ENH volcano plot

src/anndata_plot/pl/volcano_plot.py

@@ -0,0 +1,62 @@
+import holoviews as hv
+import numpy as np
+from bokeh.models import HoverTool
+from holoviews import dim
+
+
+def volcano_plot(results_df, pval_threshold=0.05, LFC_threshold=2):
+    """
+    Create a volcano plot from a results dataframe.
+
+    Parameters
+    ----------
+    results_df : pd.DataFrame
+        The results dataframe.
+    pval_threshold : float, optional
+        The p-value threshold.
+    LFC_threshold : float, optional
+        The log2 fold change threshold.
+
+    Returns
+    -------
+    p : holoviews.Scatter
+        The volcano plot.
+    """
+
+    def map_DE(a):
+        log2FoldChange, pval = a
+        if pval < pval_threshold:
+            if log2FoldChange > LFC_threshold:
+                return "Up-reg"
+            elif log2FoldChange < -LFC_threshold:
+                return "Down-reg"
+        return "Non-sig"
+
+    # Create conditions for each category
+    results_df["de_status"] = results_df[["log2FoldChange", "padj"]].apply(map_DE, axis=1)
+
+    hover = HoverTool(
+        tooltips=[
+            ("gene", "@gene"),
+        ]
+    )
+
+    p = (
+        hv.Scatter(results_df, kdims="log2FoldChange", vdims=["padj", "de_status", "gene"], label="Volcano Plot")
+        .transform(padj=-dim("padj").log10())
+        .opts(
+            color="de_status",
+            cmap={"Up-reg": "indianred", "Down-reg": "cornflowerblue", "Non-sig": "lightgrey"},
+            size=5,
+            tools=[hover],
+            width=600,
+            height=600,
+            xlabel="log2 Fold Change",
+            ylabel="-log10(adjusted p-value)",
+            title="Volcano Plot",
+        )
+        * hv.VLine(-LFC_threshold).opts(color="black", line_dash="dashed")
+        * hv.VLine(LFC_threshold).opts(color="black", line_dash="dashed")
+        * hv.HLine(-np.log10(pval_threshold)).opts(color="black", line_dash="dashed")
+    )
+    return p