents[ib], hfrac.bin_contents[ib], (hi - lo) / 2.)
hists[idir][region + '_gene_fraction_vs_mute_freq'] = hfrac
def add_gene_calls_vs_mute_freq_plots(args, hists, rebin=1.):
for idir in range(len(args.names)):
name = args.names[idir]
for region in utils.regions:
hright = hists[idir][region + '_gene_right_vs_mute_freq']
hwrong = hists[idir][region + '_gene_wrong_vs_mute_freq']
hdenom = TH1D(hright)
hdenom.Add(hwrong)
hfrac = TH1D(hright)
hfrac.Divide(hdenom)
hfrac.Scale(1. / rebin)
# print name, region
for ib in range(hfrac.GetNbinsX() + 2):
lo, hi, cached = fraction_uncertainty.err(
hright.GetBinContent(ib),
hdenom.GetBinContent(ib),
for_paper=True)
hfrac.SetBinError(ib, (hi - lo) / 2.)
# print '%5d %5d %.2f %.3f' % (hright.GetBinContent(ib), hdenom.GetBinContent(ib), hfrac.GetBinContent(ib), (hi - lo) / 2.)
hists[idir][region + '_gene_fraction_vs_mute_freq'] = hfrac
def tigger_calcs(self, position, gcounts, mean_x_icpt):
iterinfo = gcounts['tigger'].items()
obs = [d['muted'] for nm, d in iterinfo if nm < self.n_max_mutes]
if sum(obs) < self.n_obs_min: # ignore positions with only a few observed mutations
return
lohis = [fraction_uncertainty.err(d['muted'], d['total'], use_beta=True) for nm, d in iterinfo if nm < self.n_max_mutes]
errs = [(hi - lo) / 2 for lo, hi, _ in lohis]
weights = [1./(e*e) for e in errs]
freqs = [float(d['muted']) / d['total'] for nm, d in iterinfo if nm < self.n_max_mutes]
total = [d['total'] for nm, d in iterinfo if nm < self.n_max_mutes]
# for i in range(len(freqs)):
# print ' %3d / %3d = %6.2f %6.2f %6.2f' % (obs[i], total[i], freqs[i], errs[i], weights[i])
n_mutelist = [nm for nm in gcounts['tigger'].keys() if nm < self.n_max_mutes]
params, cov = numpy.polyfit(n_mutelist, freqs, 1, w=weights, cov=True)
slope, slope_err = params[0], math.sqrt(cov[0][0])
y_icpt, y_icpt_err = params[1], math.sqrt(cov[1][1])
interesting = False
if y_icpt + y_icpt_err < 1./8:
x_icpt, x_icpt_err = -y_icpt / slope, abs(y_icpt / slope) * math.sqrt((y_icpt_err/y_icpt)**2 + (slope_err/slope)**2)
mean_x_icpt['sum'] += x_icpt / x_icpt_err
mean_x_icpt['total'] += 1. / x_icpt_err
else:
x_icpt, x_icpt_err = 0, 0
interesting = True
print_str = ' %3d %9.3f +/- %-9.3f %9.3f +/- %-9.3f %7.4f +/- %7.4f %3d / %3d' % (position, x_icpt, x_icpt_err, y_icpt, y_icpt_err, slope, slope_err, sum(obs), sum(total))
if interesting:
print_str = utils.color('red', print_str)
print print_str
# for testing it's easier to make plots here:
# plotinfo = {'n_muted' : n_mutelist, 'freqs' : freqs, 'errs' : errs, 'slope' : slope, 'intercept' : y_icpt}
# plotting.make_tigger_plot('IGHVX', position, plotinfo)
return plotinfo
def frac_err(obs, total):
lo, hi = fraction_uncertainty.err(obs, total)
return 0.5 * (hi - lo)
def set_bin(numer, denom, ibin, label):
frac = float(numer) / denom
bounds = fraction_uncertainty.err(numer, denom, use_beta=True)
err = max(abs(frac - bounds[0]), abs(frac - bounds[1]))
hist.set_ibin(ibin, frac, error=err, label=label)
def set_bin(numer, denom, ibin, label):
frac = float(numer) / denom
bounds = fraction_uncertainty.err(numer, denom, use_beta=True)
err = max(abs(frac - bounds[0]), abs(frac - bounds[1]))
hist.set_ibin(ibin, frac, error=err, label=label)
def plot(self, plotdir, only_csv=False):
utils.prep_dir(plotdir, wildling=None, multilings=['*.csv', '*.svg', '*.root'])
for column in self.values:
if self.only_correct_gene_fractions and column not in bool_columns:
continue
if column in bool_columns:
right = self.values[column]['right']
wrong = self.values[column]['wrong']
errs = fraction_uncertainty.err(right, right+wrong)
print ' %s\n correct up to allele: %4d / %-4d = %4.4f (-%.3f, +%.3f)' % (column, right, right+wrong, float(right) / (right + wrong), errs[0], errs[1])
hist = plotting.make_bool_hist(right, wrong, self.name + '-' + column)
plotting.draw_no_root(hist, plotname=column, plotdir=plotdir, write_csv=True, stats='0-bin', only_csv=only_csv)
else:
# TODO this is dumb... I should make the integer-valued ones histograms as well
hist = plotting.make_hist_from_dict_of_counts(self.values[column], 'int', self.name + '-' + column, normalize=True)
log = ''
if column.find('hamming_to_true_naive') >= 0: # TODO why doesn't this just use the config dicts in plotheaders or wherever?
hist.title = 'hamming distance'
else:
hist.title = 'inferred - true'
plotting.draw_no_root(hist, plotname=column, plotdir=plotdir, write_csv=True, log=log, only_csv=only_csv)
for column in self.hists:
plotting.draw_no_root(self.hists[column], plotname=column, plotdir=plotdir, write_csv=True, log=log, only_csv=only_csv)
# per-gene support crap
for region in utils.regions:
if self.hists[region + '_allele_right_vs_per_gene_support'].integral(include_overflows=True) == 0:
continue
xvals = self.hists[region + '_allele_right_vs_per_gene_support'].get_bin_centers() #ignore_overflows=True)
right = self.hists[region + '_allele_right_vs_per_gene_support'].bin_contents
wrong = self.hists[region + '_allele_wrong_vs_per_gene_support'].bin_contents
yvals = [float(r) / (r + w) if r + w > 0. else 0. for r, w in zip(right, wrong)]
# remove values corresponding to bins with no entries
while yvals.count(0.) > 0:
iv = yvals.index(0.)
xvals.pop(iv)
right.pop(iv)
wrong.pop(iv)
yvals.pop(iv)
tmphilos = [fraction_uncertainty.err(r, r + w) for r, w in zip(right, wrong)]
yerrs = [err[1] - err[0] for err in tmphilos]
# fitting a line isn't particularly informative, actually
# params, cov = numpy.polyfit(xvals, yvals, 1, w=[1./(e*e) if e > 0. else 0. for e in yerrs], cov=True)
# slope, slope_err = params[0], math.sqrt(cov[0][0])
# y_icpt, y_icpt_err = params[1], math.sqrt(cov[1][1])
# print '%s slope: %5.2f +/- %5.2f y-intercept: %5.2f +/- %5.2f' % (region, slope, slope_err, y_icpt, y_icpt_err)
# print '%s' % region
# for iv in range(len(xvals)):
# print ' %5.2f %5.0f / %5.0f = %5.2f +/- %.3f' % (xvals[iv], right[iv], right[iv] + wrong[iv], yvals[iv], yerrs[iv])
fig, ax = plotting.mpl_init()
ax.errorbar(xvals, yvals, yerr=yerrs, markersize=10, linewidth=1, marker='.')
ax.plot((0, 1), (0, 1), color='black', linestyle='--', linewidth=3) # line with slope 1 and intercept 0
# linevals = [slope*x + y_icpt for x in [0] + xvals] # fitted line
# ax.plot([0] + xvals, linevals)
plotting.mpl_finish(ax, plotdir, region + '_allele_fraction_correct_vs_per_gene_support', xlabel='support', ylabel='fraction correct', xbounds=(-0.1, 1.1), ybounds=(-0.1, 1.1))
if not only_csv:
plotting.make_html(plotdir)