Gallery¶
Figures you could drop into a manuscript today. Every image below is produced by the code beneath it, sized for a Nature single column (89 mm) and exported with embedded fonts. Click a thumbnail to jump to the code.
Dose responsebox · points · Holm-corrected tests
Tumour growthindividual units · 95% CI · end labels
Volcano plotscatter · thresholds · highlight
Expression heatmapz-score · groups · tracks · clustering
Correlation matrixmasked triangle · values · diverging map
Rainclouddensity · box · raw data · stars
Split violinsgenotype × sex · quartiles
qPCR fold changegrouped bars · SD · reference line
Pharmacokineticslog axis · error bars · markers
Regressionper-group fits · equation · R²
Biomarker distributiondensity · KDE · sample size
Temperature anomalyfill between · split by sign
Cell composition100% stacked area · end labels
Forecast fan chartnested 50/80/95% intervals
Synthetic data, real output
All datasets are simulated with a fixed random seed so every figure is reproducible. Nothing on this page is retouched: the images are the library's output, regenerated each time the documentation is built.
Dose response¶
Cell viability after 48 h of drug exposure. Every well is visible, the box
summarises it, and .compare() runs Mann–Whitney tests against the vehicle with a
Holm correction. A light-to-dark ramp encodes the dose without needing a legend.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(7)
doses = ["Vehicle", "1 µM", "5 µM", "10 µM"]
viability = pd.DataFrame({
"dose": np.repeat(doses, 12),
"viability": np.concatenate(
[rng.normal(mean, 7, 12) for mean in (100, 91, 70, 46)]
),
})
(
sp.box(viability, x="dose", y="viability")
.points()
.compare(ref="Vehicle", correction="holm", format="stars")
.colors(dict(zip(doses, ["#C6DBEF", "#6BAED6", "#2171B5", "#08306B"])))
.labels(x="", y="Cell viability (% of vehicle)")
)
Options used: .points(), .compare(),
.colors().
Tumour growth¶
Tumour volume of eight mice per arm on a log scale. Thin lines are individual
animals (units=), the band is the 95% confidence interval of the mean, and
end labels replace the legend.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(11)
days = np.arange(0, 22, 2)
arms = {"Vehicle": 0.16, "Drug A": 0.11, "Drug A + B": 0.05}
rows = []
for arm, rate in arms.items():
for mouse in range(8):
volume = 90 * np.exp(rate * days) * rng.lognormal(0, 0.14, days.size)
rows += [
{"day": d, "volume": v, "arm": arm, "mouse": f"{arm}-{mouse}"}
for d, v in zip(days, volume)
]
tumours = pd.DataFrame(rows)
(
sp.line(tumours, x="day", y="volume", hue="arm", units="mouse", error="ci")
.endlabels()
.log("y")
.colors(["#555555", "#E69F00", "#0072B2"])
.labels(x="Days after implantation", y="Tumour volume (mm³)")
)
Volcano plot¶
Differential expression of 2,000 genes. Grey points fall inside the fold-change and significance thresholds; reference lines mark both cut-offs.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(3)
n = 2000
lfc = rng.normal(0, 1.1, n)
log_p = np.abs(lfc) ** 1.6 * rng.gamma(2.2, 1.0, n) + rng.uniform(0, 0.3, n)
genes = pd.DataFrame({"lfc": lfc, "log_p": log_p})
significant = genes["log_p"] > 2
genes["call"] = np.select(
[significant & (genes["lfc"] > 1), significant & (genes["lfc"] < -1)],
["Up", "Down"],
"Not significant",
)
(
sp.scatter(genes, x="lfc", y="log_p", hue="call")
.colors({"Up": "#C0392B", "Down": "#2471A3", "Not significant": "#C8C8C8"})
.refline(-1, 1, axis="x", linestyle=":")
.refline(2, axis="y", linestyle=":")
.theme(marker_size=3, grid="none")
.labels(x="log₂ fold change", y="−log₁₀ P")
)
Expression heatmap¶
Sixteen genes in four pathways after myocardial infarction (MI). Rows are z-scored, clustered within each pathway, and annotated with a condition track. Gene names are italic, as journals require.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(5)
genes = ["GATA4", "NKX2-5", "TBX5", "HAND2", "MYH6", "MYL7", "TNNT2", "ACTC1",
"CDK1", "MKI67", "TOP2A", "CCNB1", "COL1A1", "COL3A1", "POSTN", "FN1"]
samples = [f"Sham {i}" for i in range(1, 5)] + [f"MI {i}" for i in range(1, 5)]
values = rng.normal(5, 0.7, (16, 8))
values[:8, 4:] -= 1.6 # cardiac programme down after MI
values[8:12, 4:] += 1.2 # proliferation up
values[12:, 4:] += 2.2 # fibrosis up
expression = pd.DataFrame(values, index=pd.Index(genes, name="gene"), columns=samples)
expression["pathway"] = (
["Cardiac TF"] * 4 + ["Sarcomere"] * 4 + ["Cell cycle"] * 4 + ["Fibrosis"] * 4
)
(
sp.heatmap(expression)
.scale("row")
.groups(rows="pathway")
.cluster(cols=False)
.tracks(cols={"Condition": ["Sham"] * 4 + ["MI"] * 4})
.cmap("tol_sunset", center=0)
.colorbar("Row z-score")
.ticklabels(italic=True)
)
Correlation matrix¶
Pairwise Pearson correlations between clinical variables. The redundant upper triangle is masked and the diverging colormap is centred on zero.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(21)
variables = ["Age", "BMI", "SBP", "LDL", "HbA1c", "CRP", "eGFR"]
latent = rng.normal(size=(7, 3))
covariance = latent @ latent.T + np.eye(7)
cohort = pd.DataFrame(
rng.multivariate_normal(np.zeros(7), covariance, 300), columns=variables
)
(
sp.heatmap(cohort.corr(), square=True)
.mask("upper")
.cmap("tol_bu_rd", center=0, vmin=-1, vmax=1)
.values(".2f")
.colorbar("Pearson r")
)
Raincloud¶
Reaction times in a Stroop task. The raincloud combines a half-violin, a box and the raw trials, the honest way to show skewed data.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(42)
conditions = ["Congruent", "Neutral", "Incongruent"]
stroop = pd.DataFrame({
"condition": np.repeat(conditions, 60),
"rt": np.concatenate(
[rng.lognormal(np.log(mu), 0.18, 60) for mu in (480, 515, 610)]
),
})
(
sp.raincloud(stroop, x="condition", y="rt")
.compare(ref="Congruent", format="stars")
.colors(dict(zip(conditions, ["#44AA99", "#DDCC77", "#CC6677"])))
.labels(x="Reaction time (ms)", y="") # horizontal: values run along x
)
Split violins¶
A behavioural score by genotype, with females and males sharing each violin. Dashed lines mark the quartiles.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(8)
behaviour = pd.DataFrame({
"genotype": np.tile(np.repeat(["WT", "Het", "KO"], 40), 2),
"sex": np.repeat(["Female", "Male"], 120),
})
effect = behaviour["genotype"].map({"WT": 0, "Het": 0.6, "KO": 1.6})
behaviour["score"] = rng.normal(0, 1, 240) + effect + 0.5 * (behaviour["sex"] == "Male")
(
sp.violin(behaviour, x="genotype", y="score", hue="sex", split=True, inner="quartiles")
.colors(["#CC79A7", "#009E73"])
.labels(x="", y="Anxiety score (z)")
)
qPCR fold change¶
Relative expression of four cytokines in wild-type and knockout macrophages (mean ± SD, n = 6). The reference line at 1 marks no change.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(13)
induction = {"Il6": 6.0, "Tnf": 4.0, "Il1b": 3.0, "Cxcl10": 8.0}
qpcr = pd.DataFrame(
[
{"gene": gene, "genotype": genotype,
"fold": fold * (0.4 if genotype == "KO" else 1) * rng.lognormal(0, 0.18)}
for gene, fold in induction.items()
for genotype in ("WT", "KO")
for _ in range(6)
]
)
(
sp.bar(qpcr, x="gene", y="fold", hue="genotype", error="sd")
.refline(1, linestyle=":")
.colors(["#4D4D4D", "#E69F00"])
.ticklabels(italic=True)
.labels(x="", y="Relative expression (2^−ΔΔCt)")
)
Pharmacokinetics¶
Plasma concentration after a single oral dose at two dose levels (mean ± SD, n = 10), on a log axis where first-order elimination is a straight line.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(17)
hours = np.array([0.25, 0.5, 1, 2, 4, 6, 8, 12, 24])
rows = []
for dose, scale in (("10 mg kg⁻¹", 1.0), ("30 mg kg⁻¹", 3.0)):
for subject in range(10):
ka, ke = rng.normal(1.6, 0.2), rng.normal(0.18, 0.02)
conc = scale * 900 * ka / (ka - ke) * (np.exp(-ke * hours) - np.exp(-ka * hours))
rows += [{"hours": h, "conc": c * rng.lognormal(0, 0.12), "dose": dose}
for h, c in zip(hours, conc)]
pk = pd.DataFrame(rows)
(
sp.line(pk, x="hours", y="conc", hue="dose", error="sd")
.errorbars("bars")
.markers()
.log("y")
.colors(["#56B4E9", "#0072B2"])
.labels(x="Time after dose (h)", y="Plasma concentration (ng mL⁻¹)")
)
Regression¶
Hepatic and muscle lipid content against BMI with one least-squares fit per tissue, its 95% confidence band, and the fitted equation with R².
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(29)
biopsies = pd.DataFrame({
"bmi": rng.uniform(19, 36, 140),
"tissue": rng.choice(["Liver", "Muscle"], 140),
})
offset = np.where(biopsies["tissue"] == "Liver", 6, 0)
slope = np.where(biopsies["tissue"] == "Liver", 0.82, 0.45)
biopsies["lipid"] = slope * biopsies["bmi"] + offset + rng.normal(0, 2.2, 140)
(
sp.scatter(biopsies, x="bmi", y="lipid", hue="tissue")
.fit("linear", stats=True)
.colors(["#AA3377", "#4477AA"])
.labels(x="BMI (kg m⁻²)", y="Lipid content (mg g⁻¹)")
)
Biomarker distribution¶
Serum biomarker levels in healthy donors and patients, normalised to densities so cohorts of different sizes compare fairly. Group sizes are added to the legend automatically.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(31)
serum = pd.DataFrame({
"cohort": np.repeat(["Healthy", "Patients"], [400, 250]),
"level": np.concatenate([rng.normal(1.2, 0.25, 400), rng.normal(1.75, 0.32, 250)]),
})
(
sp.histogram(serum, x="level", hue="cohort", stat="density")
.kde()
.show_n()
.colors(["#56B4E9", "#D55E00"])
.labels(x="Serum IL-6 (log₁₀ pg mL⁻¹)", y="Density")
)
Temperature anomaly¶
Annual global temperature anomaly relative to a 1951–1980 baseline. .split()
colours the area by sign, so warm and cool years read at a glance.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(1880)
years = np.arange(1880, 2025)
trend = -0.25 + 1.45 * ((years - 1880) / 144) ** 3
climate = pd.DataFrame({
"year": years,
"anomaly": trend + 0.12 * np.sin(years / 3.7) + rng.normal(0, 0.09, years.size),
})
(
sp.fill_between(climate, x="year", y1="anomaly", y2=0)
.split(colors=("#D6604D", "#4393C3"), labels=("Warmer", "Cooler"))
.labels(x="", y="Temperature anomaly (°C)")
)
Cell composition¶
Share of immune cell types in blood after stimulation. A 100% stacked area keeps the total fixed so shifts in composition are easy to follow.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(48)
hours = np.arange(0, 49, 4)
cell_types = {"T cells": (42, -0.35), "B cells": (24, -0.05),
"Monocytes": (19, 0.42), "NK cells": (15, -0.02)}
blood = pd.DataFrame(
[
{"hours": h, "cell_type": name,
"cells": max(1.0, base + slope * h + rng.normal(0, 0.8))}
for name, (base, slope) in cell_types.items()
for h in hours
]
)
(
sp.area(blood, x="hours", y="cells", hue="cell_type", stack="percent")
.endlabels()
.colors(["#332288", "#117733", "#882255", "#999933"])
.labels(x="Hours after stimulation", y="Share of leukocytes (%)")
)
Forecast fan chart¶
Nested 50%, 80% and 95% intervals from 500 simulated trajectories of weekly hospital admissions. Darker bands are more likely.
Show code
import numpy as np
import pandas as pd
import SecretPlots as sp
sp.use(preset="nature")
rng = np.random.default_rng(2024)
weeks = np.arange(1, 21)
growth = rng.normal(0.06, 0.035, (500, 1))
paths = 120 * np.exp(np.cumsum(growth + rng.normal(0, 0.05, (500, 20)), axis=1))
simulations = pd.DataFrame({
"week": np.tile(weeks, 500),
"admissions": paths.ravel(),
})
(
sp.line(simulations, x="week", y="admissions", error="pi", estimator="median")
.errorbars(level=(0.5, 0.8, 0.95))
.colors("#AA3377")
.labels(x="Week", y="Weekly admissions")
)
Want more?¶
Each plot type has a guide with every option illustrated: bar · scatter · line · area · histogram · heatmap · box · violin & raincloud. To prepare a figure for a specific journal, read journal-ready figures.