Performance comparison of inference on cell cycle

Notebook compares metrics for velocity, latent time and GRN inference across different methods applied to cell cycle data.

Library imports

import pandas as pd

import matplotlib.pyplot as plt
import mplscience
import seaborn as sns

from rgv_tools import DATA_DIR, FIG_DIR

General settings

DATASET = "cell_cycle"

SAVE_FIGURES = False
if SAVE_FIGURES:
    (FIG_DIR / DATASET).mkdir(parents=True, exist_ok=True)

FIGURE_FORMATE = "svg"

Constants

TIME_METHODS = [
    "regvelo",
    "velovi",
    "scvelo",
    "unitvelo",
    "velovae_vae",
    "velovae_fullvb",
    "cell2fate",
    "tfvelo",
    "dpt",
]
VELO_METHODS = ["regvelo", "velovi", "scvelo", "unitvelo", "velovae_vae", "velovae_fullvb", "cell2fate", "tfvelo"]
GRN_METHODS = ["regvelo_grn", "grnboost2", "celloracle", "correlation", "splicejac", "tfvelo_grn"]

Data loading

time_df = []
cbc_df = []
confi_df = []
grn_df = []

for method in TIME_METHODS:
    df = pd.read_parquet(DATA_DIR / DATASET / "results" / f"{method}_correlation.parquet")
    df.columns = f"{method}_" + df.columns
    df = df.mean(0)
    time_df.append(df)

for method in VELO_METHODS:
    df = pd.read_parquet(DATA_DIR / DATASET / "results" / f"{method}_cbc.parquet")
    del df["State transition"]
    df.columns = f"{method}_" + df.columns
    df = df.mean(0)
    cbc_df.append(df)

for method in VELO_METHODS:
    df = pd.read_parquet(DATA_DIR / DATASET / "results" / f"{method}_confidence.parquet")
    df.columns = f"{method}_" + df.columns
    confi_df.append(df)

for method in GRN_METHODS:
    df = pd.read_parquet(DATA_DIR / DATASET / "results" / f"{method}_correlation.parquet")
    df = pd.DataFrame({"AUROC": df.iloc[0, 0]})
    df.columns = f"{method}_" + df.columns
    grn_df.append(df)

time_df = pd.concat(time_df).reset_index()
time_df.columns = ["method", "spearman_corr"]
cbc_df = pd.concat(cbc_df).reset_index()
cbc_df.columns = ["method", "CBC"]
confi_df = pd.concat(confi_df, axis=1).melt(var_name="method", value_name="confidence")
grn_df = pd.concat(grn_df, axis=1).melt(var_name="method", value_name="AUROC")

Processing results table

time_df.iloc[:, 0] = time_df.iloc[:, 0].str.removesuffix("_time")
cbc_df.iloc[:, 0] = cbc_df.iloc[:, 0].str.removesuffix("_CBC")
confi_df.iloc[:, 0] = confi_df.iloc[:, 0].str.removesuffix("_velocity_confidence")
grn_df.iloc[:, 0] = grn_df.iloc[:, 0].str.removesuffix("_AUROC")

Analysis

CBC

with mplscience.style_context():
    sns.set_style(style="whitegrid")
    fig, ax = plt.subplots(figsize=(3, 3), sharey=True)
    # Plot the second Seaborn plot on the first subplot
    sns.barplot(y="method", x="CBC", data=cbc_df, capsize=0.1, color="grey", order=VELO_METHODS, ax=ax)
    ax.set_xlabel("CBC", fontsize=14)
    ax.set_ylabel("")
    plt.xlim(0.55, 0.9)
    plt.show()

Latent time

with mplscience.style_context():
    sns.set_style(style="whitegrid")
    fig, ax = plt.subplots(figsize=(3, 3), sharey=True)
    # Plot the second Seaborn plot on the first subplot
    sns.barplot(y="method", x="spearman_corr", data=time_df, capsize=0.1, color="grey", order=TIME_METHODS, ax=ax)
    ax.set_xlabel("", fontsize=14)
    ax.set_ylabel("")
    plt.show()

Confidence

with mplscience.style_context():
    sns.set_style(style="whitegrid")
    fig, ax = plt.subplots(figsize=(3, 3), sharey=True)
    sns.violinplot(
        data=confi_df,
        ax=ax,
        orient="h",
        y="method",
        x="confidence",
        color="grey",
        order=VELO_METHODS,
    )
    # plt.legend(title='', loc='lower center', bbox_to_anchor=(0.5, -0.6), ncol=3)
    ax.set_xticks([0, 0.25, 0.5, 0.75, 1])
    ax.set_xticklabels([0, 0.25, 0.5, 0.75, 1])
    ax.set_xlabel("Velocity confidence", fontsize=14)
    ax.set_ylabel("")
    plt.show()

GRN

with mplscience.style_context():
    sns.set_style(style="whitegrid")
    fig, ax = plt.subplots(figsize=(3, 3), sharey=True)
    sns.violinplot(
        data=grn_df,
        ax=ax,
        orient="h",
        y="method",
        x="AUROC",
        color="grey",
        order=GRN_METHODS,
    )
    # plt.legend(title='', loc='lower center', bbox_to_anchor=(0.5, -0.6), ncol=3)
    ax.set_xlabel("AUROC", fontsize=14)
    ax.set_ylabel("")
    plt.show()