'''
r_table2scatter.py - R based plots and stats
============================================
:Author: Andreas Heger
:Release: $Id$
:Date: |today|
:Tags: Python
Purpose
-------
This script reads a table from a file or stdin.
It can compute various stats (correlations, ...)
and/or plot the data using R.
Usage
-----
Example::
python r_table2scatter.py --help
Type::
python r_table2scatter.py --help
for command line help.
Command line options
--------------------
'''
import os
import sys
import re
import tempfile
import math
import code
import CGAT.Experiment as E
import CGAT.MatlabTools as MatlabTools
import numpy
import CGAT.Stats as Stats
from rpy2.robjects import r as R
def readTable(lines,
assign,
separator="\t",
numeric_type=numpy.float,
take_columns="all",
headers=True,
row_names=None,
colours=None,
):
"""read a matrix. There probably is a routine for this in Numpy, which
I haven't found yet.
The advantage of using rpy's mechanism is to allow for
missing values.
row_names: column with row names
"""
# now import R as stdin has been read.
take = take_columns
handle, name = tempfile.mkstemp()
os.close(handle)
if take_columns == "all":
num_cols = len(lines[0][:-1].split("\t"))
take = list(range(0, num_cols))
elif take_columns == "all-but-first":
num_cols = len(lines[0][:-1].split("\t"))
take = list(range(1, num_cols))
outfile = IOTools.openFile(name, "w")
c = []
# delete columns with colour information/legend
to_delete = []
if row_names is not None:
legend = []
if take_columns == "all":
to_delete.append(row_names)
else:
legend = None
if colours is not None:
if take_columns == "all":
to_delete.append(colours)
to_delete.sort()
to_delete.reverse()
for x in to_delete:
del take[x]
# get column headers
if headers:
headers = lines[0][:-1].split("\t")
headers = [headers[x] for x in take]
del lines[0]
for l in lines:
data = [x.strip() for x in l[:-1].split("\t")]
if not data or not [x for x in data if x != ""]:
continue
outfile.write("\t".join([data[x] for x in take]) + "\n")
if row_names is not None:
legend.append(data[row_names])
if colours is not None:
c.append(data[colours])
outfile.close()
# rpy.set_default_mode(rpy.NO_CONVERSION)
# note that the conversion
# is not perfect. Some missing values are assigned to "nan", while
# some are -2147483648. They seem to treated correctly, though,
# within R, but note that when computing something like sum(), the
# result in python after conversion might be -2147483648.
if headers:
matrix = R("""%s <- read.table( '%s', na.string = c("NA", "na", 'nan', 'NaN'), col.names=c('%s'), sep="\t" )""" %
(assign, name, "','".join(headers)))
else:
matrix = R(
"""%s <- read.table( '%s', na.string = c("NA", "na", 'nan', 'NaN'), col.names=headers, sep="\t" )""" % (assign, name))
# rpy.set_default_mode(rpy.BASIC_CONVERSION)
os.remove(name)
return matrix, headers, c, legend
def writeMatrix(file,
matrix,
separator="\t",
headers=[],
format="%f"):
"""write a matrix to file.
if headers are given, add them to columns and rows.
"""
if headers:
file.write("\t" + "\t".join(headers) + "\n")
nrows, ncols = matrix.shape
for x in range(nrows):
if headers:
file.write(headers[x] + "\t")
file.write(
"\t".join([format % x for x in matrix[x]]) + "\n")
def FuncScatterDiagonal(data):
R.points(data)
R.abline(0, 1)
def main(argv=None):
if argv is None:
argv = sys.argv
parser = E.OptionParser(
version="%prog version: $Id: r_table2scatter.py 2782 2009-09-10 11:40:29Z andreas $")
parser.add_option("-c", "--columns", dest="columns", type="string",
help="columns to take from table. Choices are 'all', 'all-but-first' or a ','-separated list of columns.")
parser.add_option("--logscale", dest="logscale", type="string",
help="log-transform one or both axes [default=%Default].")
parser.add_option("-a", "--hardcopy", dest="hardcopy", type="string",
help="write hardcopy to file [default=%default].",
metavar="FILE")
parser.add_option("-f", "--file", dest="input_filename", type="string",
help="filename with table data [default=%default].",
metavar="FILE")
parser.add_option("-2", "--file2", dest="input_filename2", type="string",
help="additional data file [default=%default].",
metavar="FILE")
parser.add_option("-s", "--stats", dest="statistics", type="choice",
choices=("correlation", "spearman", "pearson", "count"),
help="statistical quantities to compute [default=%default]",
action="append")
parser.add_option("-p", "--plot", dest="plot", type="choice",
choices=("scatter", "pairs", "panel", "bar", "bar-stacked",
"bar-besides", "1_vs_x", "matched", "boxplot", "scatter+marginal",
"scatter-regression"),
help="plots to plot [default=%default]",
action="append")
parser.add_option("-t", "--threshold", dest="threshold", type="float",
help="min threshold to use for counting method [default=%default].")
parser.add_option("-o", "--colours", dest="colours", type="int",
help="column with colour information [default=%default].")
parser.add_option("-l", "--plot-labels", dest="labels", type="string",
help="column labels for x and y in matched plots [default=%default].")
parser.add_option("-d", "--add-diagonal", dest="add_diagonal", action="store_true",
help="add diagonal to plot [default=%default].")
parser.add_option("-e", "--plot-legend", dest="legend", type="int",
help="column with legend [default=%default].")
parser.add_option("-r", "--options", dest="r_options", type="string",
help="R plotting options [default=%default].")
parser.add_option("--format", dest="format", type="choice",
choices=("full", "sparse"),
help="output format [default=%default].")
parser.add_option("--title", dest="title", type="string",
help="""plot title [default=%default].""")
parser.add_option("", "--xrange", dest="xrange", type="string",
help="x viewing range of plot [default=%default].")
parser.add_option("", "--yrange", dest="yrange", type="string",
help="y viewing range of plot[default=%default].")
parser.add_option("--allow-empty-file", dest="fail_on_empty", action="store_false",
help="do not fail on empty input [default=%default].")
parser.add_option("--fail-on-empty", dest="fail_on_empty", action="store_true",
help="fail on empty input [default=%default].")
parser.set_defaults(
hardcopy=None,
input_filename="",
input_filename2=None,
columns="all",
logscale=None,
statistics=[],
plot=[],
threshold=0.0,
labels="x,y",
colours=None,
diagonal=False,
legend=None,
title=None,
xrange=None,
yrange=None,
r_options="",
fail_on_empty=True,
format="full")
(options, args) = E.Start(parser)
if len(args) == 1 and not options.input_filename:
options.input_filename = args[0]
if options.columns not in ("all", "all-but-first"):
options.columns = [int(x) - 1 for x in options.columns.split(",")]
if options.colours:
options.colours -= 1
if options.legend:
options.legend -= 1
table = {}
headers = []
# read data matrix
if options.input_filename:
lines = IOTools.openFile(options.input_filename, "r").readlines()
else:
# note: this will not work for interactive viewing, but
# creating hardcopy plots works.
lines = sys.stdin.readlines()
lines = [x for x in lines if x[0] != "#"]
if len(lines) == 0:
if options.fail_on_empty:
raise IOError("no input")
E.warn("empty input")
E.Stop()
return
matrix, headers, colours, legend = readTable(lines,
"matrix",
take_columns=options.columns,
headers=True,
colours=options.colours,
row_names=options.legend)
if options.input_filename2:
# read another matrix (should be of the same format.
matrix2, headers2, colours2, legend2 = readTable(
lines,
"matrix2",
take_columns=options.columns,
headers=True,
colours=options.colours,
row_names=options.legend)
R.assign("headers", headers)
ndata = R("""length( matrix[,1] )""")[0]
if options.loglevel >= 1:
options.stdlog.write("# read matrix: %ix%i\n" % (len(headers), ndata))
if colours:
R.assign("colours", colours)
for method in options.statistics:
if method == "correlation":
cor = R.cor(matrix, use="pairwise.complete.obs")
writeMatrix(sys.stdout, cor, headers=headers, format="%5.2f")
elif method == "pearson":
options.stdout.write("\t".join(("var1",
"var2",
"coeff",
"passed",
"pvalue",
"n",
"method",
"alternative")) + "\n")
for x in range(len(headers) - 1):
for y in range(x + 1, len(headers)):
try:
result = R(
"""cor.test( matrix[,%i], matrix[,%i] )""" % (x + 1, y + 1))
except rpy.RPyException as msg:
E.warn("correlation not computed for columns %i(%s) and %i(%s): %s" % (
x, headers[x], y, headers[y], msg))
options.stdout.write("%s\t%s\t%s\t%s\t%s\t%i\t%s\t%s\n" %
(headers[x], headers[y],
"na",
"na",
"na",
0,
"na",
"na"))
else:
options.stdout.write(
"%s\t%s\t%6.4f\t%s\t%e\t%i\t%s\t%s\n" %
(headers[x], headers[y],
result.rx2('estimate').rx2(
'cor')[0],
Stats.getSignificance(
float(result.rx2('p.value')[0])),
result.rx2('p.value')[0],
result.rx2('parameter').rx2(
'df')[0],
result.rx2('method')[0],
result.rx2('alternative')[0]))
elif method == "spearman":
options.stdout.write("\t".join(("var1", "var2",
"coeff",
"passed",
"pvalue",
"method",
"alternative")) + "\n")
for x in range(len(headers) - 1):
for y in range(x + 1, len(headers)):
result = R(
"""cor.test( matrix[,%i], matrix[,%i], method='spearman')""" % (x + 1, y + 1))
options.stdout.write(
"%s\t%s\t%6.4f\t%s\t%e\t%i\t%s\t%s\n" %
(headers[x], headers[y],
result['estimate']['rho'],
Stats.getSignificance(float(result['p.value'])),
result['p.value'],
result['parameter']['df'],
result['method'],
result['alternative']))
elif method == "count":
# number of shared elements > threshold
m, r, c = MatlabTools.ReadMatrix(open(options.input_filename, "r"),
take=options.columns,
headers=True)
mask = numpy.greater(m, options.threshold)
counts = numpy.dot(numpy.transpose(mask), mask)
writeMatrix(options.stdout, counts, headers=c, format="%i")
if options.plot:
# remove columns that are completely empty
if "pairs" in options.plot:
colsums = R('''colSums( is.na(matrix ))''')
take = [x for x in range(len(colsums)) if colsums[x] != ndata]
if take:
E.warn("removing empty columns %s before plotting" % str(take))
matrix = R.subset(matrix, select=[x + 1 for x in take])
R.assign("""matrix""", matrix)
headers = [headers[x] for x in take]
if legend:
legend = [headers[x] for x in take]
if options.r_options:
extra_options = ", %s" % options.r_options
else:
extra_options = ""
if options.legend is not None and len(legend):
extra_options += ", legend=c('%s')" % "','".join(legend)
if options.labels:
xlabel, ylabel = options.labels.split(",")
extra_options += ", xlab='%s', ylab='%s'" % (xlabel, ylabel)
else:
xlabel, ylabel = "", ""
if options.colours:
extra_options += ", col=colours"
if options.logscale:
extra_options += ", log='%s'" % options.logscale
if options.xrange:
extra_options += ", xlim=c(%f,%f)" % tuple(
map(float, options.xrange.split(",")))
if options.yrange:
extra_options += ", ylim=c(%f,%f)" % tuple(
map(float, options.yrange.split(",")))
if options.hardcopy:
if options.hardcopy.endswith(".eps"):
R.postscript(options.hardcopy)
elif options.hardcopy.endswith(".png"):
R.png(options.hardcopy, width=1024, height=768, type="cairo")
elif options.hardcopy.endswith(".jpg"):
R.jpg(options.hardcopy, width=1024, height=768, type="cairo")
for method in options.plot:
if ndata < 100:
point_size = "1"
pch = "o"
elif ndata < 1000:
point_size = "1"
pch = "o"
else:
point_size = "0.5"
pch = "."
if method == "scatter":
R("""plot( matrix[,1], matrix[,2], cex=%s, pch="o" %s)""" % (
point_size, extra_options))
if method == "scatter-regression":
R("""plot( matrix[,1], matrix[,2], cex=%s, pch="o" %s)""" % (
point_size, extra_options))
dat = R(
"""dat <- data.frame(x = matrix[,1], y = matrix[,2])""")
R(
"""new <- data.frame(x = seq( min(matrix[,1]), max(matrix[,1]), (max(matrix[,1]) - min(matrix[,1])) / 100))""")
mod = R("""mod <- lm( y ~ x, dat)""")
R("""predict(mod, new, se.fit = TRUE)""")
R("""pred.w.plim <- predict(mod, new, interval="prediction")""")
R("""pred.w.clim <- predict(mod, new, interval="confidence")""")
R(
"""matpoints(new$x,cbind(pred.w.clim, pred.w.plim[,-1]), lty=c(1,2,2,3,3), type="l")""")
R.mtext(
"y = %f * x + %f, r=%6.4f, n=%i" % (mod["coefficients"]["x"],
mod["coefficients"][
"(Intercept)"],
R("""cor( dat )[2]"""),
ndata),
3,
cex=1.0)
elif method == "pairs":
if options.add_diagonal:
R(
"""panel.hist <- function( x,y,... ) { points(x,y,...); abline(0,1); }""")
else:
R(
"""panel.hist <- function( x,y,... ) { points(x,y,...); }""")
# There used to be a argument na_action="na.omit", but
# removed this as there appeared error messages saying
# "na.action is not a graphical parameter" and the
# plots showed occasionally the wrong scale.
# cex=point_size also caused trouble (error message:
# "X11 used font size 8 when 2 was requested" or
# similar)
if options.colours:
R.pairs(matrix,
pch=pch,
col=colours,
main=options.title,
panel="panel.hist",
labels=headers,
cex_labels=2.0)
else:
R.pairs(matrix,
pch=pch,
panel="panel.hist",
main=options.title,
labels=headers,
cex_labels=2.0)
elif method == "boxplot":
extra_options += ",main='%s'" % options.title
# set vertical orientation
if max([len(x) for x in headers]) > 40 / len(headers):
# remove xlabel:
extra_options = re.sub(", xlab='[^']+'", "", extra_options)
extra_options += ", names.arg=headers, las=2"
R(
"""op <- par(mar=c(11,4,4,2))""") # the 10 allows the names.arg below the barplot
R("""boxplot( matrix %s)""" % extra_options)
elif method == "bar" or method == "bar-stacked":
if not options.colours:
extra_options += ", col=rainbow(5)"
# set vertical orientation
if max([len(x) for x in headers]) > 40 / len(headers):
# remove xlabel:
extra_options = re.sub(", xlab='[^']+'", "", extra_options)
extra_options += ", names.arg=headers, las=2"
R(
"""op <- par(mar=c(11,4,4,2))""") # the 10 allows the names.arg below the barplot
R("""barplot(as.matrix(matrix), %s)""" % extra_options)
elif method == "bar-besides":
if not options.colours:
extra_options += ", col=rainbow(%i)" % ndata
# set vertical orientation
if max([len(x) for x in headers]) > 40 / len(headers):
# remove xlabel:
extra_options = re.sub(", xlab='[^']+'", "", extra_options)
extra_options += ", names.arg=headers, las=2"
R(
"""op <- par(mar=c(11,4,4,2))""") # the 10 allows the names.arg below the barplot
R("""barplot(as.matrix(matrix), beside=TRUE %s)""" %
extra_options)
elif method == "scatter+marginal":
if options.title:
# set the size of the outer margins - the title needs to be added at the end
# after plots have been created
R.par(oma=R.c(0, 0, 4, 0))
R("""matrix""")
R("""
x <- matrix[,1];
y <- matrix[,2];
xhist <- hist(x, breaks=20, plot=FALSE);
yhist <- hist(y, breaks=20, plot=FALSE);
top <- max(c(xhist$counts, yhist$counts));
nf <- layout(matrix(c(2,0,1,3),2,2,byrow=TRUE), c(3,1), c(1,3), respect=TRUE );
par(mar=c(3,3,1,1)) ;
plot(x, y, cex=%s, pch="o" %s) ;
par(mar=c(0,3,1,1)) ;
barplot(xhist$counts, axes=FALSE, ylim=c(0, top), space=0 ) ;
par(mar=c(3,0,1,1)) ;
title(main='%s');
barplot(yhist$counts, axes=FALSE, xlim=c(0, top), space=0, horiz=TRUE ) ;
title(main='%s');
""" % (point_size, extra_options, xlabel, ylabel))
if options.title:
R.mtext(options.title, 3, outer=True, line=1, cex=1.5)
elif method in ("panel", "1_vs_x", "matched"):
if method == "panel":
pairs = []
for x in range(len(headers) - 1):
for y in range(x + 1, len(headers)):
pairs.append((x, y))
elif method == "1_vs_x":
pairs = []
for x in range(1, len(headers)):
pairs.append((0, x))
# print matching columns
elif method == "matched":
pairs = []
for x in range(len(headers) - 1):
for y in range(x + 1, len(headers)):
if headers[x] == headers[y]:
pairs.append((x, y))
break
w = int(math.ceil(math.sqrt(len(pairs))))
h = int(math.ceil(float(len(pairs)) / w))
PosInf = 1e300000
NegInf = -1e300000
xlabel, ylabel = options.labels.split(",")
R("""layout(matrix(seq(1,%i), %i, %i, byrow = TRUE))""" %
(w * h, w, h))
for a, b in pairs:
new_matrix = [x for x in zip(
list(matrix[a].values())[0],
list(matrix[b].values())[0])
if x[0] not in (float("nan"), PosInf, NegInf) and
x[1] not in (float("nan"), PosInf, NegInf)]
try:
R("""plot(matrix[,%i], matrix[,%i], main='%s versus %s', cex=0.5, pch=".", xlab='%s', ylab='%s' )""" % (
a + 1, b + 1, headers[b], headers[a], xlabel, ylabel))
except rpy.RException as msg:
print("could not plot %s versus %s: %s" % (headers[b], headers[a], msg))
if options.hardcopy:
R['dev.off']()
E.info("matrix added as >matrix< in R.")
if not options.hardcopy:
if options.input_filename:
interpreter = code.InteractiveConsole(globals())
interpreter.interact()
else:
E.info(
"can not start new interactive session as input has come from stdin.")
E.Stop()
if __name__ == "__main__":
main()
