Bio Data Visualization Lollipop Protein Maps

Version Compatibility

Reference examples tested with: maftools 2.18+, trackViewer 1.38+, g3-lollipop (JavaScript via R g3viz 1.2+), Bio.PDB 1.83+ (for domain coordinates). ProteinPaint is a hosted service.

Before using code patterns, verify installed versions match. If versions differ:

• R: packageVersion('<pkg>') then ?function_name
• Python: pip show <package> then help(module.function)

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Lollipop / Needle Protein Maps

"Plot mutations on a gene's protein" -> Render a horizontal protein backbone with colored domain rectangles (from UniProt/Pfam/InterPro), then stack vertical lines ("stems") at mutated amino-acid positions, capped with circles ("lollipops") whose size reflects mutation count and whose color encodes variant class. The biological story is hotspot identification — a tall stack of recurrences at a single residue (e.g., KRAS G12, PIK3CA E545/H1047) is the visual signature of a driver mutation.

• R: maftools::lollipopPlot, trackViewer::lolliplot, g3viz::g3Lollipop
• Python: pyLollipop (limited maintenance); ProteinPaint via API
• Web: cBioPortal, ProteinPaint, MutationMapper

The Single Most Important Modern Insight -- Hotspot Recurrence Drives the Plot

A lollipop plot exists to identify hotspots — residues with disproportionate recurrence. The MutSig hotspot test (Lawrence 2014 Nature 505:495) and statisticalhotspot methods (Chang 2016 Nat Biotechnol 34:155) formalize this: a residue's mutation count should exceed the gene-wide background rate × residue count. Visualizing this on a domain map IS the diagnostic.

Key practical consequences:

• Stack height ≠ frequency: a tall lollipop at residue 600 means recurrence, not population frequency. Annotate the count.
• Domain colors should encode functional class (kinase, SH2, binding), not random hue.
• Mark known activating/inactivating residues (G12 for KRAS, R175 for TP53) with bold labels.

Decision Tree by Question

maftools::lollipopPlot

Goal: Render per-gene mutation distribution on Pfam domain map with count-sized lollipops and class-colored stems.

Approach: Pass MAF and gene to lollipopPlot; maftools queries Pfam for domain coordinates automatically; outputs ggplot2 object.

library(maftools)
maf <- read.maf(maf = 'cohort.maf')

# Default lollipop
lollipopPlot(maf = maf, gene = 'TP53',
             AACol = 'HGVSp_Short',
             labelPos = c(175, 248, 273),                   # mark canonical hotspots
             labPosSize = 1.0,
             showMutationRate = TRUE,
             domainLabelSize = 1,
             printCount = TRUE,
             colors = c(Missense_Mutation = '#D55E00',
                        Nonsense_Mutation = '#000000',
                        Frame_Shift_Del   = '#0072B2',
                        Frame_Shift_Ins   = '#56B4E9',
                        Splice_Site       = '#CC79A7',
                        In_Frame_Del      = '#009E73'))

# Compare two cohorts -- one up, one down
lollipopPlot2(m1 = cohort_a, m2 = cohort_b,
              gene = 'TP53',
              m1_name = 'Cohort A',
              m2_name = 'Cohort B',
              AACol1 = 'HGVSp_Short', AACol2 = 'HGVSp_Short',
              colors = my_palette)

trackViewer::lolliplot -- Fine Control over Track Layout

library(trackViewer)
library(GenomicRanges)

# Build SNP (lollipop) and feature (domain) GRanges
snps <- GRanges('chr17', IRanges(c(175, 248, 273), width = 1, names = c('R175H', 'R248Q', 'R273H')),
                color = c('#D55E00', '#D55E00', '#D55E00'),
                score = c(45, 38, 29))                       # mutation count
features <- GRanges('chr17',
                    IRanges(c(102, 323, 363), width = c(190, 30, 30),
                            names = c('DNA-binding', 'Tetramerization', 'Regulatory')),
                    fill = c('#0072B2', '#009E73', '#CC79A7'),
                    height = 0.04)

lolliplot(snps, features, ylab = 'Mutation count',
          xaxis = TRUE, yaxis = TRUE)

trackViewer is more flexible than maftools for non-standard layouts (custom domain sources, multi-protein stacking, integration with genome coordinates).

g3viz / g3-lollipop -- Interactive HTML

library(g3viz)
mutation_data <- hgvspChange2protein(maf, gene = 'TP53')
g3Lollipop(mutation_data,
           gene.symbol = 'TP53',
           protein.change.col = 'AA_Change',
           plot.options = g3Lollipop.theme(theme.name = 'nature'),
           output.filename = 'TP53_lollipop.html')

g3-lollipop produces an interactive HTML — hover tooltips, click-to-filter, exportable. Suitable for supplementary HTML supplement; not for journal figure submission directly.

Domain Annotation Sources

For canonical isoform: maftools uses the canonical UniProt isoform by default. For specific isoform: pass refSeqID or proteinID explicitly. Mutations annotated against a different isoform will be off-by-residue.

Per-Method Failure Modes

Mutations not labeled with AA position

Trigger: MAF column HGVSp_Short missing or malformed.

Mechanism: maftools expects HGVSp_Short (e.g., 'p.R175H'); falls back to other columns inconsistently.

Symptom: "No mutations to plot" or wrong positions.

Fix: Verify HGVSp_Short column exists; reformat from HGVSp if needed. Use AACol argument to specify which column.

Isoform mismatch

Trigger: Mutations called against ENST00000269305 but plotted against canonical ENST00000288602 (TP53).

Mechanism: Residue numbering differs across isoforms.

Symptom: Known R175H plotted at R177H or in a different domain.

Fix: Annotate the isoform in the figure caption; pass proteinID to lollipopPlot to force a specific isoform.

Domain map outdated

Trigger: maftools' cached Pfam annotation is older than the protein's current Pfam release.

Mechanism: Domain coordinates can shift across Pfam versions.

Symptom: Domain boundaries off by a few residues; published-figure mismatch.

Fix: Pull domain coordinates from UniProt directly (current); pass via trackViewer::lolliplot features.

Recurrence at low-coverage region overinterpreted

Trigger: "Hotspot" identified at a residue with high coverage variance — looks recurrent but is a sequencing artifact.

Mechanism: Capture-bait coverage variability; some residues sequenced more deeply.

Symptom: "Hotspot" in untargeted region; not validated in WGS.

Fix: Verify recurrence in independent cohort (TCGA Pan-Cancer + ICGC); use MutSig hotspot test (Lawrence 2014) for formal hotspot calling.