-
Notifications
You must be signed in to change notification settings - Fork 0
/
penEst.py
460 lines (393 loc) · 16.5 KB
/
penEst.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
#!/usr/bin/python
"""Penetrance Estimator plotting application core; calculates ftilde and
ftildestar penetrance estimates.
"""
# Authors: Sang-Cheol Seok <[email protected]>
# Jo Valentine-Cooper <[email protected]>
#
# Copyright (C) 2022 Mathematical Medicine LLC
# This program is free software: you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the Free
# Software Foundation, either version 3 of the License, or (at your option)
# any later version.
# You should have received a copy of the GNU General Public License along
# with this program. If not, see <https://www.gnu.org/licenses/>.
"""
% Assumes parents are phenotypically unknown and 11 x 12
% allele '2' is the disease allele
% The disease is rare dominant disease
% There is at least one kid with '12' and affected.
% That is, the first kid is '12' and affected.
% Apply the k-model in the Hodge(1998) paper page1218.
% Prob(family ascertained | r affected) = const * (r^k + t)
% HetModel, prob(aff|12) = alpha + f - alpha * f
% where r = # of kids with 'aff' and '12'
% Use const = 1/1000 and k= -1,0, 1, 2
%Inputs
% s: # of kids
% alpha: prob(aff | '11')
% f: prob(aff | '12')
%Outputs
% ftilde
% ftildestar
"""
from math import factorial
from itertools import permutations, repeat, product
import random
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd # construct dataframe
import seaborn as sns
from sinaplot import sinaplot
# Useful constants so I don't have to keep typing the TeX over and over
FT = r"$\tilde{f}$"
FTS = r"$\tilde{f}^*$"
ALPHA = r"$\alpha$"
# Shorthand for our data printing options context.
DATA_PRINT_OPTIONS = pd.option_context(
'display.max_rows', None,
'display.max_columns', None,
'display.precision', 4)
def nCr(n,r):
return factorial(n) / factorial(r) / factorial(n-r)
def simNucPedigree(N, s, alpha, f, kvalue):
# simulate Pedigree
# FIXME: Look into ways of optimizing this because it seems to be repeating
# work a lot.
if s < 10 :
const = 1/100
else :
const = 1/1000
hetProbaff= alpha + f - alpha * f
TildeRatios= [0] * N
TildeStarRatios= [0] * N
UnAff12 = 0.0
famId =0
totalfam=0
while famId < N :
totalfam = totalfam +1
GTs = []
for i in range(0,s):
n = random.randint(1,2)
GTs.append(n)
PTs = [0] * s
for k in range(s):
if GTs[k]==1: # % '11'
if random.random() < alpha :
PTs[k]=2
else:
PTs[k]=1
else: # % '12'
if random.random() < hetProbaff :
PTs[k]=2
else:
PTs[k]=1
temp =10*np.array(PTs)+np.array(GTs)
r = sum(p == 22 for p in temp) # sum((PTs==2)&(GTs==2))
t = sum(p == 2 for p in PTs)
if r> 0:
probAsct = const * (pow(r, float(kvalue)) + t) #% prob(ascertaint) using k-method const * (r^k +t)
else:
probAsct = 0
if random.random() < probAsct:
temp =10*np.array(PTs)+np.array(GTs)
UnAff12 = UnAff12 + sum(p == 12 for p in temp)
if r > 0 :
den = (r-1 + sum(p == 12 for p in temp) )
if den > 0 :
TildeRatios[famId]= (r-1)/ den
else:
TildeRatios[famId]= -1 #TildeRatios[famId]=(r-1)/(r-1 + sum(p == 12 for p in temp) )
TildeStarRatios[famId]= r/(r + sum(p == 12 for p in temp) )
famId = famId + 1
ftilde=0
for p in TildeRatios:
if p >=0 :
ftilde= ftilde +p
countNonNegative = sum(p >=0 for p in TildeRatios)
if countNonNegative >0:
ftilde= ftilde /sum(p >=0 for p in TildeRatios) #ftilde= np.mean(TildeRatios[TildeRatios>=0])
else:
ftilde =float("NaN")
ftildestar = np.mean(TildeStarRatios)
#print (N, s, ftilde , ftildestar)
#%sprintf('Aff12=%f Aff12NoProb=%f UnAff12=%f ftilde = %f ftildestar = %f',Aff12 ,Aff12NoProb, UnAff12, ftilde,ftildestar )
return ftilde, ftildestar
def exactNucPedigree(s, alpha, f, kvalue):
# FIXME: Look into ways of optimizing this as well, for the same reasons as
# above.
hetProbaff= alpha + f - alpha * f
const = 0.001
#loop over all possible cases of phenotypes for the s-1 kids
pools = sorted(list(set(
permutations([1 for x in range(s)] + [2 for x in range(s)], s))))
numRows=len(pools)
GTs=[1] * s
totalProbGT = [0] * s #zeros(s,1)
ftildes = [] #[[0]*numRows]*s
ftildeStars = [] #[[0]*numRows]*s
probFamilies = []
for numkids in range(s): # we don't need the case for no kid has '12', so we start from at least one '12' case
GTs[numkids]=2
ftilderow=[]
ftildeStarrow=[]
#print(" numkids ", numkids, "GTs", GTs)
#otherwise, there are 2 phenotype cases
#GTs
probGT=float(nCr(s,numkids+1))/pow(2,s)
# 2**s # probability of GT is 1/2^s in kids
# nchoosek is the number of liketerms
#print ("s is ",s," numkids is ",numkids," probGT is ",probGT, "nCr(s,numkids) is ", nCr(s,numkids), "pow(2,s) is ",pow(2,s))
totalProb = [0] * numRows
#%Step2. Assigning Phenotypes
for j in range(numRows):
PTs = pools[j]
#print(" numkis ", numkids, " j is " ,j, [GTs, PTs ])
probPT_GT = 1 #% prob of PT given GT
for k in range(s):
if GTs[k]==1: # % '11'
if PTs[k]==1 :
probPT_GT = probPT_GT * (1-alpha)
else:
probPT_GT = probPT_GT * alpha
else: # % '12'
if PTs[k]==1 :
probPT_GT = probPT_GT * (1 - hetProbaff) #%(1-f)
else:
probPT_GT = probPT_GT * hetProbaff #%* f
temp =10*np.array(PTs)+np.array(GTs)
r = sum(p == 22 for p in temp) # sum((PTs==2)&(GTs==2))
t = sum(p == 2 for p in PTs)
if r> 0:
probAsct = const * (pow(r, float(kvalue)) + t) #% prob(ascertaint) using k-method const * (r^k +t)
else:
probAsct = 0 #%const * t % 0
totalProb[j] = probAsct* probGT*probPT_GT
curUnAff12 = sum(p == 12 for p in temp)
den = r-1 + curUnAff12
#print( " r is ", r, " den is ", den)
if r>0:
if den <= 0 :
ftilderow.append(-1)
else:
fract=(r-1)/den
ftilderow.append(fract)
else:
ftilderow.append(0)
ftildeStarrow.append(r/( r + curUnAff12)) #ftildeStars[numkids][j] = r/( r + curUnAff12)
ftildes.append(ftilderow)
ftildeStars.append(ftildeStarrow)
probFamilies.append(totalProb)
totalProbGT[numkids]= sum(totalProb)
#print("ftilde after numkids ", numkids)
#print(ftilderow)
probFamilies = np.asarray(probFamilies).flatten()
ftildes = np.asarray(ftildes).flatten()
totalProbAsct = probFamilies.sum()
probFamiliesStripped = probFamilies[ftildes>=0]
ftildesStripped = ftildes[ftildes>=0]
ftilde = np.dot(ftildesStripped , probFamiliesStripped) / probFamiliesStripped.sum()
ftildestar = np.dot(np.asarray(ftildeStars).flatten() , probFamilies) / totalProbAsct
return ftilde, ftildestar
def listify_noniter(var):
"""Given a variable, returns it if it's an iterable, otherwise returns that
variable as a one-item tuple.
"""
try:
iter(var)
except TypeError:
return (var, )
else:
return var
class FakeProgress(object):
"""Fakes progress bar interface."""
def tick(self):
pass
def calc_stats(alphavals, fvals, svals, kvals, Nvals=None, NumReps=None,
to_stdout=False, progress=FakeProgress()):
"""Given one or more values each for two of alpha, f, s, and k (and one
value each for the others) OR given one value each for all four plus number
of families and number of simulations of families to run, calculates ftilde
and ftildestar for all values of those variables and returns them in a
dataframe.
"""
# We have two possible operating modes:
# 1) Penetrance estimate distributions as a function of number of families.
# 2) Calculation of penetrance estimation as a function of one of the
# component variables (typically alpha, but hypothetically we could do
# others) for multiple values of another component variable (typically k)
# I (Jo) don't actually know the distinctions all that well and was
# referred to the (at the time unpublished) paper when I asked, so please
# bear with me. :)
# From a data structure perspective, though, this can be summed up as
# "we have two iterables and some constants and they're used to generate a
# pair of lists of values".
# For mode 1, the iterables are Nvals and range(NumReps).
# For mode 2, the iterables are kvals and alphavals.
# So we make the following assumptions:
# * If Nvals and/or NumReps are None, we're using operating mode 2 above.
# * Otherwise, we're using mode 1.
alphavals = listify_noniter(alphavals)
for alphaval in alphavals:
if alphaval < 0 or alphaval > 1:
raise ValueError("alpha must be between 0 and 1")
fvals = listify_noniter(fvals)
for fval in fvals:
if fval < 0 or fval > 1:
raise ValueError("f must be between 0 and 1")
svals = listify_noniter(svals)
for sval in svals:
if sval < 2:
raise ValueError("s must be greater than or equal to 2")
kvals = listify_noniter(kvals)
for kval in kvals:
if kval < -10 or kval > 10:
raise ValueError("k must be between -10 and 10")
Nvals = listify_noniter(Nvals)
for Nval in Nvals:
if Nval is not None and Nval < 1:
raise ValueError("N must be a positive integer")
if NumReps is not None:
if NumReps < 10:
raise ValueError("NumReps must be greater than or equal to 10")
itercount = [1 for vals in (alphavals, fvals, svals, kvals, Nvals)
if len(vals) > 1]
if NumReps is not None:
itercount.append(1) # NumReps is always an iterator
if len(itercount) > 2:
raise ValueError("only 2 vars may have multiple values")
# k gets reversed so the legend appears in the right order
allvars = product(alphavals, fvals, svals, reversed(kvals), Nvals,
(None,) if NumReps is None else range(NumReps))
progress.total = np.prod([len(val) for val in (alphavals, fvals, svals,
kvals, Nvals)] + [1 if NumReps is None else NumReps, ])
stats = {}
for alpha, f, s, k, N, RepNum in allvars:
#print(f"alpha={alpha}, f={f}, s={s}, k={k}, N={N}, "
# f"RepNum={RepNum}")
stats.setdefault(ALPHA, []).append(alpha)
stats.setdefault('f', []).append(f)
stats.setdefault('s', []).append(s)
stats.setdefault('k', []).append(k)
ft, fts = (2, 2)
if RepNum is not None and N is not None:
progress.message = (f"Calculating: α={alpha}, f={f}, "
f"s={s}, k={k}, N={N}, RepNum={RepNum}")
stats.setdefault("N", []).append(N)
stats.setdefault("RepNum", []).append(RepNum)
while (ft >= 2 and fts >= 2):
ft, fts = simNucPedigree(N, s, alpha, f, k)
else:
progress.message = (f"Calculating: alpha={alpha}, f={f}, "
f"s={s}, k={k}")
ft, fts = exactNucPedigree(s, alpha, f, k)
stats.setdefault(FT, []).append(ft)
stats.setdefault(FTS, []).append(fts)
progress.tick()
results = pd.DataFrame(stats)
if to_stdout:
with DATA_PRINT_OPTIONS:
print(results)
return results
def output_summarystats(df, filename):
"""Given one of our figure dataframes and a filename, outputs our summary
stats info to that file.
"""
# The "summary stats" for "as a function of alpha" (fig2) pretty much are a
# specialized form of the raw output. For the estimate distributions based
# on number of families, though, we have actual summarization. So we need
# to figure out which is which.
try:
df['N']
except KeyError:
# "as a function of alpha" - no summarization, just formatting
ftdf, ftsdf = df.pivot('k', ALPHA, FT), df.pivot('k', ALPHA, FTS)
with open(filename, "w") as outfile:
with DATA_PRINT_OPTIONS:
outfile.write("\n\n".join((
# lack of commas for first 3 lines is deliberate
"==================================\n"
"Figure2. Statistics\n"
"==================================\n"
f"f={df['f'][0]}, s={df['s'][0]}",
f"f_tilde\n{df.pivot('k', ALPHA, FT)}",
f"f_tilde_star\n{df.pivot('k', ALPHA, FTS)}"
)))
else:
# distribution plot - we need to actually summarize :)
base_ft = df.pivot('RepNum', 'N', FT)
base_fts = df.pivot('RepNum', 'N', FTS)
ft_stats = pd.DataFrame({
"mean": base_ft.mean(0),
"std": base_ft.std(0),
"med": base_ft.median(0)
}).transpose()
fts_stats = pd.DataFrame({
"mean": base_fts.mean(0),
"std": base_fts.std(0),
"med": base_fts.median(0)
}).transpose()
with open(filename, "w") as outfile:
with DATA_PRINT_OPTIONS:
outfile.write("\n\n".join((
# lack of commas for first 5 lines is deliberate
"==================================\n"
"Figure1. Statistics\n"
"==================================\n"
f"alpha={df[ALPHA][0]}, f={df['f'][0]}, "
f"s={df['s'][0]}, k={df['k'][0]}, "
f"NumReps={df['RepNum'].iat[-1] + 1}",
f"f_tilde\n{ft_stats}",
f"f_tilde_star\n{fts_stats}"
)))
def generate_and_display_test_figure(fig1df, fig2df):
"""Given two dataframes - one with our N values and our two sets of Y-axis
values (ftilde, ftildestar), and the other with our alpha and k
values and our two sets of Y-axis values, plots the first dataset's values
by N in sinaplots and the second dataset's values by alpha in lineplots,
and display the results.
"""
plt.figure(1)
plt.subplot(2, 2, 1)
sinaplot(x="N", y=FT, data=fig1df, edgecolor="black", alpha=.5,
violin=False)
plt.ylabel(FT, rotation=0)
plt.ylim([0,1])
plt.axhline(fig1df['f'][0], alpha=0.5, dashes=(5,2))
plt.subplot(2, 2, 2)
sinaplot(x="N", y=FTS, data=fig1df, edgecolor="black", alpha=.5,
violin=False)
plt.ylabel(FTS, rotation=0)
plt.ylim([0,1])
plt.axhline(fig1df['f'][0], alpha=0.5, dashes=(5,2))
plt.subplot(2, 2, 3)
sns.lineplot(data=fig2df.pivot(ALPHA, "k", FT), markers=True)
plt.ylabel(FT, rotation=0)
plt.ylim([0,1])
plt.axhline(fig2df['f'][0], alpha=0.5, dashes=(5,2))
plt.subplot(2, 2, 4)
sns.lineplot(data=fig2df.pivot(ALPHA, "k", FTS), markers=True)
plt.ylabel(FTS, rotation=0)
plt.ylim([0,1])
plt.axhline(fig2df['f'][0], alpha=0.5, dashes=(5,2))
plt.tight_layout()
plt.show()
if __name__ == "__main__":
# Default "test" input values
# Figure 1 which goes to the 1st row
alpha = 0.0
f = 0.5
s = 2
k = 1
N = [1, 2, 4, 6, 8, 10, 30, 50]
NumReps = 100
fig1df = calc_stats(alpha, f, s, k, N, NumReps)
# Figure 2 which goes to the 2nd row.
alpha = [0.001, 0.01, 0.1, 0.2, 0.3, 0.4, 0.5]
# now up to 0.5 and these values are fixed. 5/30/2022
f = 0.6
s = 3
k = [-1, 0, 1, 2]
fig2df = calc_stats(alpha, f, s, k)
generate_and_display_test_figure(fig1df, fig2df)