Back to results
US 2026/0311802 A1Published application
System and method for dual-threshold classification and refinement in streamed measurement datasets
AssigneesGuardant Health, Inc., Dana-Farber Cancer Institute, Inc.
InventorsRichard B. LANMAN, Geoffrey R. OXNARD
Life of this patent
Estimating the term…
Abstract
A computer-implemented method for classifying measurement data using a dual-threshold filter is disclosed. The method includes receiving streamed input data, applying first and second threshold filters to assign classification states to individual data points, generating a classification mapping structure, and refining intermediate classifications using dynamic input updates and reference dataset comparisons. Classified data is output in a structured format for downstream indexing and retrieval. The system supports adaptive thresholding and high-throughput classification workflows in distributed environments.
Claims30
A computer-implemented method for classifying genomic loci in cell-free DNA (cfDNA) sequencing data, executed by a processor in a computer system, the method comprising:
- receiving cfDNA sequence reads via a communication interface and storing the cfDNA sequence reads as cfDNA sequencing data in memory;
- for each genomic locus, determining a copy-number-variation (CNV) status by computing a standard deviation (STDEV) of allele-fraction (AF) values within a heterozygous band for the subject and comparing the STDEV to a predetermined CNV threshold;
- performing an initial binning of genomic loci having a CNV status less than the predetermined CNV threshold into (i) presumed somatic origin, (ii) presumed germline origin, or (iii) indeterminate origin, based on comparison with reference variant datasets comprising known germline and known somatic variants;
- generating a first frequency distribution of AF values from the genomic loci of presumed somatic origin and a second frequency distribution of AF values from the genomic loci of presumed germline origin;
- deriving a first threshold from the first frequency distribution and a second threshold from the second frequency distribution;
- applying a dual-threshold genomic filter to the cfDNA sequencing data for the genomic loci having a CNV status less than the predetermined CNV threshold, wherein,
- a first threshold filter is applied to the cfDNA sequence reads, classifying those with allele fraction (AF) values less than or equal to the first threshold as somatic,
- a second threshold filter is applied to the cfDNA sequence reads, classifying those with AF values greater than or equal to the second threshold as germline, and
- cfDNA sequence reads with AF values between the first and second thresholds are retained for further analysis;
- generating a classification matrix that assigns each genomic locus a classification state corresponding to its somatic, germline, or ambiguous status, wherein genomic loci having a CNV status greater than or equal to the predetermined CNV threshold are assigned an ambiguous classification state without applying the dual-threshold genomic filter;
- refining ambiguous classifications by,
- reassessing AF values in relation to one or more updated sequencing datasets and
- comparing ambiguous loci to reference variant datasets stored in a classification database;
- and
- outputting a binary genomic classification dataset, wherein classified loci are stored in an indexed genomic database for subsequent medical diagnostics.
12 more elements
2-9. (canceled)
The computer-implemented method of claim 7, wherein generating the first distribution of somatic variants is based at least in part on a first bin of genomic loci having presumed somatic origin.
The computer-implemented method of claim 7, wherein generating the second distribution of germline variants is based at least in part on a second bin of genomic loci having presumed germline origin.
One or more non-transitory computer-readable media storing processor-executable instructions that, when executed by at least one processor, cause the at least one processor to:
- receive cfDNA sequence reads via a communication interface and storing the cfDNA sequence reads as cfDNA sequencing data in memory;
- for each genomic locus, determine a copy-number-variation (CNV) status by computing a standard deviation (STDEV) of allele-fraction (AF) values within a heterozygous band for the subject and comparing the STDEV to a predetermined CNV threshold;
- perform an initial binning of genomic loci having a CNV status less than the predetermined CNV threshold into (i) presumed somatic origin, (ii) presumed germline origin, or (iii) indeterminate origin, based on comparison with reference variant datasets comprising known germline and known somatic variants;
- generate a first frequency distribution of AF values from the genomic loci of presumed somatic origin and a second frequency distribution of AF values from the genomic loci of presumed germline origin;
- derive a first threshold from the first frequency distribution and a second threshold from the second frequency distribution;
- apply a dual-threshold genomic filter to the cfDNA sequencing data for the genomic loci having a CNV status less than the predetermined CNV threshold, wherein,
- a first threshold filter is applied to the cfDNA sequence reads, classifying those with allele fraction (AF) values less than or equal to the first threshold as somatic,
- a second threshold filter is applied to the cfDNA sequence reads, classifying those with AF values greater than or equal to the second threshold as germline, and
- cfDNA sequence reads with AF values between the first and second thresholds are retained for further analysis;
- generate a classification matrix that assigns each genomic locus a classification state corresponding to its somatic, germline, or ambiguous status, wherein genomic loci having a CNV status greater than or equal to the predetermined CNV threshold are assigned an ambiguous classification state without applying the dual-threshold genomic filter;
- refine ambiguous classifications by,
- reassessing AF values in relation to updated sequencing datasets and
- comparing ambiguous loci to reference variant datasets stored in a classification database;
- and
- output a binary genomic classification dataset, wherein classified loci are stored in an indexed genomic database for subsequent medical diagnostics.
12 more elements
An apparatus comprising:
- one or more processors;
- and
- a memory storing processor-executable instructions that, when executed by the one or more processors, cause the apparatus to:
- receive cfDNA sequence reads via a communication interface and storing the cfDNA sequence reads as cfDNA sequencing data in memory;
- for each genomic locus, determine a copy-number-variation (CNV) status by computing a standard deviation (STDEV) of allele-fraction (AF) values within a heterozygous band for the subject and comparing the STDEV to a predetermined CNV threshold;
- perform an initial binning of genomic loci having a CNV status less than the predetermined CNV threshold into (i) presumed somatic origin, (ii) presumed germline origin, or (iii) indeterminate origin, based on comparison with reference variant datasets comprising known germline and known somatic variants;
- generate a first frequency distribution of AF values from the genomic loci of presumed somatic origin and a second frequency distribution of AF values from the genomic loci of presumed germline origin;
- derive a first threshold from the first frequency distribution and a second threshold from the second frequency distribution;
- apply a dual-threshold genomic filter to the cfDNA sequencing data for the genomic loci having a CNV status less than the predetermined CNV threshold, wherein,
- a first threshold filter is applied to the cfDNA sequence reads, classifying those with allele fraction (AF) values less than or equal to the first threshold as somatic,
- a second threshold filter is applied to the cfDNA sequence reads, classifying those with AF values greater than or equal to the second threshold as germline, and
- cfDNA sequence reads with AF values between the first and second thresholds are retained for further analysis;
- generate a classification matrix that assigns each genomic locus a classification state corresponding to its somatic, germline, or ambiguous status, wherein genomic loci having a CNV status greater than or equal to the predetermined CNV threshold are assigned an ambiguous classification state without applying the dual-threshold genomic filter;
- refine ambiguous classifications by,
- reassessing AF values in relation to updated sequencing datasets and
- comparing ambiguous loci to reference variant datasets stored in a classification database;
- and
- output a binary genomic classification dataset, wherein classified loci are stored in an indexed genomic database for subsequent medical diagnostics.
15 more elements
21-22. (canceled)
Dependencies are inferred from the claim text.