Most people picture a cancer test as something that happens to a tissue sample under a microscope. A liquid biopsy is the alternative pitch: draw blood, isolate the tiny fragments of DNA that tumors shed into the bloodstream, and read them for the mutations that identify and track a cancer. Doing that reliably is less a single invention than a chain of engineering problems — getting clean DNA out of plasma, labeling each molecule so you can tell signal from error, and computing which faint variants are real. Guardant Health's patent record maps that chain step by step, and the week of 21 April 2026 re-anchored the tile at the center of it.

On 21 April 2026, the U.S. patent office granted US12606874B2, "Systems and methods to detect rare mutations and copy number variation." This is the family that has run through Guardant's estate for years — earlier members issued as far back as 2017 — and the newest grant re-states the same broad workflow: isolate cell-free DNA from a bodily fluid, sequence it, and apply bioinformatics to call rare mutations and copy-number changes against a reference. The abstract lays out the scope in plain terms:

The present disclosure provides a system and method for the detection of rare mutations and copy number variations in cell free polynucleotides.— Systems and methods to detect rare mutations and copy number variation, US12606874B2

The CPC classifications cluster in C12Q 1/6886 (nucleic-acid testing for cancer), C12Q 1/6869 (sequencing methods), and the G16B bioinformatics codes for sequence analysis — the signature of a record that binds the wet-lab assay to the software that interprets it. For a business reader, the relevant fact is that this family has been continuously refiled and re-granted: the April 2026 grant is the latest issued member of a lineage that also includes US12624400B2, granted weeks earlier on 12 May 2026, and older grants such as US10501810B2 from 2019. A company that keeps issuing patents in the same family is extending enforceable coverage over its core method across overlapping term windows.

The footprint traces the whole assay

What turns a single grant into a coverage map is the surrounding estate, and Guardant's runs to more than 140 U.S. grants concentrated in the same CPC neighborhood. The tiles fill in each step of the workflow. On sample handling, US12630882B2 covers compositions and methods for isolating cell-free DNA, including capturing it with a "sequence-variable target region set and an epigenetic target region set" sequenced to different depths — coverage on the front-end chemistry that gets usable DNA out of plasma. On error suppression, US9920366B2 covers tagging double-stranded fragments with duplex tags so that variant calls can be confirmed across both complementary strands — the molecular-barcoding idea that lets the assay distinguish a true low-frequency mutation from a sequencing artifact.

The newest grant in the estate, US12655469B2 (issued 16 June 2026), covers methods for analyzing cell-free DNA in "methylation partitioning assays" — partitioning a sample by the proportion of modified cytosines, which is the chemistry behind reading a tumor's epigenetic signature rather than only its mutations. Upstream of that, US9850523B1 covers multi-resolution analysis using paired bait sets that enrich nucleosome-associated regions, coverage on how the assay selectively captures the genomic regions that reveal a cell's tissue of origin. Detection method, cfDNA isolation, duplex tagging, methylation partitioning, bait enrichment: the issued tiles trace the path of a blood sample from plasma to a called variant.

The chronology of those grants is itself part of the map. The filing dates behind them run from the mid-2010s — when the foundational rare-mutation and bait-set work was first prosecuted — through to grants issued across 2025 and into mid-2026, including US12630882B2 in May 2026 on cfDNA isolation and the June 2026 methylation-partitioning grant. By volume, the estate's grant activity rose steadily through that period, with more than a hundred of the company's grants issuing from 2020 onward according to the year-by-year record. For a general reader, the takeaway is not that any one patent is decisive but that the issued coverage has been added in layers over a decade, each new grant attaching to a different stage of the same assay rather than re-covering the same ground.

What the workflow coverage buys

The business read on a coverage map like this is about where the freedom-to-operate pressure sits. A granted claim is the enforceable part of an estate, and Guardant's grants concentrate on the steps any blood-based-testing company has to perform — extract the DNA, label the molecules, suppress the noise, enrich the informative regions, and compute the variant calls. Coverage distributed across those steps is a different kind of position from a single product patent: it can put a competitor's workflow within reach of an issued claim even where the competitor's instrument and software are its own. That is the practical meaning of an estate built around a process, and it shows up commercially in licensing talks, in the design choices rivals make to engineer around issued method claims, and in the way diagnostics firms describe their patent positions to partners.

The usual caveats apply to any coverage read. Issued claims describe what was granted, not how broadly a court would construe a method step or whether a given claim survives a validity challenge; sequencing and bioinformatics methods in particular sit in a crowded prior-art field. The continuously refiled "rare mutations" family also means multiple grants describe closely related scope on staggered term clocks, which is a feature of the strategy rather than a contradiction. But the pattern across Guardant's record is consistent: the company holds issued claims spread across the entire liquid-biopsy chain, from the blood draw to the variant call. The 21 April 2026 grant re-anchors the detection method at the center of that map.