Systems and methods to detect rare mutations and copy number variation

What is this patent about?

’899 is related to the field of genetic analysis and bioinformatics, specifically focusing on the detection of rare mutations and copy number variations within cell-free polynucleotides. In clinical diagnostics, identifying low-frequency genetic aberrations—such as those shed by tumors into the bloodstream—is often hindered by the inherent noise and distortion introduced during PCR amplification and high-throughput sequencing. The background context involves the challenge of distinguishing true somatic variants from technical artifacts when the signal from diseased tissue is extremely dilute compared to healthy germline DNA.

The underlying idea behind ’899 is the use of a controlled tagging strategy to enable the reconstruction of original parent molecules from a pool of noisy sequencing reads. By applying a specific number of molecular barcodes relative to the expected number of duplicate fragments (those sharing identical start and stop positions), the system ensures that individual molecules can be uniquely identified and tracked. This allows the software to collapse multiple progeny reads into a single consensus sequence, effectively filtering out random errors and providing a high-fidelity digital representation of the initial genetic material.

The claims of ’899 focus on a method for detecting cancer by converting a sample of polynucleotides, such as cell-free DNA, into a population of tagged parent molecules using a defined range of barcodes. Specifically, the independent claims require tagging the molecules with n different barcodes, where n is at least 2 but no more than 10,000*z, with z representing the mean number of duplicate molecules having identical start and stop positions. This mathematical constraint on barcode diversity is central to the claimed process of amplifying and sequencing these tagged molecules to identify the presence or absence of malignancy.

In practice, the invention works by extracting cell-free DNA from a bodily fluid and attaching these molecular barcodes before any amplification occurs. By sampling the amplified progeny at a sufficient depth—often 5- to 10-fold coverage per parent molecule—the system can group reads into families derived from the same ancestor. This familial grouping allows the bioinformatics pipeline to assign confidence scores to base calls, ensuring that a mutation is only reported if it appears consistently across the family, rather than appearing as a one-off error from the sequencer.

This approach differs from prior methods by optimizing the conversion efficiency of the library preparation and using the relationship between fragment start/stop positions and barcode diversity to resolve molecular identity. While traditional sequencing might mask a 0.1% tumor signal within a 2% baseline error rate, this method reduces the effective error rate by several orders of magnitude. By focusing on the quantitative measure of unique families rather than raw read counts, the invention also provides a more accurate assessment of copy number variations, negating the effects of uneven amplification bias.