def correctness_abacus_plot(output_directory, file_prefix, df,
x_series_index, y_series_index, facet_index, peptide_index, series_color, plot_title = '', x_axis_label = '', y_axis_label = '',
fcorrect_x_cutoff = 1.0, fcorrect_y_cutoff = 1.0,
min_experimental_ddg = None,
max_experimental_ddg = None):
try:
os.mkdir(output_directory)
except:
pass
assert (os.path.exists(output_directory))
#first_peptide = df.ix[:, peptide_index].min()
#last_peptide = df.ix[:, peptide_index].max()
df['Categorization'] = df.apply(lambda r: _determine_fraction_correct_class(r[x_series_index], r[y_series_index])[0], axis = 1)
categorization_index = len(df.columns.values) - 1
df['CategorizationShape'] = df.apply(lambda r: _determine_fraction_correct_class(r[x_series_index], r[y_series_index])[1], axis = 1)
categorization_shape_index = len(df.columns.values) - 1
df['CategorizationColor'] = df.apply(lambda r: _determine_fraction_correct_class(r[x_series_index], r[y_series_index])[2], axis = 1)
categorization_color_index = len(df.columns.values) - 1
# Create the R script
boxplot_r_script = '''
library(ggplot2)
library(gridExtra)
library(scales)
library(qualV)
# PNG generation
png('%(file_prefix)s.png', width=2560, height=2048, bg="white", res=600)
txtalpha <- 0.6
redtxtalpha <- 0.6
%(png_plot_commands)s
'''
xy_table_filename = '{0}.txt'.format(file_prefix)
xy_table_filepath = os.path.join(output_directory, xy_table_filename)
header_names = df.columns.values
#x_series = header_names[x_series_index]
#y_series = header_names[y_series_index]
facet_series = header_names[facet_index]
peptide_series = header_names[peptide_index]
#categorization_series = header_names[categorization_index]
#print(x_series,y_series, facet_series, peptide_series, categorization_series)
data_table = df.to_csv(header = True, index = False)
print(data_table)
df = df.sort_values([facet_series, peptide_series])
data_table = df.to_csv(header = True, index = False)
print(data_table)
write_file(xy_table_filepath, data_table)
main_plot_script = '''
# Set the margins
par(mar=c(5, 5, 1, 1))
xy_data <- read.csv('%(xy_table_filename)s', header=T)
names(xy_data)[%(x_series_index)d + 1] <- "xvalues"
names(xy_data)[%(y_series_index)d + 1] <- "yvalues"
names(xy_data)[%(facet_index)d + 1] <- "facets"
names(xy_data)[%(peptide_index)d + 1] <- "peptides"
names(xy_data)[%(categorization_index)d + 1] <- "categorization"
names(xy_data)[%(categorization_shape_index)d + 1] <- "categorization_shape"
names(xy_data)[%(categorization_color_index)d + 1] <- "categorization_color"
xy_data[%(peptide_index)d + 1]
peptide_names <- sort(xy_data[[%(peptide_index)d + 1]])
peptide_names
class(peptide_names)
first_peptide = peptide_names[1]
last_peptide = peptide_names[length(peptide_names)]
xlabel <- "%(x_axis_label)s"
ylabel <- "%(y_axis_label)s"
plot_title <- "%(plot_title)s"
xy_data
# Set graph limits and the position for the correlation value
miny <- min(0.0, min(xy_data$xvalues) - 0.1) # "X-axis" values are plotted on to Y-axis
maxy <- max(1.0, max(xy_data$xvalues) + 0.1)
'''
if min_experimental_ddg != None:
main_plot_script += '''
miny <- min(miny - 0.2, %(min_experimental_ddg)f - 0.2)
'''
if min_experimental_ddg != None:
main_plot_script += '''
maxy <- max(maxy + 0.5, %(min_experimental_ddg)f + 0.5)
first_peptide
last_peptide
'''
main_plot_script += '''
#aes(color = categorization_color, shape = categorization_shape)
p <- ggplot(data=xy_data, aes(x=peptides, y = xvalues, color = categorization_color, shape = categorization_color, group = facets)) +
theme(legend.position = "none") + # hide the legend
annotate("rect", xmin = first_peptide, xmax = last_peptide, ymin = -1, ymax = +1, alpha = .2) +
xlab(xlabel) +
labs(title = "%(plot_title)s") +
theme(plot.title = element_text(color = "#555555", size=rel(0.55))) +
labs(x = xlabel, y = ylabel) +
theme(axis.text.x = element_text(angle = 90, hjust = 1, size = 3)) +
geom_point() +
scale_colour_manual(values = c("black", "blue", "green", "red")) +
scale_shape_manual(values = c(16, 18, 25, 17)) +
facet_wrap(~facets)
# Plot graph
p
dev.off()
'''
# Create the R script
plot_type = 'png'
png_plot_commands = main_plot_script % locals()
boxplot_r_script = boxplot_r_script % locals()
r_script_filename = '{0}.R'.format(file_prefix)
r_script_filepath = os.path.join(output_directory, r_script_filename)
write_file(r_script_filepath, boxplot_r_script)
# Run the R script
run_r_script(r_script_filename, cwd = output_directory)
def error_by_error_scatterplot(output_directory, file_prefix, df,
reference_series_index, x_series_index, y_series_index,
x_color, y_color,
x_series_name = None, y_series_name = None,
plot_title = '', x_axis_label = '', y_axis_label = '', similarity_range = 0.25,
add_similarity_range_annotation = True,
shape_by_category = False, shape_category_series_index = None, shape_category_title = 'Case',
label_series_index = None, label_outliers = True,
use_geom_text_repel = True,
):
""" Creates a scatterplot of error versus error intended to show which computational method (X or Y) has the least amount of error relative to a reference series.
The difference vectors (reference_series - x_series, reference_series - y_series) are created and these differences (errors)
are plotted against each other.
:param output_directory: The output directory.
:param file_prefix: A prefix for the generated files. A CSV file with the plot points, the R script, and the R output is saved along with the plot itself.
:param df: A pandas dataframe. Note: The dataframe is zero-indexed.
:param reference_series_index: The numerical index of the reference series e.g. experimental data.
:param x_series_index: The numerical index of the X-axis series e.g. predictions from a computational method.
:param y_series_index: The numerical index of the Y-axis series e.g. predictions from a second computational method.
:param x_color: The color of the "method X is better" points.
:param y_color: The color of the "method Y is better" points.
:param x_series_name: A name for the X-series which is used in the the classification legend.
:param y_series_name: A name for the Y-series which is used in the the classification legend.
:param plot_title: Plot title.
:param x_axis_label: X-axis label.
:param y_axis_label: Y-axis label.
:param similarity_range: A point (x, y) is considered as similar if |x - y| <= similarity_range.
:param add_similarity_range_annotation: If true then the similarity range is included in the plot.
:param shape_by_category: Boolean. If set then points are shaped by the column identified with shape_category_series_index. Otherwise, points are shaped by classification ("X is better", "Y is better", or "Similar")
:param shape_category_series_index: The numerical index of the series used to choose point shapes.
:param shape_category_title: The title of the shape legend.
:param label_series_index: The numerical index of the series label_series_index
:param label_outliers: Boolean. If set then label outliers using the column identified with label_series_index.
:param use_geom_text_repel: Boolean. If set then the ggrepel package is used to avoid overlapping labels.
This function was adapted from the Kortemme Lab covariation benchmark (https://github.com/Kortemme-Lab/covariation).
todo: I need to check that ggplot2 is respecting the color choices. It may be doing its own thing.
"""
try:
os.mkdir(output_directory)
except:
pass
assert (os.path.exists(output_directory))
if not isinstance(shape_category_series_index, int):
shape_by_category = False
if not isinstance(label_series_index, int):
label_outliers = False
assert(x_series_name != None and y_series_name != None)
df = df.copy()
headers = df.columns.values
num_categories = len(set(df.ix[:, shape_category_series_index].values))
legal_shapes = range(15,25+1) + range(0,14+1)
if num_categories > len(legal_shapes):
colortext.warning('Too many categories ({0}) to plot using meaningful shapes.'.format(num_categories))
shape_by_category = False
else:
legal_shapes = legal_shapes[:num_categories]
df['X_error'] = abs(df[headers[reference_series_index]] - df[headers[x_series_index]])
x_error_index = len(df.columns.values) - 1
df['Y_error'] = abs(df[headers[reference_series_index]] - df[headers[y_series_index]])
y_error_index = len(df.columns.values) - 1
# Get the list of domains common to both runs
df['Classification'] = df.apply(lambda r: _classify_smallest_error(r['X_error'], r['Y_error'], similarity_range, x_series_name, y_series_name), axis = 1)
error_classification_index = len(df.columns.values) - 1
# Create the R script
boxplot_r_script = '''
library(ggplot2)
library(gridExtra)
library(scales)
library(qualV)
library(grid)'''
if use_geom_text_repel:
boxplot_r_script +='''
library(ggrepel) # install with 'install.packages("ggrepel")' inside the R interactive shell.
'''
boxplot_r_script += '''
# PNG generation
png('%(file_prefix)s.png', width=2560, height=2048, bg="white", res=600)
txtalpha <- 0.8
redtxtalpha <- 0.8
%(png_plot_commands)s
'''
xy_table_filename = '{0}.txt'.format(file_prefix)
xy_table_filepath = os.path.join(output_directory, xy_table_filename)
data_table = df.to_csv(header = True, index = False)
write_file(xy_table_filepath, data_table)
main_plot_script = '''
# Set the margins
par(mar=c(5, 5, 1, 1))
xy_data <- read.csv('%(xy_table_filename)s', header=T)
names(xy_data)[%(x_error_index)d + 1] <- "xerrors"
names(xy_data)[%(y_error_index)d + 1] <- "yerrors"
'''
if label_outliers:
main_plot_script +='''names(xy_data)[%(label_series_index)d + 1] <- "outlier_labels"'''
main_plot_script +='''
names(xy_data)[%(shape_category_series_index)d + 1] <- "categories"
xy_data[%(x_error_index)d + 1]
xy_data[%(y_error_index)d + 1]
# coefs contains two values: (Intercept) and yerrors
coefs <- coef(lm(xerrors~yerrors, data = xy_data))
fitcoefs = coef(lm(xerrors~0 + yerrors, data = xy_data))
fitlmv_yerrors <- as.numeric(fitcoefs[1])
lmv_intercept <- as.numeric(coefs[1])
lmv_yerrors <- as.numeric(coefs[2])
lm(xy_data$yerrors~xy_data$xerrors)
xlabel <- "%(x_axis_label)s"
ylabel <- "%(y_axis_label)s"
plot_title <- "%(plot_title)s"
rvalue <- cor(xy_data$yerrors, xy_data$xerrors)
# Alphabetically, "Similar" < "X" < "Y" so the logic below works
countsim <- paste("Similar =", dim(subset(xy_data, Classification=="Similar"))[1])
countX <- paste("%(x_series_name)s =", dim(subset(xy_data, Classification=="%(x_series_name)s"))[1])
countY <- paste("%(y_series_name)s =", dim(subset(xy_data, Classification=="%(y_series_name)s"))[1])
countX
countY
countsim
# Set graph limits and the position for the correlation value
minx <- min(0.0, min(xy_data$xerrors) - 0.1)
miny <- min(0.0, min(xy_data$yerrors) - 0.1)
maxx <- max(1.0, max(xy_data$xerrors) + 0.1)
maxy <- max(1.0, max(xy_data$yerrors) + 0.1)
# Create a square plot (x-range = y-range)
minx <- min(minx, miny)
miny <- minx
maxx <- max(maxx, maxy)
maxy <- maxx
xpos <- maxx / 25.0
ypos <- maxy - (maxy / 25.0)
ypos_2 <- maxy - (2 * maxy / 25.0)
plot_scale <- scale_color_manual(
"Counts",
values = c( "Similar" = '#444444', "%(x_series_name)s" = '%(x_color)s', "%(y_series_name)s" ='%(y_color)s'),
labels = c( "Similar" = countsim, "%(x_series_name)s" = countX, "%(y_series_name)s" = countY) )'''
if add_similarity_range_annotation:
main_plot_script += '''
# Polygon denoting the similarity range. We turn off plot clipping below (gt$layout$clip) so we need to be more exact than using 4 points when defining the region
boxy_mc_boxface <- data.frame(
X = c(minx - 0, maxx - %(similarity_range)f, maxx + 0, maxx + 0, 0 + %(similarity_range)f, 0),
Y = c(minx - 0 + %(similarity_range)f, maxx + 0, maxx + 0, maxx + 0 -%(similarity_range)f, 0, 0 )
)'''
else:
main_plot_script += '''
# Polygon denoting the similarity range. We turn off plot clipping below (gt$layout$clip) so we need to be more exact than using 4 points when defining the region
boxy_mc_boxface <- data.frame(
X = c(minx - 1, maxx + 1, maxx + 1, minx - 1),
Y = c(minx - 1 + %(similarity_range)f, maxx + 1 + %(similarity_range)f, maxx + 1 - %(similarity_range)f, minx - 1 - %(similarity_range)f)
)'''
if shape_by_category:
main_plot_script += '''
# Plot
p <- qplot(main="", xerrors, yerrors, data=xy_data, xlab=xlabel, ylab=ylabel, alpha = I(txtalpha), shape=factor(categories), col=factor(Classification)) +'''
else:
main_plot_script += '''
# Plot
p <- qplot(main="", xerrors, yerrors, data=xy_data, xlab=xlabel, ylab=ylabel, alpha = I(txtalpha), shape=factor(Classification), col=factor(Classification)) +'''
main_plot_script += '''
geom_polygon(data=boxy_mc_boxface, aes(X, Y), fill = "#bbbbbb", alpha = 0.4, color = "darkseagreen", linetype="blank", inherit.aes = FALSE, show.legend = FALSE) +
plot_scale +
geom_point() +
guides(col = guide_legend()) +
labs(title = "%(plot_title)s") +
theme(plot.title = element_text(color = "#555555", size=rel(0.75))) +
theme(axis.title = element_text(color = "#555555", size=rel(0.6))) +
theme(legend.title = element_text(color = "#555555", size=rel(0.45)), legend.text = element_text(color = "#555555", size=rel(0.4))) +
coord_cartesian(xlim = c(minx, maxx), ylim = c(miny, maxy)) + # set the graph limits
annotate("text", hjust=0, size = 2, colour="#222222", x = xpos, y = ypos, label = sprintf("R = %%0.2f", round(rvalue, digits = 4))) + # add correlation text; hjust=0 sets left-alignment. Using annotate instead of geom_text avoids blocky text caused by geom_text being run multiple times over the series'''
if label_outliers:
if use_geom_text_repel:
main_plot_script += '''
# Label outliers
geom_text_repel(size=1.5, segment.size = 0.15, color="#000000", alpha=0.6, data=subset(xy_data, abs(yerrors - xerrors) > maxx/3 & xerrors <= maxx / 2 & yerrors >=maxy/2), aes(xerrors, yerrors-maxy/100, label=outlier_labels)) +
geom_text_repel(size=1.5, segment.size = 0.15, color="#000000", alpha=0.6, data=subset(xy_data, abs(yerrors - xerrors) > maxx/3 & xerrors <= maxx / 2 & yerrors < maxy/2), aes(xerrors, yerrors+2*maxy/100, label=outlier_labels)) +
geom_text_repel(size=1.5, segment.size = 0.15, color="#000000", alpha=0.6, data=subset(xy_data, abs(yerrors - xerrors) > maxx/3 & xerrors > maxx / 2 & yerrors >=maxy/2), aes(xerrors, yerrors-maxy/100, label=outlier_labels)) +
geom_text_repel(size=1.5, segment.size = 0.15, color="#000000", alpha=0.6, data=subset(xy_data, abs(yerrors - xerrors) > maxx/3 & xerrors > maxx / 2 & yerrors < maxy/2), aes(xerrors, yerrors+2*maxy/100, label=outlier_labels)) +'''
else:
main_plot_script += '''
# Label outliers
geom_text(hjust = 0, size=1.5, color="#000000", alpha=0.6, data=subset(xy_data, abs(yerrors - xerrors) > maxx/3 & xerrors <= maxx / 2 & yerrors >=maxy/2), aes(xerrors, yerrors-maxy/100, label=outlier_labels)) +
geom_text(hjust = 0, size=1.5, color="#000000", alpha=0.6, data=subset(xy_data, abs(yerrors - xerrors) > maxx/3 & xerrors <= maxx / 2 & yerrors < maxy/2), aes(xerrors, yerrors+2*maxy/100, label=outlier_labels)) +
geom_text(hjust = 1, size=1.5, color="#000000", alpha=0.6, data=subset(xy_data, abs(yerrors - xerrors) > maxx/3 & xerrors > maxx / 2 & yerrors >=maxy/2), aes(xerrors, yerrors-maxy/100, label=outlier_labels)) +
geom_text(hjust = 1, size=1.5, color="#000000", alpha=0.6, data=subset(xy_data, abs(yerrors - xerrors) > maxx/3 & xerrors > maxx / 2 & yerrors < maxy/2), aes(xerrors, yerrors+2*maxy/100, label=outlier_labels)) +'''
counts_title = 'Counts'
if add_similarity_range_annotation:
counts_title += '*'
main_plot_script += '''
#geom_text(hjust = 0, size=1.5, color="#000000", alpha=0.6, data=subset(xy_data, abs(yvalues - xvalues) > 2 & xvalues <= 0), aes(xvalues, yvalues+0.35, label=Origin_of_peptide), check_overlap = TRUE) + # label outliers
#geom_text(hjust = 1, size=1.5, color="#000000", alpha=0.6, data=subset(xy_data, abs(yvalues - xvalues) > 2 & xvalues > 0), aes(xvalues, yvalues+0.35, label=Origin_of_peptide), check_overlap = TRUE) + # label outliers
scale_colour_manual('%(counts_title)s', values = c('#444444', '%(x_color)s', '%(y_color)s'),
labels = c( "Similar" = countsim, "%(x_series_name)s" = countX, "%(y_series_name)s" = countY)) +'''
if shape_by_category:
legal_shapes_str = ', '.join(map(str, legal_shapes))
main_plot_script += '''
scale_shape_manual('%(shape_category_title)s', values = c(%(legal_shapes_str)s),
labels = c( "Similar" = countsim, "%(x_series_name)s" = countX, "%(y_series_name)s" = countY))'''
else:
main_plot_script += '''
scale_shape_manual('%(counts_title)s', values = c(18, 16, 15),
labels = c( "Similar" = countsim, "%(x_series_name)s" = countX, "%(y_series_name)s" = countY))'''
if add_similarity_range_annotation:
main_plot_script += '''+
# Add a caption
annotation_custom(grob = textGrob(gp = gpar(fontsize = 5), hjust = 0, sprintf("* Similar \\u225d \\u00b1 %%0.2f", round(%(similarity_range)f, digits = 2))), xmin = maxx + (2 * maxx / 10), ymin = -1, ymax = -1)'''
main_plot_script += '''
# Plot graph
p
'''
if add_similarity_range_annotation:
main_plot_script += '''
# Code to override clipping
gt <- ggplot_gtable(ggplot_build(p))
gt$layout$clip[gt$layout$name=="panel"] <- "off"
grid.draw(gt)'''
main_plot_script +='''
dev.off()
'''
# Create the R script
plot_type = 'png'
png_plot_commands = main_plot_script % locals()
boxplot_r_script = boxplot_r_script % locals()
r_script_filename = '{0}.R'.format(file_prefix)
r_script_filepath = os.path.join(output_directory, r_script_filename)
write_file(r_script_filepath, boxplot_r_script)
# Run the R script
run_r_script(r_script_filename, cwd = output_directory)
def multicategory_scatterplot(
output_directory,
file_prefix,
df,
x_series_index,
y_series_index,
category_series_index,
series_color,
plot_title="",
x_axis_label="",
y_axis_label="",
min_predicted_ddg=None,
max_predicted_ddg=None,
min_experimental_ddg=None,
max_experimental_ddg=None,
):
"""This function was adapted from the covariation benchmark."""
# todo: Abstract this graph from the current usage (DDG measurements).
# todo: make the capped value for unquantified but classified measurements (e.g. DDG > 7 kcal/mol) parameterizable
# todo: add an option to identify outliers by standard deviations (over the set of errors |x - y|) rather than by fixed value
# todo: add an option to use geom_text_repel to avoid/reduce overlapping text
# todo: allow users to provide colors for the facets / categories
# Changeset
# todo: Change it to take in a pandas dataframe instead of the data_table_headers + data_table parameters.
# todo: Add exception if number of cases > 2 so the general case can be implemented once we have test data.
# todo: use one column as the category e.g. "PDB". assert that there is a maximum number of categories. Test with > 2 categories
# todo: remove all references to SNX27 and NHERF1 below and loop over the set of categories instead
# print(df[facet_index])
color_map = {}
categories = list(df.ix[:, category_series_index].unique())
print(type(categories))
num_categories = len(categories)
category_colors = get_spaced_plot_colors(num_categories)
for x in xrange(num_categories):
color_map[categories[x]] = "#" + category_colors[x]
df["CategorizationColor"] = df.apply(lambda r: color_map[r[category_series_index]], axis=1)
categorization_color_index = len(df.columns.values) - 1
# Monday: continue here
print(df)
sys.exit(0)
try:
os.mkdir(output_directory)
except:
pass
assert os.path.exists(output_directory)
df["Categorization"] = df.apply(
lambda r: _determine_fraction_correct_class(r[x_series_index], r[y_series_index])[0], axis=1
)
categorization_index = len(df.columns.values) - 1
df["CategorizationShape"] = df.apply(
lambda r: _determine_fraction_correct_class(r[x_series_index], r[y_series_index])[1], axis=1
)
categorization_shape_index = len(df.columns.values) - 1
# Create the R script
boxplot_r_script = """
library(ggplot2)
library(gridExtra)
library(scales)
library(qualV)
# PNG generation
png('%(file_prefix)s.png', width=2560, height=2048, bg="white", res=600)
txtalpha <- 0.6
redtxtalpha <- 0.6
%(png_plot_commands)s
"""
xy_table_filename = "{0}.txt".format(file_prefix)
xy_table_filepath = os.path.join(output_directory, xy_table_filename)
write_file(xy_table_filepath, "\n".join(",".join(map(str, line)) for line in [data_table_headers] + data_table))
single_plot_commands = """
# Set the margins
par(mar=c(5, 5, 1, 1))
xy_data <- read.csv('%(xy_table_filename)s', header=T)
names(xy_data)[%(x_series_index)d + 1] <- "xvalues"
names(xy_data)[%(y_series_index)d + 1] <- "yvalues"
# coefs contains two values: (Intercept) and yvalues
coefs <- coef(lm(xvalues~yvalues, data = xy_data))
fitcoefs = coef(lm(xvalues~0 + yvalues, data = xy_data))
fitlmv_yvalues <- as.numeric(fitcoefs[1])
lmv_intercept <- as.numeric(coefs[1])
lmv_yvalues <- as.numeric(coefs[2])
lm(xy_data$yvalues~xy_data$xvalues)
xlabel <- "%(x_axis_label)s"
ylabel <- "%(y_axis_label)s"
plot_title <- "%(plot_title)s"
rvalue <- cor(xy_data$yvalues, xy_data$xvalues)
rvalue
xy_data
#3QDO = SNX27
#1G9O = NHERF1
valid_xy_data <- xy_data[which(xy_data$xvalues < 6.99),]
rvalue <- cor(valid_xy_data$yvalues, valid_xy_data$xvalues)
rvalue
valid_xy_data
valid_xy_data_NHERF1 <- xy_data[which(xy_data$xvalues < 6.99 & xy_data$PDB == '1G9O'),]
rvalue_NHERF1 <- cor(valid_xy_data_NHERF1$yvalues, valid_xy_data_NHERF1$xvalues)
rvalue_NHERF1
valid_xy_data_NHERF1
coefs_NHERF1 <- coef(lm(xvalues~yvalues, data = valid_xy_data_NHERF1))
lmv_intercept_NHERF1 <- as.numeric(coefs_NHERF1[1])
lmv_yvalues_NHERF1 <- as.numeric(coefs_NHERF1[2])
valid_xy_data_SNX27 <- xy_data[which(xy_data$xvalues < 6.99 & xy_data$PDB == '3QDO'),]
rvalue_SNX27 <- cor(valid_xy_data_SNX27$yvalues, valid_xy_data_SNX27$xvalues)
rvalue_SNX27
valid_xy_data_SNX27
coefs_SNX27 <- coef(lm(xvalues~yvalues, data = valid_xy_data_SNX27))
lmv_intercept_SNX27 <- as.numeric(coefs_SNX27[1])
lmv_yvalues_SNX27 <- as.numeric(coefs_SNX27[2])
lmv_intercept
lmv_yvalues
lmv_intercept_NHERF1
lmv_yvalues_NHERF1
lmv_intercept_SNX27
lmv_yvalues_SNX27
# Set graph limits and the position for the correlation value
minx <- min(0.0, min(xy_data$xvalues) - 0.1)
miny <- min(0.0, min(xy_data$yvalues) - 0.1)
maxx <- max(1.0, max(xy_data$xvalues) + 0.1)
maxy <- max(1.0, max(xy_data$yvalues) + 0.1)
"""
if min_predicted_ddg != None:
single_plot_commands += """
miny <- min(miny - 0.2, %(min_predicted_ddg)f - 0.2)
"""
if max_predicted_ddg != None:
single_plot_commands += """
maxy <- max(maxy + 0.5, %(max_predicted_ddg)f + 0.5)
miny <- -6
maxy <- 12.5
"""
if min_experimental_ddg != None:
single_plot_commands += """
minx <- min(minx, %(min_experimental_ddg)f)
"""
if max_experimental_ddg != None:
single_plot_commands += """
maxx <- max(maxx, %(max_experimental_ddg)f) + 0.2
"""
single_plot_commands += """
xpos <- minx + 0.2
ypos <- maxy - 1
ypos_SNX27 <- ypos - 1
ypos_NHERF1 <- ypos_SNX27 - 1
lrt <- expression('R'^tst)
p <- qplot(main="", xvalues, yvalues, data=xy_data, xlab=xlabel, ylab=ylabel, shape = PDB, alpha = I(txtalpha)) +
geom_point(aes(color = PDB), alpha = 0.6) +
scale_colour_manual(name="", values = c("1G9O"="orange", "3QDO"="blue", "3"="red", "value3"="grey", "value2"="black")) +
labs(title = "%(plot_title)s") +
theme(plot.title = element_text(color = "#555555", size=rel(0.75))) +
# Correlation fit lines (global + one per facet
geom_abline(size = 0.125, color="black", intercept = lmv_intercept, slope = lmv_yvalues, alpha=0.2) +
geom_abline(size = 0.125, color="orange", intercept = lmv_intercept_NHERF1, slope = lmv_yvalues_NHERF1, alpha=0.4) +
geom_abline(size = 0.125, color="blue", intercept = lmv_intercept_SNX27, slope = lmv_yvalues_SNX27, alpha=0.4) +
geom_abline(slope=1, intercept=0, linetype=3, size=0.25, alpha=0.4) + # add a diagonal (dotted)
coord_cartesian(xlim = c(minx, maxx), ylim = c(miny, maxy)) + # set the graph limits
geom_text(hjust = 0, size=1.5, color="#000000", alpha=0.6, data=subset(xy_data, abs(yvalues - xvalues) > 2 & xvalues <= 0), aes(xvalues, yvalues+0.35, label=Origin_of_peptide), check_overlap = TRUE) + # label outliers
geom_text(hjust = 1, size=1.5, color="#000000", alpha=0.6, data=subset(xy_data, abs(yvalues - xvalues) > 2 & xvalues > 0), aes(xvalues, yvalues+0.35, label=Origin_of_peptide), check_overlap = TRUE) + # label outliers
geom_text(hjust=0, size=2, colour="black", aes(x = xpos, y = ypos, label = sprintf("R == %%0.2f", round(rvalue, digits = 4))), parse = TRUE) +
geom_text(hjust=0, size=2, colour="darkorange", aes(x = xpos, y = ypos_NHERF1, label = sprintf("R[NHERF] == %%0.2f", round(rvalue_NHERF1, digits = 4))), parse = TRUE) +
geom_text(hjust=0, size=2, colour="blue", aes(x = xpos, y = ypos_SNX27, label = sprintf("R[SNX27] == %%0.2f", round(rvalue_SNX27, digits = 4))), parse = TRUE) +
theme(legend.position = "none")
# geom_text(hjust=0, size=2, colour="black", aes(xpos, ypos, fontface="plain", family = "sans", label=paste(sprintf("R = %%0.2f%%s", round(rvalue, digits = 4), lrt), expression('R'[3]) ))) # add correlation text; hjust=0 sets left-alignment
#geom_text(hjust=0, size=3, colour="black", aes(xpos, ypos, fontface="plain", family = "sans", label=sprintf("R = %%0.2f", round(rvalue, digits = 4)))) # add correlation text; hjust=0 sets left-alignment
# geom_text(hjust=0, size=3, colour="black", aes(xpos, ypos, fontface="plain", family = "sans", label=sprintf("R = %%0.2f", round(rvalue, digits = 4)))) # add correlation text; hjust=0 sets left-alignment
# Plot graph
p
dev.off()
"""
# geom_point(aes(color = C)) +
# color = "%(series_color)s"
# Create the R script
plot_type = "png"
png_plot_commands = single_plot_commands % locals()
boxplot_r_script = boxplot_r_script % locals()
r_script_filename = "{0}.R".format(file_prefix)
r_script_filepath = os.path.join(output_directory, r_script_filename)
write_file(r_script_filepath, boxplot_r_script)
# Run the R script
run_r_script(r_script_filename, cwd=output_directory)
