Granted Patent B2

US 12,706,180 · App. 19/251,587 · Granted Aug 11, 2026

Inventor: Alexander Halley Stram (Altadena, CA)

Assignee: Personalis, Inc.

G16B50/50G06F16/174G06N3/04G16B20/00G16B30/10G16B50/20

Abstract

Claims (44)

1 . A method of regenerating a FASTQ file from a compressed sequence alignment map file, the compressed sequence alignment map file including alignment data for a plurality of sequence strings for more than one aligned reads of clusters on a flow cell, the method comprising:

(a) determining an fqsum of an original FASTQ file on which the compressed sequence alignment file is based, wherein each entry of the original FASTQ file is defined by at least four line-separated fields per sequence and the fqsum represents an order invariant checksum of the original FASTQ file;

(b) storing a quality score of each base call for individual instances of the plurality of sequence strings;

(c) sorting the compressed sequence alignment map file by one or more reference genome coordinates;

(d) extracting primary alignments from the compressed sequence alignment map file;

(e) ensuring all FASTQ records from the original FASTQ file are stored in the compressed SAM file;

(f) reconstructing original base quality scores by at least one of retrieving the stored quality score or applying an inverse of a model of the stored quality score;

(g) writing at least three of the four line-separated fields per sequence of the compressed sequence alignment map file to a regenerated FASTQ file; and

(h) compressing the regenerated FASTQ file.

2 . The method of claim 1 , further comprising comparing the determined fqsum with an fqsum of the regenerated FASTQ file.

3 . The method of claim 1 , wherein said determining the fqsum of the original FASTQ file further comprises: for a set A and a set B of the original FASTQ file, fqsum ({A})=fqsum (fqsum ({A})), fqsum ({A, B})=fqsum ({B, A}), and fqsum({fqsum ({A, B}), C})=fqsum ({A, fqsum({B, C})}).

4 . The method of claim 1 , wherein said storing the quality score comprises generating a base quality score recalibration model.

5 . The method of claim 1 , wherein said storing the quality score of each base call for individual instances of the plurality of sequence strings comprises storing a Phred scaled score.

6 . The method of claim 5 , wherein said storing the quality score of each base call for individual instances of the plurality of sequence strings comprises storing a string of integers, equal in length to a selected sequence string of the plurality of sequence strings.

7 . The method of claim 1 , wherein said ensuring all FASTQ records from the original FASTQ file are stored in the compressed SAM file further comprises: (i) removing sequencing duplicates and off-target reads, and (ii) storing the sequencing duplicates and off-target reads in an auxiliary compressed sequence alignment map file.

8 . The method of claim 1 , wherein the at least four line-separated fields per sequence of each entry of the original FASTQ file comprises: (i) a sequence identifier field, (ii) a sequence field, (iii) a quality score identifier field, and (iv) a quality score field.

9 . The method of claim 1 , wherein the compressed sequence alignment map file comprises at least one of a binary alignment map file and a compressed reference-oriented map file.

10 . The method of claim 1 , wherein a file size of the compressed sequence alignment map file is smaller than a file size of either of the original FASTQ file and the regenerated FASTQ file.

11 . A method of regenerating a FASTQ file from a compressed sequence alignment map file, the compressed sequence alignment map file including alignment data for a plurality of sequence strings for more than one aligned reads of clusters on a flow cell, the method comprising:

(a) determining an fqsum of an original FASTQ file on which the compressed sequence alignment file is based, wherein each entry of the original FASTQ file is defined by at least four line-separated fields per sequence and the fqsum represents an order invariant checksum of the original FASTQ file;

(b) storing a quality score of each base call for individual instances of the plurality of sequence strings;

(c) sorting the compressed sequence alignment map file by sequence string;

(d) extracting primary alignments from the compressed sequence alignment map file;

(e) ensuring all FASTQ records from the original FASTQ file are stored in the compressed SAM file;

(f) reconstructing original base quality scores by at least one of retrieving the stored quality score or applying an inverse of a model of the stored quality score;

(g) writing at least three of the four line-separated fields per sequence of the compressed sequence alignment map file to a regenerated FASTQ file; and

(h) compressing the regenerated FASTQ file.

12 . The method of claim 11 , further comprising comparing the determined fqsum with an fqsum of the regenerated FASTQ file.

13 . The method of claim 11 , wherein said determining the fqsum of the original FASTQ file further comprises: for a set A and a set B of the original FASTQ file, fqsum ({A})=fqsum (fqsum ({A})), fqsum ({A, B})=fqsum ({B, A}), and fqsum({fqsum ({A, B}), C})=fqsum ({A, fqsum({B, C})}).

14 . The method of claim 11 , wherein said storing the quality score comprises generating a base quality score recalibration model.

15 . The method of claim 11 , wherein said storing the quality score of each base call for individual instances of the plurality of sequence strings comprises storing a Phred scaled score.

16 . The method of claim 15 , wherein said storing the quality score of each base call for individual instances of the plurality of sequence strings comprises storing a string of integers, equal in length to a selected sequence string of the plurality of sequence strings.

17 . The method of claim 11 , wherein the at least four line-separated fields per sequence of each entry of the original FASTQ file comprises: (i) a sequence identifier field, (ii) a sequence field, (iii) a quality score identifier field, and (iv) a quality score field.

18 . The method of claim 10 , wherein the compressed sequence alignment map file comprises at least one of a binary alignment map file and a compressed reference-oriented map file.

19 . The method of claim 11 , wherein a file size of the compressed sequence alignment map file is smaller than a file size of either of the original FASTQ file and the regenerated FASTQ file.

20 . A computer program product for regenerating a FASTQ file from a compressed sequence alignment map file, the compressed sequence alignment map file including alignment data for a plurality of sequence strings for more than one aligned reads of clusters on a flow cell, the computer program product being encoded on one or more machine-readable storage media and comprising instructions for:

(i) determining an fqsum of an original FASTQ file on which the compressed sequence alignment file is based, wherein each entry of the original FASTQ file is defined by at least four line-separated fields per sequence and the fqsum represents an order invariant checksum of the original FASTQ file;

(ii) storing a quality score of each base call for individual instances of the plurality of sequence strings;

(iii) sorting the compressed sequence alignment map file by one or more reference genome coordinates;

(iv) extracting primary alignments from the compressed sequence alignment map file;

(v) ensuring all FASTQ records from the original FASTQ file are stored in the compressed SAM file;

(vi) reconstructing original base quality scores by at least one of retrieving the stored quality score or applying an inverse of a model of the stored quality score;

(vii) writing at least three of the four line-separated fields per sequence of the compressed sequence alignment map file to a regenerated FASTQ file; and

(viii) compressing the regenerated FASTQ file.

Assignments (1)

Continuity (2)

References Cited (400)