Genome encodes functional and dysfunctional information not only in its main chapters of linear sequences and primary chemical modifications such as methylations, also in its dynamic multi-dimensional structures, organizations, functional derivatives in forms of aberrant orders, binding and interactions, loops and folds, locations and shapes.
The genome we inherit and the genome we live by -
The germline genome is a historical document, written across thousands of generations, carrying the migration and adaptation of the ancestors we never had the chance to meet. It is written before we take our first breath - that is not even our doing, just as our parents were not our choice. Everything that follows is a journey - and somatic disease happens along that journey.
For twenty years, we have progressed to be good at answering one question: " what does this person's inherited genome say? "
That question has a simpler structure. One genome, shared by almost every cell, in theory. Averaging across millions of cells is the correct method. Deep short- and long-read sequencing plus mapping answers it with remarkable approximation and falling cost, and the informatics built on top of it is some of the best engineering in modern biology.
Disease along the journey after birth is a different question.
A tumor or an aging body is not one genome. It is thousands or millions of genomes, diverging in real time — aneuploid, poly-aneuploid, restructured by chromothripsis and chromoplexy, extrachromosomal DNA (ecDNA). Aneuploidy is a hallmark of cancer, present in up to 90% of solid tumors. Poly-aneuploid cancer cells drive therapy resistance and progression.
Under pressure, the genome taps into a mechanisms older than humans: DNA detaches from chromosomes, forms loops, and turns copy numbers into adjustable knobs. Bacteria have operated this way for billions of years as plasmids. Cancer didn't invent this mechanism, only providing pressure to make it profitable. In preliminary research screening, ecDNA is reported in a large fraction of the most aggressive cancers (17%-60%), including glioblastoma and small cell lung cancer.
Due to lack of centromeres, ecDNA are randomly segregated to offspring during each cycle of cell division, varies from zero to hundreds of copies between neighboring cells, causing the tumor's composition to be continuously rewritten. They are evolving multi-specie mini-cancer factories carrying fully functional oncogenes and randomly recruited regulatory elements from different sections of the genome, unlike tiny circulating DNA fragments in blood - they are large random page-recombined "rogue" chapters of the book of life, rewriting the narratives of the fate, driving about one-third of human cancers with therapy resistance, relapses, aggressiveness, immuno-escape and worse survival rate.
In germline genomics, the consensus is the answer. In somatic genomics, the consensus is a genome that no cell actually has. In that setting, averaging stops being compression and becomes destruction.
This is not a failure of sequencing. It's a mismatch between an input and a question. Standard preparation — mixing, sonication, fragmentation, stripping of structural context — dissolves exactly the information that defines a somatic genome: which pieces were attached to which, what was linear or circular, what belonged to the same strand, compartment, nucleus, cell, how do they interact with each other to drive different functions. No amount of sequencing depth recovers it, because the information was gone before the instrument saw the sample.
Yet, the cancer and aging cells' "hidden" genome - a powerful layer of cancer biology, such an important biomarker is largely missing in today's testing, limited by the current technology. Conventional sequencing workflow cannot truly reveal them easily and completely. Because ecDNA has same sequence source from the host genome, the proportion of ecDNA accounts for less than 0.01%-2% of total genomic mass in a cell. In mixed samples, structurally different ecDNA ring sequences, often compressed into an averaged consensus amplicon - and those emerging species of less abundance might be precisely the ones that survive the treatment to cause recurrence.
This is where we come in. Before building the library, Dimension Genomics physically separates extrachromosomal DNA from chromosomal DNA based on size and conformation. This method does not require prior knowledge of what it is looking for—so it is a "discovery" rather than a "confirmation." What is obtained is not a consensus amplicon but a parsed lineage: which rings exist, what abundances each, and which novel junctions are carried.
This same principle applies to all spontaneously emerging aberrant components of the somatic genome.