A noise measure for copy number analysis on targeted panel sequencing data

A method of reducing a statistical noise level in a calculation of a subject's genetic copy number value in massively parallel nucleic acid sequencing data derived from a sample, comprising:

• (a) obtaining massively parallel sequencing information of a subject's sample and of a cohort of normal sample for a defined set of genomic loci as nucleic acid sequence reads;
• (b) aligning said nucleic acid sequencing reads to a reference sequence;
• (c) determining the genetic copy number value using a median normalized ratio of sample read count data and normal read count data as obtained in step (a);
• (d) determining the ratio of a standard error value or the ratio of a standard deviation value to the genetic copy number value as determined in step (c) at each locus of said defined set of genomic loci;
• (e) adding the ratios as determined in step (d), optionally multiplying each ratio with a weight factor corresponding to the clinical relevance of the locus;
• (f) obtaining a value for the statistical noise level in the subject's genetic copy number wherein said value for the statistical noise level is the sum of added ratios as obtained in step (e);
• and
• (g) normalizing the value obtained in step (f) by the size of the defined set of genomic loci or by the sum of all applied weights for the defined set of genomic loci if different weights are applied,
• wherein said normal sample is selected from a heterogeneous cohort of normal samples, wherein said normal sample yields the lowest final noise measure among all normal samples of said cohort with respect to the subject's sample as analyzed, or wherein said normal sample is a sub-cohort selected from a heterogeneous cohort of normal samples, wherein said sub-cohort comprises those normal samples which yield the 2, 3, 4 or 5 lowest final noise measures among all normal samples of said heterogenous cohort with respect to the subject's sample as analyzed, and wherein a reduced value for the statistical noise level in the subject's genetic copy number is obtained.

A method for determining a subject's genetic copy number value in massively parallel nucleic acid sequencing data derived from a sample, comprising:

• (a) obtaining massively parallel sequencing information of a subject's sample and of a cohort of normal sample for a defined set of genomic loci as nucleic acid sequence reads;
• (b) aligning said nucleic acid sequencing reads to a reference sequence;
• (c) determining the genetic copy number value using a median normalized ratio of sample read count data and normal read count data as obtained in step (a);
• (d) determining the ratio of a standard error value or the ratio of a standard deviation value to the genetic copy number value as determined in step (c) at each locus of said defined set of genomic loci;
• (e) adding the ratios as determined in step (d), optionally multiplying each ratio with a weight factor corresponding to the clinical relevance of the locus;
• (f) obtaining a value for the statistical noise level in the subject's genetic copy number wherein said value for the statistical noise level is the sum of added ratios as obtained in step (e);
• (g) normalizing the value obtained in step (f) by the size of the defined set of genomic loci or by the sum of all applied weights for the defined set of genomic loci if different weights are applied;
• and
• (h) adjusting the determination of the subject's genetic copy number value by applying a value for the statistical noise level in the subject's genetic copy number.

A method to determine a subject's genetic copy number value for stratifying the subject for cancer therapy, comprising:

• (a) performing a massively parallel nucleic acid sequencing of nucleic acids extracted from a subject's tumor sample and from a cohort of normal sample for a defined set of genomic loci to obtain nucleic acid sequence reads;
• (b) aligning said nucleic acid sequencing reads to a reference sequence;
• (c) determining the genetic copy number value using a median normalized ratio of sample read count data and normal read count data as obtained in step (a);
• (d) determining the ratio of a standard error value or the ratio of a standard deviation value to the genetic copy number value as determined in step (c) at each locus of said defined set of genomic loci;
• (e) adding the ratios as determined in step (d), optionally multiplying each ratio with a weight factor corresponding to the clinical relevance of the locus;
• (f) obtaining a value for the statistical noise level in the subject's genetic copy number wherein said value for the statistical noise level is the sum of added ratios as obtained in step (e);
• and
• (g) normalizing the value obtained in step (f) by the size of the defined set of genomic loci or by the sum of all applied weights for the defined set of genomic loci if different weights are applied;
• and
• (h) adjusting the determination of the subject's genetic copy number value by applying a value for the statistical noise level in the subject's genetic copy number;
• and
• (i) attributing the determined subject's genetic copy number value to a group of increased, normal or decreased genetic copy number values, which can guide a treatment decision,
• wherein said adjusting of the determination of the subject's genetic copy number value in (h) comprises an exclusion from further usage of samples which have been identified as being higher than the threshold obtained according to a procedure comprising the steps of:
• (j) selecting a set of normal samples as a calibration test set, wherein all copy numbers are known to have a value of 2;
• (k) selecting a disjunct set of normal samples as a calibration reference set, wherein all copy numbers are known to have a value of 2;
• (l) analyzing each normal sample in the calibration test set to determine the copy number value of all loci of said defined set of genomic loci, the copy number standard deviation and the copy number standard error by using the calibration reference set of normal samples of step (k) for normalization;
• (m) calculating the noise measure value for each of the calibration test samples, wherein the calculation comprises:
• (i) determining the ratio of a standard error value or the ratio of a standard deviation value to the genetic copy number value as determined in step (l) at each locus of said defined set of genomic loci;
• (ii) adding the ratios as determined in step (m)(i), optionally multiplying each ratio with a weight factor corresponding to the clinical relevance of the locus;
• (iii) obtaining a value for the statistical noise level in the subject's genetic copy number wherein said value for the statistical noise level is the sum of added ratios as obtained in step (m)(ii);
• and
• (iv) normalizing the value obtained in step (m)(iii) by the size of the defined set of genomic loci or by the sum of all applied weights for the defined set of genomic loci if different weights are applied;
• (n) calculating a direct noise measure value of the copy number value of step (l) by forming a vector of deviations to a copy number value 2, wherein each component of the vector is the deviation of the copy number value of a single locus as determined in step (l) from the copy number value 2;
• (o) calculating a p-norm of the vector obtained in step (n) for values of p=1 and/or p->00 (maximum norm);
• (p) for the case of p=1 said vector is normalized by the size of the defined set of genomic loci to obtain a mean of absolute deviation to copy number 2 for all genetic copy numbers in said defined set of genomic loci;
• (q) plotting the noise measure of each calibration test sample against the mean absolute deviation to copy number 2 or maximum deviation to copy number 2 and observing correlations for both plots;
• and
• (r) using a significance criterion for a threshold in terms of the mean absolute deviation to 2 or maximum deviation to 2 and determining to which value of noise measure it translates,
• wherein said adjusting of the determination of the subject's genetic copy number value in (h) further comprises a modification of the statistical significance of a calculated genetic copy number subject to the value for the statistical noise level in a subject's genetic copy number.