-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathGASeqDesign.py
executable file
·193 lines (166 loc) · 8.92 KB
/
GASeqDesign.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
#!/usr/bin/env python
import warnings
import MYUTILS
import sys
import argparse
parser = argparse.ArgumentParser(description='Takes as input one or more lists of ports of servers, weights for training, and parameters for optimization and outputs a collection of sequences and outputs.')
parser.add_argument('-ids',dest='ids', metavar='<ids>',help='Model IDs, comma separated', required=True);
parser.add_argument('-p',dest='ports', metavar='<ports>',help='Model ports, comma separated, same order as -ids', required=True);
parser.add_argument('-hosts',dest='hosts', metavar='<hosts>',help='Model hosts, comma separated, same order as -ids - can also be one entry for all [default="127.0.0.1"]', default = "127.0.0.1", required=False);
parser.add_argument('-w',dest='weights', metavar='<weights>',help='Model weights, comma separated, same order as -ids', required=True);
parser.add_argument('-b',dest='batches', metavar='<batches>',help='Batches of sequences', required=True);
parser.add_argument('-nb',dest='seqsPerBatch', metavar='<seqsPerBatch>',help='Sequences taken per batch', required=True);
parser.add_argument('-pop',dest='popSize', metavar='<popSize>',help='Population size [default=100]',default="100", required=False);
parser.add_argument('-gen',dest='gen', metavar='<generations>',help='Number of generations [default=100]',default="100", required=False);
parser.add_argument('-mr',dest='mutationRate', metavar='<mutationRate>',help='Mutation rate [default=0.02]',default="0.02", required=False);
parser.add_argument('-rf',dest='recombFreq', metavar='<recombFreq>',help='Recombination rate [default=0.1]',default="0.1", required=False);
parser.add_argument('-ws',dest='weightedSum', action='count',help='Use a weighted sum for fitness rather than NSGA-II (multi-objective)?', required=False, default=0);
parser.add_argument('-o',dest='outFP', metavar='<outFile>',help='Where to output results [default=stdout]', required=False);
parser.add_argument('-l',dest='logFP', metavar='<logFile>',help='Where to output errors/warnings [default=stderr]', required=False);
parser.add_argument('-v',dest='verbose', action='count',help='Verbose output?', required=False, default=0);
#args = lambda: None; args.inFP = "thisScript.py -i ...".split() # debug settings
args = parser.parse_args();
import numpy as np;
from deap import creator, base, tools, algorithms
from tqdm import tqdm
import socket
import json
args.ports = [int(x) for x in args.ports.split(",")]
args.ids = args.ids.split(",")
args.hosts = args.hosts.split(",")
args.weights = [float(x) for x in args.weights.split(",")]
assert len(args.ids)==len(args.weights), "Must be same number of IDs as weights"
assert len(args.ids)==len(args.ports), "Must be same number of IDs as ports"
assert len(args.ids)==len(args.weights), "Must be same number of IDs as weights"
assert len(args.ids)==len(args.hosts) or len(args.hosts)==1, "Must be same number of IDs as hosts, or only one host"
if (args.logFP is not None):
logFile=MYUTILS.smartGZOpen(args.logFP,'w');
sys.stderr=logFile;
if (args.outFP is None):
outFile= sys.stdout;
else:
if args.verbose>0: sys.stderr.write("Outputting to file "+args.outFP+"\n");
outFile = MYUTILS.smartGZOpen(args.outFP,'w');
argsGA = {'sequence_length' : 80 , 'nucleotide_frequency' :[0.25,0.25,0.25,0.25] }
LHS = list("TGCATTTTTTTCACATC")
RHS = list("GGTTACGGCTGTT");
MAXBYTES=1000000
args.batches = int(args.batches);
args.mutationRate = float(args.mutationRate);
args.recombFreq = float(args.recombFreq);
args.popSize = int(args.popSize);
args.seqsPerBatch = int(args.seqsPerBatch);
args.gen = int(args.gen);
randomizer=np.random
def random_sequence_generator(randomizer,args) :
return randomizer.choice(list('ACGT') , p=args['nucleotide_frequency'] )
def mutation(individual, indpb):
for i in xrange(len(individual)):
if np.random.random() < indpb:
if individual[i]=='A' :
individual[i] = (randomizer.choice(list('CGT') , p=[ argsGA['nucleotide_frequency'][1]/(1- argsGA['nucleotide_frequency'][0]) , argsGA['nucleotide_frequency'][2]/(1- argsGA['nucleotide_frequency'][0]) , argsGA['nucleotide_frequency'][3]/(1- argsGA['nucleotide_frequency'][0]) ] ) )
elif individual[i]=='C' :
individual[i] = (randomizer.choice(list('AGT') , p=[ argsGA['nucleotide_frequency'][0]/(1- argsGA['nucleotide_frequency'][1]) , argsGA['nucleotide_frequency'][2]/(1- argsGA['nucleotide_frequency'][1]) , argsGA['nucleotide_frequency'][3]/(1- argsGA['nucleotide_frequency'][1]) ] ) )
elif individual[i]=='G' :
individual[i] = (randomizer.choice(list('CGT') , p=[ argsGA['nucleotide_frequency'][2]/(1- argsGA['nucleotide_frequency'][2]) , argsGA['nucleotide_frequency'][1]/(1- argsGA['nucleotide_frequency'][2]) , argsGA['nucleotide_frequency'][3]/(1- argsGA['nucleotide_frequency'][2]) ] ) )
elif individual[i]=='T' :
individual[i] = (randomizer.choice(list('CGT') , p=[ argsGA['nucleotide_frequency'][0]/(1- argsGA['nucleotide_frequency'][3]) , argsGA['nucleotide_frequency'][1]/(1- argsGA['nucleotide_frequency'][3]) , argsGA['nucleotide_frequency'][2]/(1- argsGA['nucleotide_frequency'][3]) ] ) )
return individual,
def getPrediction(sequences, socketInfo):
global MAXBYTES
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.connect(socketInfo)
s.send(json.dumps(sequences))
predStr = s.recv(MAXBYTES);
nerr=0
while True:
try:
preds = json.loads(predStr)
break;
except ValueError as e:
nerr=nerr+1
if nerr>10:
print predStr;
raise e
predStr = predStr + s.recv(MAXBYTES).strip()
s.close();
return preds;
def fitness(sequences):
global socketInfo, RHS, LHS
seqsFull= []
for ind in range(len(sequences)) :
seqsFull.append(LHS + sequences[ind] + RHS )
allPreds = [[] for x in range(0,len(seqsFull))]
for i in range(0,len(socketInfo)):
curPreds = getPrediction(seqsFull, socketInfo[i])
for j in range(0,len(seqsFull)):
allPreds[j].append(curPreds[j]);
for j in range(0,len(seqsFull)):
allPreds[j] = tuple(allPreds[j])
return(allPreds)
def weightedSum(sequences):
preds = fitness(sequences);
global args
preds2 = [];
for i in range(0, len(preds)):
preds2.append((sum(np.array(preds[i])*np.array(args.weights)),))
return preds2;
if args.weightedSum>0:
creator.create("FitnessMax", base.Fitness, weights=((1.0,)))
else:
creator.create("FitnessMax", base.Fitness, weights=(tuple(args.weights)))
#creator.create("FitnessMax", base.Fitness, weights=(0,))
creator.create("Individual", list , fitness=creator.FitnessMax)
toolbox = base.Toolbox()
toolbox.register("base", random_sequence_generator , randomizer , argsGA)
toolbox.register("individual", tools.initRepeat, creator.Individual, toolbox.base, n=argsGA['sequence_length'])
toolbox.register("population", tools.initRepeat, list, toolbox.individual)
if args.weightedSum>0:
toolbox.register("evaluate", weightedSum)
else:
toolbox.register("evaluate", fitness)
toolbox.register("mate", tools.cxTwoPoint)
toolbox.register("mutate", mutation, indpb=args.mutationRate)
if len(args.weights)==1 or args.weightedSum>0:
toolbox.register("select", tools.selTournament, tournsize=3)
else:
toolbox.register("select", tools.selNSGA2)
outFile.write("seq\tbatch\t%s\n"%("\t".join(args.ids)));
if len(args.hosts)< len(args.ports):
args.hosts = args.hosts*len(args.ports);
socketInfo = zip(args.hosts,args.ports)
args.weights = np.array(args.weights)
if args.verbose>1: sys.stderr.write("Running optimization\n")
for b in tqdm(range(args.batches)):
population = toolbox.population(n=args.popSize)
fits = toolbox.evaluate(population)
for fit, ind in zip(fits, population): # below is where Fitness.setValues is actually run
ind.fitness.setValues(fit)# somehow this sets ind.fitness to the unweighted values,
if args.verbose>1: sys.stderr.write("batch=%i\n"%(b))
for gen in range(args.gen):
if args.verbose>1: sys.stderr.write("\tgen=%i\n"%(gen))
#make offspring from survivors of last round
offspring = algorithms.varAnd(population, toolbox, cxpb=args.recombFreq, mutpb=1.0)
#print "".join(offspring[0])
#calculate fitness of offspring
fits = toolbox.evaluate(offspring)
if args.verbose>1:
if args.weightedSum>0:
sys.stderr.write("\tBest fitness: %g\n"%(max([f for f in fits])));
else:
sys.stderr.write("\tBest fitness: %g\n"%(max([np.sum(np.array(f)*args.weights) for f in fits])));
for fit, ind in zip(fits, offspring): # below is where Fitness.setValues is actually run
ind.fitness.setValues(fit)# somehow this sets ind.fitness to the unweighted values,
population = toolbox.select(population+offspring, k=len(population))
if len(args.weights)==1 or args.weightedSum>0:
top = tools.selBest(population, k=args.seqsPerBatch)
else:
top = tools.selNSGA2(population, k=args.seqsPerBatch)
topFitnesses = fitness(top)
top = ["".join(x) for x in top]
for i in range(0, len(top)):
outFile.write("%s\t%i\t%s\n"%(top[i], b, "\t".join([str(x) for x in topFitnesses[i]])))
#raise Exception("Reached bad state=%d for '%s.%d' '%s' at line '%s'" %(state,mid,ver,tfid,line));
outFile.close();
if (args.logFP is not None):
logFile.close();