1 #!/usr/bin/env python
2 # -*- coding: utf-8 -*-
4 from numpy import inf
5 from numpy.matlib import zeros
6 import pdb
8 # 3 * 30 supporting points: x_1 ... x_30 => 3 * 30 parameters (1 .. 90): y_1 ... y_30
9 # piecewise linear function:
10 # take score from SVM vektor (don_supp: case 1, acc_supp: case 2) and compute length of intron: case 3
11 # these are our values x
12 #
13 # | y_1 if x <= x_1
14 # |
15 # | x_i+1 - x x - x_i
16 # f(x) = | y_i * ----------- + y_i+1 * ----------- if x_i <= x <= x_i+1
17 # | x_i+1 - x_i x_i+1 - x_i
18 # |
19 # | y_30 if x_n <= x
20 #
21 # y_i and y_i+1 parameters, so the fractions are saved in the weight vectors!
23 def calculateWeights(plf, scores):
24 currentWeight = zeros((30,1))
26 for k in range(len(scores)):
27 value = scores[k]
28 Lower = len([elem for elem in plf.limits if elem <= value])
29 # because we count from 0 in python
30 Lower -= 1
31 Upper = Lower+1 ; # x-werte bleiben fest
33 print value,Lower,Upper
35 if Lower == 0:
36 currentWeight[0] += 1
37 elif Lower == len(plf.limits):
38 currentWeight[-1] += 1
39 else:
40 weightup = 1.0*(value - plf.limits[Lower]) / (plf.limits[Upper] - plf.limits[Lower])
41 weightlow = 1.0*(plf.limits[Upper] - value) / (plf.limits[Upper] - plf.limits[Lower])
42 currentWeight[Upper] = currentWeight[Upper] + weightup
43 currentWeight[Lower] = currentWeight[Lower] + weightlow
45 print plf.limits[Lower],plf.limits[Upper]
46 print weightup,weightlow,currentWeight[Upper],currentWeight[Lower]
48 return currentWeight
50 def computeSpliceWeights(d, a, h, SpliceAlign, don_supp, acc_supp):
51 ####################################################################################
52 # 1. Donor: In don_supp stehen Werte der SVM., in SpliceAlign die Einsen
53 ####################################################################################
55 # Picke die Positionen raus, an denen eine Donorstelle ist
56 DonorScores = [elem for pos,elem in enumerate(don_supp) if SpliceAlign[pos] == 1]
57 assert not ( -inf in DonorScores )
59 #print 'donor'
60 weightDon = calculateWeights(d,DonorScores)
62 ####################################################################################
63 # 2. Acceptor: In acc_supp stehen Werte der SVM., in SpliceAlign die Einsen
64 ####################################################################################
66 #Den Vektor Acceptorstellen durchgehen und die Gewichtsvektoren belasten:
67 AcceptorScores = [elem for pos,elem in enumerate(acc_supp) if pos > 0 and SpliceAlign[pos-1] == 2]
68 assert not ( -inf in AcceptorScores )
70 #print 'acceptor'
71 weightAcc = calculateWeights(a,AcceptorScores)
73 ####################################################################################
74 # 3. Intron length: SpliceAlign: Gaps zaehlen und auf Gapgewichte addieren
75 ####################################################################################
77 intron_starts = []
78 intron_ends = []
79 for pos,elem in enumerate(SpliceAlign):
80 if elem == 1:
81 intron_starts.append(pos)
83 if elem == 2:
84 intron_ends.append(pos)
86 assert len(intron_starts) == len(intron_ends)
88 for i in range(len(intron_starts)):
89 assert intron_starts[i] < intron_ends[i]
91 values = [0.0]*len(intron_starts)
92 for pos in range(len(intron_starts)):
93 values[pos] = intron_ends[pos] - intron_starts[pos] + 1
95 weightIntron = calculateWeights(h,values)
97 return weightDon, weightAcc, weightIntron