def tsv_plot_output_aa(wtfile,
cond,
df,
df_se=pd.DataFrame(),
outfilename='test.pdf'):
"""
Read in score and se dataframes, output plots accordingly.
wtfile: WT sequence file, used for y-labeling
cond: The experiment condition, used for title info
df: enrich score dataframe, from plot_input folder, can also used customized df
df_se: enrich score dataframe, from plot_input folder, will match df (if customized) automatically
outfilename: outputfile path + filename, where to store the output
This function is customized for aa.
"""
row_number, col_number = df.shape
# Create Figure object that consists of two axes:
grid = GridSpec(2,
1,
height_ratios=[row_number, 1],
hspace=1 / (row_number + 1) * 2)
fig = plt.figure()
fig.set_size_inches((col_number + 8) * 0.2, (10 + row_number) * 0.2)
# Later, will use subplots_adjust to make sure each square is 1*0.2^2 inches^2
### Up to now, there is no axes object created, therefore, no plot will be shown.
# Create two axes object subplot.
mesh_ax = plt.subplot(grid[0])
cbar_ax = plt.subplot(grid[1])
# Adjust subplots to make room for add-on information:
## top: (title + amino acid grouping): 5 unit (each unit is 1*0.2 inch)
## padding between main figure and bar: 1 unit
## bottom: 3 units
## left 2.5 units
## right: 3.5 units
plt.gcf().subplots_adjust(top=(row_number + 5) / (row_number + 10),
bottom=3 / (row_number + 10),
left=5 / (col_number + 8),
right=(col_number + 5) / (col_number + 8))
# Replace 'deplete' with an arbituraliy large value, say, 1000
df_raw = df.set_index(['pos'])
for col in df_raw.columns:
df_raw[col] = df_raw.apply(lambda row: float(row[col])
if not row[col] == 'deplete' else 1000,
axis=1)
# Find score range:
ls = df_raw.values.reshape(1, -1).tolist()[0]
ls = [x for x in ls if not np.isnan(x)]
_ls = sorted(set(ls))
vmin = _ls[0]
if not _ls[-1] == 1000:
vmax = _ls[-1]
else:
vmax = _ls[-2]
# Prepare a numpy array and mask NaNs for later plotting:
arr_masked = np.ma.array(df_raw, mask=np.isnan(df_raw))
# Get color map and set colors:
cmap = plt.get_cmap("RdBu_r")
colorlim = max(-vmin, vmax)
# Recenter color map:
cmap = recentered_cmap(cmap, -colorlim, colorlim)
colors = [cmap(i) for i in range(65, 200)] # R -> G -> B
cmap_new = LinearSegmentedColormap.from_list('place_holder', colors)
# The new cmap_new will take advantage of old cmap, "shrink" the spectrum to make it lighter for better printing.
# Set grey to NaN values, and black to totally depleted ones:
cmap_new.set_bad("#808080", 1)
rgba = cmap_new(0) # The darkest in the new cmap_new, equal to cmap(65)
cmap_new.set_over(rgba)
# Plot the heatmap: return the value as mesn_pcolor as a mappable object in order to add color_bar:
mesh_pcolor = mesh_ax.pcolormesh(arr_masked,
cmap=cmap_new,
vmin=-colorlim,
vmax=colorlim)
## Below are modifications to plotting:
# Add in the color bar
cbar = fig.colorbar(mesh_pcolor, cax=cbar_ax, orientation='horizontal')
cbar.set_label("Enrich2 Score")
# Set mesh_ax y labels:
mesh_ax.set_ylabel("Position in WT")
# Set mesh_ax title:
mesh_ax.set_title("Amino Acid Distribution Map for Experiment: " + cond,
pad=50)
# pad uses points, not sure the relationship between point and inch.
# 50 works nicely it seems.
# Add in column information:
for i, x in enumerate(list(df_raw.columns)):
mesh_ax.text(i + 0.5,
len(df_raw.index) + 1,
x,
horizontalalignment="center",
verticalalignment="center",
rotation=90)
# Add in amino acid grouping information:
new_AA_GROUPS = []
cstart = 0
cend = -1
for i in range(0, len(AA_GROUPS)):
aaProperty = AA_GROUPS[i][0]
count = 0
for item in df.columns.values:
if item in Property2AA[aaProperty]:
count = count + 1
if count == 1: cstart = cend = cend + 1
else:
gap = count - 1
cstart = cend + 1
cend = cstart + gap
if cend >= cstart:
newTuple = (aaProperty, cstart, cend)
new_AA_GROUPS.append(newTuple)
for codon, start, end in new_AA_GROUPS:
mesh_ax.text((end - start + 1) / 2 + start,
row_number + 2.5,
codon,
horizontalalignment="center",
verticalalignment="center")
bar = Line2D([start + 0.125, end + 1 - 0.125],
[row_number + 2, row_number + 2],
color="black")
bar.set_clip_on(False)
mesh_ax.add_line(bar)
WT_codon = [
str(wt_codon(wtfile, int(x * 3 - 2))) for x in df['pos'].values
]
wt_aa = [str(CODON2AA[codon][0]) for codon in WT_codon]
# Add in mesh_ax y-label: aa coordinates in WT sequence:
ypos = np.arange(len(df['pos'])) + 0.5
mesh_ax.set_yticks(ypos)
labelPos = list(map(str, list(map(int, df['pos']))))
labelAA = wt_aa
labelCombine = ['-'.join(item) for item in zip(labelPos, labelAA)]
mesh_ax.set_yticklabels(labelCombine, ha='right')
# Add in deliminator horizontally:
for i in range(0, df_raw.shape[0]):
delimBar = Line2D([0, df_raw.shape[1]], [i, i],
transform=mesh_ax.transData,
color="white")
delimBar.set_clip_on(False)
mesh_ax.add_line(delimBar)
# Add in WT label onto corresponding cell:
WT_aa = [CODON2AA[x][0] for x in WT_codon]
x_coordinate = [
list(df.columns).index(x) if x in df.columns else np.nan for x in WT_aa
]
y_coordinate = list(range(df.shape[1]))
wt_xy = zip(x_coordinate, y_coordinate)
for x, y in wt_xy:
if not x == np.nan:
mesh_ax.add_patch(
Circle((x + 0.5, y + 0.5),
.1666,
fill=True,
facecolor="black",
edgecolor="none",
alpha=0.5))
# Make the figure cleaner by removing ticks:
mesh_ax.tick_params(bottom=False, left=False)
mesh_ax.get_xaxis().set_visible(False)
cbar_ax.tick_params(bottom=False)
# Add in SE if specified:
if not df_se.shape == pd.DataFrame().shape: # Add in SE
# Subset df_se to match df that might be truncated:
for col in df_se.columns:
if not col in df.columns: df_se.drop(columns=col, inplace=True)
df_se = df_se[df_se['pos'].isin(df['pos'])]
# Below is to find max se value:
se_df = df_se.drop(columns=['pos'])
tem = sorted(se_df.values.reshape(1, -1).tolist()[0])
tem = [x for x in tem if not np.isnan(x)]
se_max = tem[-1]
for row in range(len(df_se['pos'])):
for col in range(len(df_se.columns) - 1):
se_value = df_se.iloc[row, col]
corner_dist = (se_max - se_value) / (2 * se_max)
corner_dist = (1 - se_value) / 2
diag = Line2D([col + corner_dist, col + 1 - corner_dist],
[row + corner_dist, row + 1 - corner_dist],
color="black")
if se_value > 0.02 and df_raw.iloc[
row,
col] != 1000: # se_value below 0.02 will not be displayed so as totally depleted ones
mesh_ax.add_line(diag)
pylab.savefig(outfilename)
def tsv_plot_output_aa_double(wtfile, wt_mut, cond, df, df_se=pd.DataFrame(), outfilename='test.pdf', version=1, scale = 'max'):
# Same as tsv_plot_output_double but for aa mode.
row_number, col_number = df.shape
# Create Figure object that consists of two axes:
grid = GridSpec(2,1, height_ratios=[row_number,1], hspace=1/(row_number+1)*2)
fig = plt.figure()
fig.set_size_inches((col_number+10)*0.2, (10+row_number)*0.2)
# Later, will use subplots_adjust to make sure each square is 1*0.2^2 inches^2
### Up to now, there is no axes object created, therefore, no plot will be shown.
# Create two axes object subplot.
mesh_ax = plt.subplot(grid[0])
cbar_ax = plt.subplot(grid[1])
# Adjust subplots to make room for add-on information:
## top: (title + amino acid grouping): 5 unit (each unit is 1*0.2 inch)
## padding between main figure and bar: 1 unit
## bottom: 3 units
plt.gcf().subplots_adjust(top=(row_number+4.2)/(row_number+10),
bottom=2.2/(row_number+10),
left=2/(col_number+10),
right=(col_number+2)/(col_number+10))
# Replace 'deplete' with an arbituraliy large value, say, 1000
df_raw = df.set_index(['pos'])
for col in df_raw.columns:
df_raw[col] = df_raw.apply(lambda row: float(row[col])
if not row[col] == 'deplete'
else 1000, axis=1)
# Find score range:
ls = df_raw.values.reshape(1,-1).tolist()[0]
ls = [x for x in ls if not np.isnan(x)]
_ls = sorted(set(ls))
vmin = _ls[0]
if not _ls[-1] == 1000:
vmax = _ls[-1]
else:
vmax = _ls[-2]
### Try to fix the color overflow issue:
if scale == 'max':
colorlim = max(-vmin, vmax)
else:
colorlim = float(scale)
for col in df_raw.columns:
df_raw[col] = df_raw.apply(lambda row: colorlim if row[col] > colorlim and not row[col] == 1000 else row[col], axis=1)
# Prepare a numpy array and mask NaNs for later plotting:
arr_masked = np.ma.array(df_raw, mask=np.isnan(df_raw))
# Get color map and set colors:
cmap = plt.get_cmap("RdBu_r")
# Recenter color map:
cmap = recentered_cmap(cmap, -colorlim, colorlim)
# Rescale the color by cutting a fragment from it:
colors = [cmap(i) for i in range(65,200)] # R -> G -> B
cmap_new = LinearSegmentedColormap.from_list('place_holder', colors)
# The new cmap_new will take advantage of old cmap, "shrink" the spectrum to make it lighter for better printing.
# Set grey to NaN values, and black to totally depleted ones:
cmap_new.set_bad("#808080",1)
rgba = cmap_new(0) # The darkest in the new cmap_new, equal to cmap(65)
cmap_new.set_over(rgba)
# Plot the heatmap: return the value as mesn_pcolor as a mappable object in order to add color_bar:
mesh_pcolor = mesh_ax.pcolormesh(arr_masked, cmap=cmap_new, vmin=-colorlim, vmax=colorlim)
## Below are modifications to plotting:
# Add in the color bar
cbar = fig.colorbar(mesh_pcolor, cax=cbar_ax, orientation='horizontal')
cbar.set_label("Enrich2 Score")
# Set mesh_ax title:
WT_codon = [str(wt_codon(wtfile, wt_mut[0][0])), str(wt_codon(wtfile, wt_mut[0][1]))]
WT_site1 = np.nan
WT_site2 = np.nan
# These lines for determining WT label coordinates: site1 determines row position while site2 determines column position
Col_site = " (Column: " + str(int((wt_mut[0][1] + 2)/3)) + "-" + CODON2AA[WT_codon[1]][0] + " (" + WT_codon[1] + "))"
mesh_ax.set_title("Amino Acid Enrichment (Double) for : " + cond + "\n" + Col_site, pad=50)
# pad uses points, not sure the relationship between point and inch.
# 50 works nicely it seems.
# Reoder the rows according to Property order list:
PropertyOrderList = Property2AA['Polar'] + Property2AA['Charged'] + Property2AA['Non-polar'] + ('Stop',)
df_row_order = pd.DataFrame()
for aa in PropertyOrderList:
pos = [group[1] for group in CODON_GROUPS if group[0] == aa][0]
row = df[df['pos'] == pos]
df_row_order = df_row_order.append(row)
df_raw = df_row_order.copy()
df_raw.set_index('pos')
# Add in aa information: columns
for i, x in enumerate(list(df_raw.columns[:-1])):
mesh_ax.text(i + 0.5, len(df_raw.index) + 1, x,
horizontalalignment="center",
verticalalignment="center",
rotation = 90)
if x == CODON2AA[WT_codon[1]][0]: WT_site2 = i
# Add in amino acid property grouping information: columns
new_AA_GROUPS = []
cstart = 0
cend = -1
for i in range(0, len(AA_GROUPS)):
aaProperty = AA_GROUPS[i][0]
count = 0
for item in df.columns.values:
if item in Property2AA[aaProperty]:
count = count + 1
if count == 1: cstart = cend = cend + 1
else:
gap = count - 1
cstart = cend + 1
cend = cstart + gap
if cend >= cstart:
newTuple = (aaProperty, cstart, cend)
new_AA_GROUPS.append(newTuple)
for codon, start, end in new_AA_GROUPS:
if version == 2: codon = CODON321[codon]
mesh_ax.text((end - start + 1) / 2 + start,
row_number + 2.5, codon,
horizontalalignment="center",
verticalalignment="center")
bar = Line2D([start + 0.125, end + 1 - 0.125],
[row_number + 2, row_number + 2], color="black")
bar.set_clip_on(False)
mesh_ax.add_line(bar)
# Add in the deliminator for amino acid property groups: columns
delimBar = Line2D([end + 1, end + 1],
[0, len(df_raw.index) + 1], color="white")
delimBar.set_clip_on(False)
mesh_ax.add_line(delimBar)
# Add in aa information: row
for i, x in enumerate(list(df_raw['pos'][::-1])):
aa_text = [group[0] for group in CODON_GROUPS if group[1] == x][0] # determine the corresponding aa
mesh_ax.text(col_number, i+1-0.5, aa_text, # Note that if -0.5, visually not perfect but this is the best achievable
horizontalalignment="center",
verticalalignment="center",)
if CODON2AA[CodonList[int(x) - 1]][0] == CODON2AA[WT_codon[0]][0]:
posInPropertyList = [i for i, x in enumerate(PropertyOrderList[::-1]) if x == CODON2AA[WT_codon[0]][0]][0]
WT_site1 = posInPropertyList
# Add in amino acid property grouping information: rows
new_AA_GROUPS = []
cstart = 0
cend = -1
for i in range(0, len(AA_GROUPS)):
aaProperty = AA_GROUPS[i][0]
count = 0
for item in df.columns.values:
if item in Property2AA[aaProperty]:
count = count + 1
if count == 1: cstart = cend = cend + 1
else:
gap = count - 1
cstart = cend + 1
cend = cstart + gap
if cend >= cstart:
newTuple = (aaProperty, cstart, cend)
new_AA_GROUPS.append(newTuple)
for codon, start, end in new_AA_GROUPS:
if version == 2: codon = CODON321[codon]
y = row_number - (end - start)/2 - start - 0.5
mesh_ax.text(col_number+1.25, y, codon,
horizontalalignment="left",
verticalalignment="center")
x = [col_number + 1, col_number + 1]
y = [row_number - start - 1 + 0.875, row_number - end - 0.875]
bar = Line2D(x, y, color="black")
bar.set_clip_on(False)
mesh_ax.add_line(bar)
# Add in the deliminator for amino acid property groups: rows
x = [0, col_number]
y = [row_number - end - 1, row_number - end - 1]
delimBar = Line2D(x, y, color="white")
delimBar.set_clip_on(False)
mesh_ax.add_line(delimBar)
"""
# Legacy, add to figure title (colomn site) and to the right margin (row site)
# Add in mutation site info onto the left y-axis:
x = -2
y = row_number / 3 * 2
mesh_ax.text(x, y, "Row:\n" + str(int((wt_mut[0][0] + 2)/3)) + "-" + CODON2AA[WT_codon[0]][0] + "\n(" + WT_codon[0] + ")",
horizontalalignment="center",
verticalalignment="center")
mesh_ax.text(x, y/2, "Column:\n" + str(int((wt_mut[0][1] + 2)/3)) + "-" + CODON2AA[WT_codon[1]][0] + "\n(" + WT_codon[1] + ")",
horizontalalignment="center",
verticalalignment="center")
"""
# Add in first mutation site (per row) info onto figure right margin
x = col_number + 5
y = row_number / 2
mesh_ax.text(x, y, "(Row: " + str(int((wt_mut[0][0] + 2)/3)) + "-" + CODON2AA[WT_codon[0]][0] + " (" + WT_codon[0] + "))",
horizontalalignment="center",
verticalalignment="center",
rotation = 270)
# Add in WT label onto corresponding cells:
wt_aa = [str(CODON2AA[codon][0]) for codon in WT_codon]
if not WT_site1 == np.nan and not WT_site2 == np.nan:
mesh_ax.add_patch(Circle((WT_site2 + 0.5, WT_site1 + 0.5), .1666,
fill=True, facecolor="black",
edgecolor="none", alpha=0.5))
# Make the figure cleaner by removing ticks:
mesh_ax.tick_params(bottom=False, left=False)
mesh_ax.get_xaxis().set_visible(False)
mesh_ax.get_yaxis().set_visible(False)
cbar_ax.tick_params(bottom=False)
# Add in SE if specified:
if not df_se.shape == pd.DataFrame().shape: # Add in SE
# Subset df_se to match df that might be truncated:
for col in df_se.columns:
if not col in df.columns: df_se.drop(columns=col, inplace=True)
df_se = df_se[df_se['pos'].isin(df['pos'])]
# Below is to find max se value:
se_df = df_se.drop(columns=['pos'])
tem = sorted(se_df.values.reshape(1,-1).tolist()[0])
tem = [x for x in tem if not np.isnan(x)]
se_max = tem[-1]
for row in range(len(df_se['pos'])):
for col in range(len(df_se.columns)-1):
se_value = df_se.iloc[row, col]
corner_dist = (se_max - se_value)/(2 * se_max)
corner_dist = (1 - se_value) / 2
diag = Line2D([col + corner_dist, col + 1 - corner_dist],
[row + corner_dist, row + 1 - corner_dist], color="black")
if se_value > 0.02 and df_raw.iloc[row, col] != 1000: # se_value below 0.02 will not be displayed so as totally depleted ones
mesh_ax.add_line(diag)
pylab.savefig(outfilename)