p, c, dtype, ctissues = ("UBFD1", "TP63", "crispr", ["Lung"])
plot_df = pd.concat(
[
drespo.loc[[c]].T.add_suffix("_y")
if dtype == "drug" else crispr.loc[[c]].T.add_suffix("_y"),
prot.loc[[p]].T.add_suffix("_prot"),
gexp.loc[[p]].T.add_suffix("_gexp"),
prot_obj.ss["tissue"],
],
axis=1,
sort=False,
).dropna(subset=[f"{c}_y", f"{p}_prot"])
# Protein
ax = GIPlot.gi_tissue_plot(f"{p}_prot", f"{c}_y", plot_df)
if dtype == "drug":
ax.axhline(np.log(drespo_maxc[c]),
ls="--",
lw=.3,
color=CrispyPlot.PAL_DTRACE[1])
ax.set_xlabel(f"{p}\nProtein intensities")
ax.set_ylabel(
f"{c}\n{'Drug response IC50' if dtype == 'drug' else 'CRISPR-Cas9 (log2 FC)'}"
)
plt.savefig(
f"{RPATH}/TopHits_{p}_{c}_{dtype}_regression_tissue_plot.pdf",
bbox_inches="tight",
)
(drespo.T < dmax[drespo.index]).sum().rename("nsamples"),
drespo.T[drespo.T < dmax[drespo.index]].median().rename("dependency"),
ml_scores,
],
axis=1,
)
drug_selective["name"] = [i.split(";")[1] for i in drug_selective.index]
drug_selective_set = set(drug_selective.query("skew < -2").index)
# Scatter
grid = GIPlot.gi_regression(
"skew",
"median",
drug_selective,
size="dependency",
size_inverse=True,
size_legend_title="Median IC50",
plot_reg=False,
plot_annot=False,
)
grid.ax_joint.axvline(-1, c=GIPlot.PAL_DTRACE[1], lw=0.3, ls="--")
g_highlight_df = drug_selective.query("skew < -1").sort_values("skew").head(5)
labels = [
grid.ax_joint.text(
row["skew"], row["median"], row["name"], color="k", fontsize=4
)
for _, row in g_highlight_df.iterrows()
]
adjust_text(
labels,
drug_pca_df.corr(),
cmap="Spectral",
annot=True,
center=0,
fmt=".2f",
annot_kws=dict(size=4),
lw=0.05,
figsize=(3, 3),
)
plt.savefig(f"{RPATH}/drug_pca_clustermap.pdf", bbox_inches="tight", dpi=600)
plt.close("all")
#
y_var = "PC1"
g = GIPlot.gi_regression("growth", y_var, drug_pca_df, lowess=True)
g.set_axis_labels("Growth rate", f"{y_var} ({drug_vexp[y_var]*100:.1f}%)")
plt.savefig(f"{RPATH}/drug_pca_regression_growth.pdf", bbox_inches="tight", dpi=600)
plt.close("all")
# Covariates
#
# CRISPR
covs_crispr = LMModels.define_covariates(
institute=crispr_obj.merged_institute,
medium=True,
cancertype=False,
tissuetype=False,
mburden=False,
ploidy=True,
signature=[i[0] if len(i) > 0 else "All" for i in plot_df["signature"]]
)
ax_min = plot_df[["gexp_corr", "prot_corr"]].min().min() * 1.1
ax_max = plot_df[["gexp_corr", "prot_corr"]].max().max() * 1.1
discrete_pal = pd.Series(
sns.color_palette("tab10").as_hex()[: len(signatures)], index=signatures
)
discrete_pal["All"] = CrispyPlot.PAL_DTRACE[0]
grid = GIPlot.gi_regression_marginal(
"gexp_corr",
"prot_corr",
"signature",
plot_df,
plot_reg=False,
plot_annot=False,
scatter_kws=dict(edgecolor="w", lw=0.1, s=8),
discrete_pal=discrete_pal,
)
grid.ax_joint.plot([ax_min, ax_max], [ax_min, ax_max], "k--", lw=0.3)
grid.ax_joint.set_xlim(ax_min, ax_max)
grid.ax_joint.set_ylim(ax_min, ax_max)
labels = [
grid.ax_joint.text(row["gexp_corr"], row["prot_corr"], i, color="k", fontsize=4)
for i, row in plot_df.query("signature != 'All'")
.sort_values("attenuation", ascending=False)
.head(15)
.iterrows()
for g in plot_df.index])
plot_df = plot_df.assign(
signature=[i[0] if len(i) > 0 else "All" for i in plot_df["signature"]])
ax_min = plot_df[["gexp_corr", "prot_corr"]].min().min() * 1.1
ax_max = plot_df[["gexp_corr", "prot_corr"]].max().max() * 1.1
discrete_pal = pd.Series(sns.color_palette("tab10").as_hex()[:len(signatures)],
index=signatures)
discrete_pal["All"] = CrispyPlot.PAL_DTRACE[0]
grid = GIPlot.gi_regression_marginal(
"gexp_corr",
"prot_corr",
"signature",
plot_df,
plot_reg=False,
plot_annot=False,
scatter_kws=dict(edgecolor="w", lw=0.1, s=8),
discrete_pal=discrete_pal,
)
grid.ax_joint.plot([ax_min, ax_max], [ax_min, ax_max], "k--", lw=0.3)
grid.ax_joint.set_xlim(ax_min, ax_max)
grid.ax_joint.set_ylim(ax_min, ax_max)
labels = [
grid.ax_joint.text(row["gexp_corr"],
row["prot_corr"],
i,
color="k",
fontsize=4)
dsets = ["crispr", "gexp", "prot"]
order = natsorted(set(df_corr_ppi["n_ppi"]))
pal = pd.Series(sns.color_palette("Blues_d", n_colors=len(order)).as_hex(),
index=order)
_, axs = plt.subplots(1, len(dsets), figsize=(2 * len(dsets), 2), dpi=600)
for i, dt in enumerate(dsets):
ax = axs[i]
ax = GIPlot.gi_classification(
dt,
"n_ppi",
df_corr_ppi,
orient="h",
palette=pal.to_dict(),
order=order,
ax=ax,
)
if dt == "crispr":
xlabel, title = "Gene essentiality\n(mean scaled FC)", "CRISPR-Cas9"
elif dt == "gexp":
xlabel, title = "Gene expression\n(mean voom)", "RNA-Seq"
else:
xlabel, title = "Protein abundance\n(mean intensities)", "SWATH-MS"
ax.set_xlabel(xlabel)
ax.set_ylabel("Number of protein interactions" if i == 0 else None)
ax.set_title(title)
cbar_pos=None,
figsize=np.array(plot_df.shape) * 0.275,
)
plt.savefig(
f"{RPATH}/ProteinTranscriptSample_cfeatures_clustermap.pdf",
bbox_inches="tight",
transparent=True,
)
plt.close("all")
#
for z_var in ["CopyNumberInstability", "ploidy"]:
ax = GIPlot.gi_continuous_plot("prot_corr",
"gexp_prot_corr",
z_var,
satt_corr,
mid_point_norm=False)
ax.set_xlabel("Sanger&CMRI\nProtein ~ Copy number (Pearson's R)")
ax.set_ylabel("Sanger&CMRI\nProtein ~ Transcript (Pearson's R)")
plt.savefig(
f"{RPATH}/ProteinTranscriptSample_prot_gexp_regression_{z_var}.pdf",
bbox_inches="tight",
)
plt.close("all")
#
x_var, y_var, z_var = "ploidy", "CopyNumberInstability", "size"
ax = GIPlot.gi_continuous_plot(x_var,
y_var,
z_var,
plot_df = pd.concat(
[
mofa.factors[[f_x, f_y]],
gexp.loc[["CDH1", "VIM"]].T.add_suffix("_transcriptomics"),
prot.loc[["CDH1", "VIM"]].T.add_suffix("_proteomics"),
ss["Tissue_type"],
],
axis=1,
sort=False,
)
# Tissue plot
ax = GIPlot.gi_tissue_plot(f_x,
f_y,
plot_df,
plot_reg=False,
pal=PALETTE_TTYPE)
ax.set_xlabel(f"Factor {f_x[1:]}")
ax.set_ylabel(f"Factor {f_y[1:]}")
plt.savefig(f"{RPATH}/MultiOmics_{f_x}_{f_y}_tissue_plot.pdf",
bbox_inches="tight")
plt.savefig(f"{RPATH}/MultiOmics_{f_x}_{f_y}_tissue_plot.png",
bbox_inches="tight",
dpi=600)
plt.close("all")
# Continous annotation
for z in ["VIM_proteomics", "CDH1_proteomics"]:
ax = GIPlot.gi_continuous_plot(f_x,
f_y,
plot_df = pd.concat(
[
drespo.loc[[c]].T.add_suffix("_y")
if dtype == "drug" else crispr.loc[[c]].T.add_suffix("_y"),
prot.loc[[p]].T.add_suffix("_prot"),
gexp.loc[[p]].T.add_suffix("_gexp"),
ss["Tissue_type"],
],
axis=1,
sort=False,
).dropna(subset=[f"{c}_y", f"{p}_prot"])
# Protein
ax = GIPlot.gi_tissue_plot(f"{p}_prot",
f"{c}_y",
plot_df,
pal=PALETTE_TTYPE)
if dtype == "drug":
ax.axhline(np.log(dmaxc[c]),
ls="--",
lw=0.3,
color=CrispyPlot.PAL_DTRACE[1])
ax.set_xlabel(f"{p}\nProtein intensities")
ax.set_ylabel(
f"{c}\n{'Drug response IC50' if dtype == 'drug' else 'CRISPR-Cas9 (log2 FC)'}"
)
plt.savefig(
f"{RPATH}/TopHits_{p}_{c}_{dtype}_regression_tissue_plot.pdf",
bbox_inches="tight",
crispr.apply(skew, axis=1).rename("skew"),
crispr.median(1).rename("median"),
(crispr < -0.5).sum(1).rename("nsamples"),
crispr[crispr < -0.5].median(1).rename("dependency").abs(),
ml_scores,
],
axis=1,
)
crispr_selective_set = set(crispr_selective.query("skew < -3").index)
# Scatter
grid = GIPlot.gi_regression(
"skew",
"median",
crispr_selective,
size="dependency",
plot_reg=False,
plot_annot=False,
)
grid.ax_joint.axvline(-3, c=GIPlot.PAL_DTRACE[1], lw=0.3, ls="--")
g_highlight_df = crispr_selective.query("skew < -3").sort_values(
"skew").head(15)
labels = [
grid.ax_joint.text(row["skew"],
row["median"],
i,
color="k",
fontsize=4) for i, row in g_highlight_df.iterrows()
]
adjust_text(