A circular grammar for whole-genome data: stacked tracks of mutations, copy number, and methylation arc around a banded ideogram, with ribbons drawing the rearrangements between distant loci.
A genome browser shows one chromosome at a time. Many biological events span larger contexts: the Philadelphia translocation creates a derivative chromosome 22 carrying BCR::ABL1 in leukemia; structural variants can connect distant loci; blocks of synteny were shuffled between mouse and human over roughly 90 million years.
The circos plot, designed by Martin Krzywinski in 2009, gives every chromosome equal visual access to every other. Chromosomes wrap the circle as banded arcs. Concentric tracks layer data inside the ring (gene density, mutations, copy number, methylation). Ribbons across the middle connect distant loci, encoding where breakpoints fall and how much sequence was rearranged.
Below is a simulated tumor genome. All 24 chromosomes form the outer ring, rounded from GRCh38 chromosome lengths and banded with a low-resolution G-stain ideogram. Four data tracks fill the inner rings: selected oncogenes, mutation density, copy-number log2 ratio (red gain, blue loss), and methylation. Ribbons mark structural variants. One ribbon, labeled in red, is the BCR::ABL1 t(9;22)(q34;q11.2) translocation associated with the Philadelphia chromosome.
Figure 1. A simulated tumor genome. Rings (outer-in): cytogenetic ideogram with 25-Mb axis ticks, gene labels for selected oncogenes, mutation-density histogram, copy-number scatter, methylation heatmap. Ribbons across the middle are structural variants, colored by type and sized by the length of the rearranged segment.
That figure folds together several views a clinical-genomics team might inspect separately: karyotype, copy-number profile, mutation map, methylation summary, and a list of structural variants. The point is not exact coordinate lookup. It is seeing whether the genome has a quiet background, a few local events, or many long-range breakpoints.
Strip away the tracks and ribbons and you're left with the ideogram: a ring of arcs, one per chromosome, sized by length in megabases. Inside each arc are cytobands, the alternating dark and light Giemsa-stained bands used in cytogenetics. Those bands are approximate landmarks, historically tied to microscopy and FISH mapping rather than base-pair-perfect sequence features. The dark red band near the middle is the centromere, the boundary between the short p arm and long q arm.
Figure 2. The bare scaffold: 24 human chromosomes (22 autosomes + X + Y) with low-resolution cytobands anchored to GRCh38 chromosome lengths and centromere positions. The outer tick ring marks 25-Mb intervals, labeled every 50 Mb. Hover any chromosome for a mini-karyogram in the center: cytobands, p/q arms, and pinned oncogenes.
The circular layout gives every chromosome the same radial budget and allows concentric data tracks without extending the figure horizontally.
Inside the ideogram are data tracks. Circos defines five primitives:
The hero figure used four of these (tile, histogram, scatter, heatmap). All five are stacked below; toggle them to see how each ring contributes. Track order is convention: schematic annotations outside, quantitative signals inside.
Figure 3. Five track types stacked on the ideogram. Toggle any combination — the rings rebuild from outer to inner in the order you've enabled them.
The interior of the disc is reserved for links, connections between two genomic positions. Drawn as thin chords, links show only that two loci are related; drawn as ribbons, the width of each end encodes the size of the rearranged segment, and the color encodes the type. The same set of structural variants is drawn both ways below.
Figure 4. Same structural variants drawn two ways. Thin links treat every event equally; ribbons make a 50-Mb inversion impossible to miss next to a 1-Mb duplication.
The BCR::ABL1 ribbon from the hero figure has its own history. In 1960, Peter Nowell and David Hungerford described an unusually small chromosome in chronic myeloid leukemia. In 1973, Janet Rowley showed that it was not a simple deletion: material had been reciprocally exchanged between chromosomes 9 and 22. The derivative chromosome 22 carries BCR::ABL1, an oncogenic tyrosine kinase. Imatinib, designed to inhibit that kinase, helped turn CML into a manageable chronic disease for many patients.
To make a single event legible, give it the whole circle. Below: chromosomes 9 and 22 alone, ABL1 and BCR labeled near their cytogenetic loci, and the t(9;22) translocation as a single thick ribbon. Copy number is shown for context.
Figure 5. The Philadelphia chromosome as a Circos zoom. Restricting the layout to the two chromosomes involved gives the breakpoints, chr9q34 for ABL1 and chr22q11.2 for BCR, room to breathe.
Circos was originally built for comparative genomics. The same ribbon idiom that connects breakpoints in one genome can connect conserved blocks between two. Below: the first six human chromosomes on the right, the first six mouse chromosomes on the left, with ribbons for blocks preserved across ~90 million years of mammalian evolution.
Where ribbons land in the same rank order, evolution left the ancestral order alone. Where they tangle, inversions and translocations rearranged the genome. Color follows the source human chromosome.
Figure 6. Simulated human–mouse synteny. Each ribbon is a conserved block; ribbon thickness encodes block size. The same visual primitive — ribbon = pair of intervals + width — serves cancer rearrangements and species comparison alike.
Circos plots work when the pattern is the message: how many connections there are, where they cluster, and which chromosomes behave like hubs. They fail at precise quantitative reading. Comparing arc lengths is difficult; comparing bar heights on a Cartesian plot is easy.
Use a circos plot for genome-wide connections at a glance or comparative-genomics overviews. Use a genome browser or Manhattan plot for precise coordinates and accurate values.
With a map of genomic alterations in hand, the next question is clinical: do these changes predict patient outcomes? See survival curves.