Regional plots

GWASLab provides functions for creating regional plots.

.plot_mqq(mode="r")

mysumstats.plot_mqq(mode="r",
          region = None,
          ...
          ):

GWASLab regional plot function is based on plot_mqq(). Most options are largely the same as Manhattan plot.

Options

mysumstats.plot_mqq(
    mode="r",
    region=(7, 156938803, 157038803),
    vcf_path="path/to/reference.vcf.gz",
    ld_link=True,
    ld_link_alpha_scale=0.2,
    ld_link_linewidth=1.0,
    ld_link_sig_level=5e-8  # Only show lines for significant variants
)

Examples

gl.plot_stacked_mqq()

Creates stacked Manhattan-QQ plots or regional plots for multiple GWAS datasets, allowing side-by-side comparison of multiple studies or traits.

gl.plot_stacked_mqq(objects, **kwargs)

Parameters

Required Parameters

Plot Mode Parameters

Layout Parameters

Regional Plot Specific Parameters

Styling Parameters

Output Parameters

Additional Parameters

All parameters from plot_mqq() can be passed via **mqq_kwargs to customize individual panels. These include:

• highlight, anno_set, pinpoint for variant annotation
• colors, scatter_kwargs for styling
• sig_line, suggestive_sig_line for significance lines
• region_ld_threshold, region_ld_colors for LD coloring
• And many more (see Manhattan plot documentation)

Plot Modes

Regional Mode (mode="r")

Creates stacked regional plots with LD information and gene tracks. Each panel shows a regional plot for a specific genomic region. Requires:

• region parameter specifying the genomic region
• vcfs parameter for LD reference (optional but recommended)

Manhattan Mode (mode="m")

Creates stacked Manhattan plots across the genome. All panels share the same x-axis (genomic position). Useful for comparing multiple traits or studies genome-wide.

Manhattan-QQ Mode (mode="mqq")

Creates stacked panels with Manhattan plot on the left and QQ plot on the right for each dataset. Useful for quality control and comparing multiple studies.

Panel Types

The function automatically detects panel types based on input data:

• Manhattan panels (pm="m"): For DataFrames with P or MLOG10P columns
• Credible set panels (pm="pip"): For DataFrames with PIP column (e.g., from fine-mapping)

Examples

Stacked regional plots for multiple studies:

import gwaslab as gl

# Load multiple sumstats
sumstats1 = gl.Sumstats("study1.txt.gz")
sumstats2 = gl.Sumstats("study2.txt.gz")

# Create stacked regional plot
gl.plot_stacked_mqq(
    objects=[sumstats1, sumstats2],
    mode="r",
    region=(7, 156538803, 157538803),
    vcfs=["ld_ref.vcf.gz"],  # Single VCF used for all panels
    titles=["Study 1", "Study 2"],
    build="38"
)

Stacked Manhattan plots:

# Compare multiple traits genome-wide
gl.plot_stacked_mqq(
    objects=[trait1_sumstats, trait2_sumstats, trait3_sumstats],
    mode="m",
    titles=["Trait 1", "Trait 2", "Trait 3"],
    colors=["#1f77b4", "#ff7f0e", "#2ca02c"],
    sig_line=True,
    sig_level_plot=5e-8
)

Stacked Manhattan-QQ plots:

# Quality control comparison
gl.plot_stacked_mqq(
    objects=[study1, study2],
    mode="mqq",
    titles=["Study 1", "Study 2"],
    mqqratio=3
)

Regional plot with credible sets:

# Include fine-mapping results
mysumstats = gl.Sumstats("gwas.txt.gz")
finemap_results = pd.read_csv("finemap_results.txt")  # Contains PIP column

gl.plot_stacked_mqq(
    objects=[mysumstats, finemap_results],
    mode="r",
    region=(7, 156538803, 157538803),
    vcfs=["ld_ref.vcf.gz"],
    titles=["GWAS", "Fine-mapping"],
    build="38"
)

Custom styling per panel:

# Different colors and styles for each panel
gl.plot_stacked_mqq(
    objects=[sumstats1, sumstats2],
    mode="m",
    titles=["Panel 1", "Panel 2"],
    colors=[["#1f77b4"], ["#ff7f0e"]],  # Different colors per panel
    scatter_kwargs=[{"s": 10}, {"s": 20}]  # Different sizes per panel
)

Notes

• When mode="r" (regional), the number of VCF files must match the number of objects, or a single VCF can be provided which will be used for all panels.
• For credible set panels (pm="pip"), VCF files are automatically set to "NA" as LD information is not applicable.
• The function automatically detects panel types based on column names (P/MLOG10P for Manhattan, PIP for credible sets).
• All parameters from plot_mqq() can be passed to customize individual panels. Use lists to specify different values for each panel.

Notes and Troubleshooting

Gene track limitations:

The gene track only displays protein-coding genes from the reference GTF files. Non-coding genes, pseudogenes, and other gene types are excluded from the visualization.

Missing exons in gene track:

Very short exons may not appear in the gene track if they are too small to render at the current resolution. To display these exons, you can either increase the figure DPI (dots per inch) when saving the plot or reduce the length of the genomic region being plotted.

LD calculation errors:

LD (linkage disequilibrium) cannot be calculated when the reference variant in the VCF file is mono-allelic (i.e., has only one allele present in the reference panel). This will result in an error even if both variants are present in the reference VCF file. Ensure your reference VCF contains biallelic variants for accurate LD calculation.