Methods and systems for detecting genetic variants

What is this patent about?

’221 is related to the field of genetic analysis and bioinformatics, specifically focusing on the detection of rare genetic variants such as copy number variations (CNV) and single nucleotide variants (SNV) within cell-free DNA (cfDNA) samples. In clinical diagnostics, identifying these mutations is often hampered by the low concentration of target molecules and the noise introduced during library preparation and sequencing. The technology addresses the need for high-sensitivity monitoring of disease states, such as cancer, by improving the accuracy of molecule counting and error correction in heterogeneous polynucleotide populations.

The underlying idea behind ’221 is that the true count of original DNA fragments in a sample can be mathematically inferred by tracking the recovery of complementary strands through the use of duplex tagging. By labeling the Watson and Crick strands of a double-stranded molecule so they can be distinguished after amplification, the system can categorize sequence reads into pairs (where both strands are recovered) and singlets (where only one is recovered). This distribution allows for a statistical estimation of the unseen molecules—those that were present in the original sample but failed to be sequenced—thereby correcting for sampling bias and improving the quantitative accuracy of genetic assays.

The claims of ’221 focus on a method for processing cfDNA by attaching duplex tags containing molecular barcodes to both ends of the fragments and subsequently generating consensus sequences. The independent claims specifically cover the reduction of redundancy by sorting sequence reads into paired and unpaired reads based on whether both complementary strands of an original parent molecule were successfully detected. Furthermore, the claims describe a non-unique tagging approach where the number of distinct barcodes is intentionally smaller than the number of DNA fragments at a specific locus, using the combination of barcodes and endogenous sequence information to uniquely identify parent molecules.

In practice, the invention works by ligating specialized adapters to cfDNA fragments at high conversion efficiencies, often exceeding 50%. These adapters contain degenerate or semi-degenerate barcodes that survive the amplification process, allowing a computer processor to group progeny reads into families derived from the same original molecule. By comparing the reads within these families, the system performs consensus calling, which filters out stochastic errors introduced by polymerases or sequencers. If a mutation appears in both the Watson and Crick strands of a paired read, the confidence that the variant is a true biological mutation rather than an artifact is significantly increased.

This approach differs from prior methods by moving beyond simple unique identification to a more sophisticated statistical reconstruction of the original sample composition. While traditional barcoding helps identify duplicates, ’221 uses the relationship between paired and unpaired strands to calculate a more precise total molecule count at specific genomic loci. This enables the detection of rare variants at concentrations below 1% with a specificity greater than 99.9%, providing a robust framework for monitoring tumor burden and identifying drug-resistance mutations in patients undergoing targeted therapy.