upstream-extended/community/tools/rpy_statistics_collection/partialR_square.py
1#!/usr/bin/env python
2
3# from galaxy import eggs
4
5import sys
6
7import rpy2.rlike.container as rlc
8import rpy2.robjects as robjects
9
10# from rpy import *
11
12
13r = robjects.r
14
15# export PYTHONPATH=~/galaxy/lib/
16# running command python partialR_square.py reg_inp.tab 4 1,2,3 partialR_result.tabular
17
18
19def stop_err(msg):
20 sys.stderr.write(msg)
21 sys.exit()
22
23
24def sscombs(s):
25 if len(s) == 1:
26 return [s]
27 else:
28 ssc = sscombs(s[1:])
29 return [s[0]] + [s[0] + comb for comb in ssc] + ssc
30
31
32infile = sys.argv[1]
33y_col = int(sys.argv[2]) - 1
34x_cols = sys.argv[3].split(",")
35outfile = sys.argv[4]
36
37print("Predictor columns: %s; Response column: %d" % (x_cols, y_col + 1))
38fout = open(outfile, "w")
39
40for i, line in enumerate(file(infile)): # noqa F821
41 line = line.rstrip("\r\n")
42 if len(line) > 0 and not line.startswith("#"):
43 elems = line.split("\t")
44 break
45 if i == 30:
46 break # Hopefully we'll never get here...
47
48if len(elems) < 1:
49 stop_err(
50 "The data in your input dataset is either missing or not formatted properly."
51 )
52
53y_vals = []
54x_vals = []
55x_vector = []
56for k, col in enumerate(x_cols):
57 x_cols[k] = int(col) - 1
58 x_vals.append([])
59 """
60 try:
61 float( elems[x_cols[k]] )
62 except Exception:
63 try:
64 msg = "This operation cannot be performed on non-numeric column %d containing value '%s'." %( col, elems[x_cols[k]] )
65 except Exception:
66 msg = "This operation cannot be performed on non-numeric data."
67 stop_err( msg )
68 """
69NA = "NA"
70for ind, line in enumerate(file(infile)): # noqa F821
71 if line and not line.startswith("#"):
72 try:
73 fields = line.split("\t")
74 try:
75 yval = float(fields[y_col])
76 except Exception:
77 yval = r("NA")
78 # print >>sys.stderr, "ey = %s" %ey
79 y_vals.append(yval)
80 for k, col in enumerate(x_cols):
81 try:
82 xval = float(fields[col])
83 except Exception:
84 xval = r("NA")
85 # print >>sys.stderr, "ex = %s" %ex
86 x_vals[k].append(xval)
87 x_vector.append(xval)
88 except Exception:
89 pass
90
91# x_vals1 = numpy.asarray(x_vals).transpose()
92# dat= r.list(x=array(x_vals1), y=y_vals)
93
94# set_default_mode(NO_CONVERSION)
95# try:
96# full = r.lm(r("y ~ x"), data= r.na_exclude(dat)) #full model includes all the predictor variables specified by the user
97# except Exception as rex:
98# stop_err("Error performing linear regression on the input data.\nEither the response column or one of the predictor columns contain no numeric values.")
99# set_default_mode(BASIC_CONVERSION)
100
101fv = robjects.FloatVector(x_vector)
102m = r["matrix"](fv, ncol=len(x_cols), byrow=True)
103# ensure order for generating formula
104od = rlc.OrdDict([("y", robjects.FloatVector(y_vals)), ("x", m)])
105dat = robjects.DataFrame(od)
106# convert dat.names: ["y","x.1","x.2"] to formula string: 'y ~ x.1 + x.2'
107formula = " + ".join(dat.names).replace("+", "~", 1)
108try:
109 full = r.lm(formula, data=r["na.exclude"](dat))
110except Exception:
111 stop_err(
112 "Error performing linear regression on the input data.\nEither the response column or one of the predictor columns contain only non-numeric or invalid values."
113 )
114
115
116summary = r.summary(full)
117# fullr2 = summary.get('r.squared','NA')
118fullr2 = summary.rx2("r.squared")[0]
119
120if fullr2 == "NA":
121 stop_err("Error in linear regression")
122
123if len(x_vals) < 10:
124 s = ""
125 for ch in range(len(x_vals)):
126 s += str(ch)
127else:
128 stop_err("This tool only works with less than 10 predictors.")
129
130print("#Model\tR-sq\tpartial_R_Terms\tpartial_R_Value", file=fout)
131all_combos = sorted(sscombs(s), key=len)
132all_combos.reverse()
133for j, cols in enumerate(all_combos):
134 # if len(cols) == len(s): #Same as the full model above
135 # continue
136 if len(cols) == 1:
137 # x_vals1 = x_vals[int(cols)]
138 x_v = x_vals[int(cols)]
139 else:
140 x_v = []
141 for col in cols:
142 # x_v.append(x_vals[int(col)])
143 x_v.extend(x_vals[int(col)])
144 # x_vals1 = numpy.asarray(x_v).transpose()
145 # dat= r.list(x=array(x_vals1), y=y_vals)
146 # set_default_mode(NO_CONVERSION)
147 # red = r.lm(r("y ~ x"), data= dat) #Reduced model
148 # set_default_mode(BASIC_CONVERSION)
149 fv = robjects.FloatVector(x_v)
150 m = r["matrix"](fv, ncol=len(cols), byrow=False)
151 # ensure order for generating formula
152 od = rlc.OrdDict([("y", robjects.FloatVector(y_vals)), ("x", m)])
153 dat = robjects.DataFrame(od)
154 # convert dat.names: ["y","x.1","x.2"] to formula string: 'y ~ x.1 + x.2'
155 formula = " + ".join(dat.names).replace("+", "~", 1)
156 try:
157 red = r.lm(formula, data=r["na.exclude"](dat))
158 except Exception:
159 stop_err(
160 "Error performing linear regression on the input data.\nEither the response column or one of the predictor columns contain only non-numeric or invalid values."
161 )
162
163 summary = r.summary(red)
164 # redr2 = summary.get('r.squared','NA')
165 redr2 = summary.rx2("r.squared")[0]
166
167 try:
168 partial_R = (float(fullr2) - float(redr2)) / (1 - float(redr2))
169 except Exception:
170 partial_R = "NA"
171 col_str = ""
172 for col in cols:
173 col_str = col_str + str(int(x_cols[int(col)]) + 1) + " "
174 col_str.strip()
175 partial_R_col_str = ""
176 for col in s:
177 if col not in cols:
178 partial_R_col_str = partial_R_col_str + str(int(x_cols[int(col)]) + 1) + " "
179 partial_R_col_str.strip()
180 if len(cols) == len(s): # full model
181 partial_R_col_str = "-"
182 partial_R = "-"
183 try:
184 redr2 = "%.4f" % (float(redr2))
185 except Exception:
186 pass
187 try:
188 partial_R = "%.4f" % (float(partial_R))
189 except Exception:
190 pass
191 print("%s\t%s\t%s\t%s" % (col_str, redr2, partial_R_col_str, partial_R), file=fout)