capable-plots

Capable Labs' house style for matplotlib/seaborn/plotly figures, plus the shared dose-response curve math used across our functional assays.

It is a styling library: you draw with normal matplotlib/seaborn or plotly, and capable-plots makes it look right and saves it correctly. It is not a plotting wrapper — the only drawing helpers are the handful of figures we make constantly (dose-response, group box + strip).

Install

pip install -e .            # from a clone, for development
pip install -e '.[seaborn]' # if you use group_box
pip install -e '.[plotly]'  # if you draw with plotly

Quick start (matplotlib)

import matplotlib.pyplot as plt
import capable_plots as cap

with cap.house:
    fig, ax = plt.subplots(figsize=cap.figsize("house-slide"))
    ax.plot(x, y, color=cap.CAPABLE)
    cap.style_axis(ax)                              # spine/tick cleanup
cap.save(fig, "figure1")                            # 300dpi PNG + editable SVG

Quick start (plotly)

import plotly.express as px
import capable_plots as cap

fig = px.scatter(df, x="dose", y="response", template=cap.plotly_house)
fig.update_layout(**cap.plotly_figsize("house-slide"))
cap.plotly_save(fig, "figure1")                     # PNG + SVG via kaleido

Set as the plotly default globally or scope it:

import plotly.io as pio
pio.templates.default = cap.plotly_house            # global

with cap.plotly_house_ctx():                        # scoped, auto-restored
    ...

Theme

Theme	Draws with	Look	Source style guide
cap.house	matplotlib/seaborn	pitch-deck: serif, transparent bg, thick lines	Capable house style
cap.plotly_house	plotly	same, as a plotly Template	Capable house style

One theme by design — simplicity first; more can be added later as additional Theme / template instances. For matplotlib use as a context manager (with cap.house:), globally (cap.house.apply()), or just pull sizes/colors (cap.figsize(...), cap.colors(...), cap.CAPABLE). For plotly, pass template=cap.plotly_house per figure, set pio.templates.default, or use with cap.plotly_house_ctx():.

Customizing

house is the default; derive a tweaked theme with customize() — it returns a new theme and never mutates house:

light = cap.house.customize(
    background="white",        # "transparent"/"none", or any color ("white", "#fff")
    font="Helvetica",          # a name or fallback list; DejaVu Sans appended as backup
    font_size=10,
    line_width=1.0,
    palette=cap.colors("colorblind"),
    rc={"figure.dpi": 200},    # escape hatch: any raw rcParams
)
with light:
    ...

Available palettes: "capable_pair" (brand placebo/treatment), "colorblind" (Okabe-Ito, 8-color), and "vibrant" — a 9-hue energetic qualitative set for pitch decks/marketing contexts where CAPABLE/PLACEBO read as too muted, with parallel "vibrant_light" (background tints) and "vibrant_dark" (text-safe shades) variants at the same index per hue.

Every knob is optional — omit one to inherit house's value.

Assay curve math

from capable_plots.assay import curves, plots

fit = curves.fit_4pl(x, y, direction="descending",   # or "ascending"
                     ns_mean=ns, ref_top=ref_top)
plots.dose_response(ax, x, y, fit)                    # log axis + EC50/Emax annotation

One canonical 4PL fitter for every modality. Whether the signal rises or falls with dose is a single explicit direction= argument — not three diverging copies.