def get_fold_change_distribution(nr: NameResolver):
lab = "winata"
methods = ["star", "salmon", "kallisto"]
condition = [24, 72]
colors = palette.get_color_list
p_value = 0.05
for m, c in zip(methods, colors):
fn = nr.deseq2_results(method=m, lab=lab, condition=condition,
lfc=True)
df = pd.read_csv(fn)
data = df
data = data[data[DESEQ2_PADJ] < p_value]
data = data[DESEQ2_LOG2_CHANGE].values
density = gaussian_kde(data)
xs = np.linspace(-20, 20, 1000)
plt.plot(xs, density(xs), color=c, label=m, lw=2)
plt.fill_between(xs, density(xs), color=c, alpha=0.1)
plt.axvline(0, ls="--", color=palette.black())
plt.ylabel(f"Probability Density (p-value<{p_value})")
plt.xlabel("Log$_2$ Fold Change")
plt.legend(loc=0)
plt.title(
f"{condition[0]} v {condition[1]} ({lab} lab)")
plt.xlim(-20, 20)
plt.grid(axis="both", zorder=0, ls=":")
plt.ylim(0)
plt.tight_layout()
plt.savefig("density.png", dpi=300)
plt.show()
def run():
nr = NameResolver("config.json")
genes = INTERESTED_GENES
# geometric_mean(nr)
# check_gene_expression(nr, genes)
base = "/mnt/windows/Enigma/Zebrafish/data/deseq2/mutant"
file_wt = f"{base}/AnalysiWTvsTbx5a/star.result_lfc.csv"
search_constant_genes(nr, file_wt)
def count_pca(nr: NameResolver):
all_conditions = sorted([30, 48, 72])
lab = "hills"
methods = ["stringtie"]
combinations = list(itertools.combinations(all_conditions, 2))
dfs = []
r = {}
for pair in combinations:
for method in methods:
f = nr.deseq2_vst(method, lab=lab, condition=list(pair))
df = pd.read_csv(f).set_index("gene_id")
col_change = {}
for c in df.columns:
col_change[c] = f"{c}_{method}"
df = df.rename(columns=col_change)
cols = df.columns
r[f"{pair[0]}_{method}"] = list(cols[: int(len(cols) / 2)])
r[f"{pair[1]}_{method}"] = list(cols[int(len(cols) / 2):])
dfs.append(df)
colors = []
columns = []
labels = {}
time_colors = {}
for i, t in enumerate(all_conditions):
time_colors[str(t)] = palette.get_color_list[i]
for color, key in zip(palette.get_color_list[:len(r)], r):
labels[key] = color
columns.extend(r[key])
# colors.extend([color] * len(r[key]))
colors.extend([time_colors[key.split("_")[0]]] * len(r[key]))
df = pd.concat(dfs, axis=1, join="outer", sort=False)
df = df.loc[:, ~df.columns.duplicated()]
df = df[columns].fillna(0)
pca = PCA()
a1 = pca.fit(df.values)
var = a1.components_
percentages = [round(x, 2) for x in a1.explained_variance_ratio_ * 100]
plt.scatter(var[0, :], var[1, :], color=colors)
plt.xlabel(f"PC1: {percentages[0]} %")
plt.ylabel(f"PC2: {percentages[1]} %")
hdl = []
for key in time_colors:
hdl.append(Patch(color=time_colors[key],
label=f"{key} hpf"))
plt.legend(handles=hdl, loc=0)
plt.title(f"Counts with {methods} ({lab} lab)")
plt.tight_layout()
plt.savefig("pca.png", dpi=300)
plt.show()
def mean_sd_plot(nr: NameResolver, *, vst: bool):
lab = "winata"
condition = [24, 48]
method = "kallisto"
fn = nr.deseq2_counts(method=method, lab=lab, condition=condition)
if vst:
fn = nr.deseq2_vst(method=method, lab=lab, condition=condition)
data = pd.read_csv(fn).set_index("gene_id")
data["temp"] = data.sum(axis=1)
plt.scatter(data.mean(axis=1), data.std(axis=1), marker=".",
c=data["temp"])
plt.xlabel("Mean")
plt.ylabel("Standard Deviation")
clb = plt.colorbar()
clb.ax.set_title("Total Counts")
plt.title(f"{condition[0]} vs {condition[1]} ({method}, {lab} lab)")
plt.tight_layout()
plt.savefig("mean_sd.png", dpi=300)
plt.show()
def plot_volcano(nr: NameResolver):
lab = "winata"
method = "salmon"
condition = [24, 48]
fn = nr.deseq2_results(method=method, lab=lab,
condition=condition,
lfc=True)
volcano_plot(fn)
plt.ylim(0, 100)
plt.xlim(-15, 15)
plt.axvline(0, ls="--", color=palette.black())
plt.title(f"{condition[0]} vs {condition[1]} ({method}, {lab} lab)")
plt.tight_layout()
plt.savefig("volcano.png", dpi=300)
plt.show()
def plot_all_volcano(nr: NameResolver):
methods = ["salmon", "kallisto", "stringtie"]
conditions = [[30, 48], [48, 72], [30, 72]]
max_cols = 3
lab = "hills"
txt_cond = []
data = []
for m in methods:
txt_cond.extend([f"{x[0]}v{x[1]} {m}" for x in conditions])
for con in conditions:
fn = nr.deseq2_results(m, lab=lab, condition=con, lfc=True)
data.append(fn)
condition_array = txt_cond
dfs = data
rows = int(np.ceil(len(condition_array) / max_cols))
offset = max_cols * rows - len(condition_array)
temp = np.zeros(len(condition_array) + offset)
if max_cols > len(condition_array):
max_cols = len(condition_array)
temp = np.zeros(len(condition_array))
temp = temp.reshape(-1, max_cols)
gs = gridspec.GridSpec(temp.shape[0], temp.shape[1])
fig = plt.figure()
for ind, condition in enumerate(condition_array):
ax = fig.add_subplot(gs[ind]) # type:plt.Axes
volcano_plot(dfs[ind], ax=ax,
add_labels=False)
ax.set_title(condition)
ax.set_xlim(-10, 10)
ax.set_ylim(0, 50)
fig.add_subplot(111, frameon=False)
plt.tick_params(labelcolor='none',
top=False,
bottom=False,
left=False,
right=False)
plt.ylabel("-Log$_{10}$ Adj P value")
plt.xlabel("Log$_2$ Fold change")
plt.tight_layout()
plt.savefig("all_volcano.png", dpi=300)
plt.show()
def plot_fold_change(nr: NameResolver, genes: list):
p = Palette()
lab = "winata"
method = "star"
con = [24, 72]
fn = nr.deseq2_results(method=method, lab=lab, condition=con, lfc=True)
d = pd.read_csv(fn)
d = convert_id_to_gene(nr, d)
d = d[d['gene_id'].isin(genes)].fillna(1989)
d = d[d[DESEQ2_PADJ] <= 0.05]
d = d[d[DESEQ2_LOG2_CHANGE] > -1]
d = d[d[DESEQ2_LOG2_CHANGE] < 1]
plt.hist(d['log2FoldChange'].values, bins=50, color=p.ultramarine())
plt.ylabel("Frequency")
plt.xlabel("Log$_2$Fold Change")
plt.title(f"Genes with Log$_2$Fold change (p<0.05)\n{lab} {method} {con}")
plt.savefig("genes.png", dpi=300)
plt.show()
def count_genes(nr: NameResolver):
lab = "winata"
method = "star"
condition = [24, 72]
def __assign(x):
if x[DESEQ2_PADJ] > 0.05:
return palette.gray()
if -1 < x[DESEQ2_LOG2_CHANGE] < 1:
return palette.blue()
else:
return palette.red()
# genes = list(set(INTERESTED_GENES).difference(BASE_GENES))
genes = HOUSE_KEEPING
fn = nr.deseq2_results(method=method,
lab=lab,
condition=condition,
lfc=True)
data = convert_id_to_gene(nr, pd.read_csv(fn))
data = data[data["gene_id"].isin(genes)]
data["color"] = data.apply(lambda x: __assign(x), axis=1)
data = data.sort_values(by="gene_id")
ind = range(len(data["gene_id"].values))
plt.barh(ind,
data[DESEQ2_LOG2_CHANGE],
color=data["color"],
xerr=data[DESEQ2_LFCSE],
error_kw=dict(ecolor=palette.gray(shade=70), capsize=5),
zorder=100)
plt.yticks(ind, data["gene_id"])
plt.axvspan(-1, 1, color=palette.gray(shade=20), zorder=0)
plt.axvline(1, ls="--", color=palette.black(), zorder=0)
plt.axvline(-1, ls="--", color=palette.black(), zorder=0)
plt.axvline(0, color=palette.black(), zorder=100)
plt.grid(axis="both", ls=":", color=palette.gray(), zorder=0)
plt.title(f"{condition[0]} vs {condition[1]} ({method}, {lab} lab)")
plt.xlabel("Log$_2$ Fold Change")
plt.ylabel("Gene")
plt.tight_layout()
plt.savefig("genes.png", dpi=300)
plt.show()
def get_average_dataframe(nr: NameResolver, lab: str, time: int,
genotype: str) -> pd.DataFrame:
method = "stringtie"
dfs = []
for r in _extract_runs(lab, time, genotype):
df = pd.read_csv(nr.run_output_file(r, method), sep="\t")
df = df.sort_values(by="TPM", ascending=False)
df = df.drop_duplicates(subset="Gene ID").reset_index(drop=True)
df = df.set_index("Gene ID")
dfs.append(df)
tps = [tp[["TPM"]] for tp in dfs]
tps = pd.concat(tps, join="inner", axis=1)
name = "TPM_AVG"
tps[name] = tps.mean(axis=1)
tps = tps[[name]]
dfs = dfs[0]
del dfs["TPM"]
dfs = pd.concat([dfs, tps], join="inner", axis=1)
dfs = dfs.rename(columns={name: "TPM"}).reset_index()
return dfs
def single_ma(nr: NameResolver):
method = "star"
lab = "hills"
condition = [30, 48]
_, ax = plt.subplots()
filename = nr.deseq2_results(method,
lab=lab,
condition=condition,
lfc=True)
ma_plot(filename,
color=palette.cerulean(shade=60),
accent_color=palette.green_light(),
ax=ax,
marker=".")
ax.set_xlim(0, 1000)
ax.set_ylim(-15, 15)
ax.set_title(f"{condition[0]} v {condition[1]} ({method}, {lab} lab)")
plt.tight_layout()
plt.savefig("ma.png", dpi=300)
plt.show()
def search_constant_genes_in_multiple(nr: NameResolver, lab, method,
conditions):
# lab = "winata"
# method = "star"
# conditions = [[24, 72]]
dfs = []
for con in conditions:
fn = nr.deseq2_results(method=method, lab=lab, condition=con, lfc=True)
d = pd.read_csv(fn)
d = d[d[DESEQ2_PADJ] <= 0.05]
d = d[d[DESEQ2_LOG2_CHANGE] > -1]
d = d[d[DESEQ2_LOG2_CHANGE] < 1]
d = d[["gene_id", DESEQ2_LOG2_CHANGE]]
d = d.set_index("gene_id")
d = d.rename(columns={DESEQ2_LOG2_CHANGE: f"{con}"})
dfs.append(d)
df = pd.concat(dfs, axis=1, sort=False, join="inner").reset_index()
df = convert_id_to_gene(nr, df).set_index("gene_id")
df = df.apply(lambda x: pow(x, 2))
df["sum"] = df.sum(axis=1)
df = df.sort_values(by="sum")
needed_genes = df.index.values
return needed_genes
def run():
nr = NameResolver("config.json")
plot_volcano(nr)