U.S. patent application number 10/502761 was filed with the patent office on 2005-12-22 for methods of validating snps and compiling libraries of assays.
Invention is credited to De La Vega, Francisco, Isaac, Hadar I., Scafe, Charles R., Spier, Eugene G., Ziegle, Janet S..
Application Number | 20050282162 10/502761 |
Document ID | / |
Family ID | 27671405 |
Filed Date | 2005-12-22 |
United States Patent
Application |
20050282162 |
Kind Code |
A1 |
De La Vega, Francisco ; et
al. |
December 22, 2005 |
Methods of validating snps and compiling libraries of assays
Abstract
Libraries of assays and methods of compiling the libraries are
provided. The assays can identify Single Nucleotide Polymorphisms
(SNPs). Methods of validating SNPs are provided. Methods of
constructing linkage disequilibrium maps using sets or subsets of
SNPs are also provided.
Inventors: |
De La Vega, Francisco; (San
Mateo, CA) ; Ziegle, Janet S.; (Berkeley, CA)
; Isaac, Hadar I.; (Foster City, CA) ; Scafe,
Charles R.; (San Francisco, CA) ; Spier, Eugene
G.; (Palo Alto, CA) |
Correspondence
Address: |
Leonard D Bowersox
Kilyk & Bowersox
3603 E Chain Bridge Road
Fairfax
VA
22030
US
|
Family ID: |
27671405 |
Appl. No.: |
10/502761 |
Filed: |
April 5, 2005 |
PCT Filed: |
January 27, 2003 |
PCT NO: |
PCT/US03/02240 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
60352039 |
Jan 25, 2002 |
|
|
|
60352356 |
Jan 28, 2002 |
|
|
|
60369127 |
Apr 1, 2002 |
|
|
|
60369657 |
Apr 3, 2002 |
|
|
|
60370921 |
Apr 9, 2002 |
|
|
|
60376171 |
Apr 26, 2002 |
|
|
|
60380057 |
May 6, 2002 |
|
|
|
60383627 |
May 28, 2002 |
|
|
|
60383954 |
May 29, 2002 |
|
|
|
60390708 |
Jun 21, 2002 |
|
|
|
60394115 |
Jul 5, 2002 |
|
|
|
60399860 |
Jul 31, 2002 |
|
|
|
Current U.S.
Class: |
435/6.11 ;
702/20 |
Current CPC
Class: |
G06Q 30/0641 20130101;
B01L 2300/021 20130101; C12Q 1/6858 20130101; G16B 35/20 20190201;
G16B 50/10 20190201; G16B 25/10 20190201; B01L 3/545 20130101; C12Q
2600/156 20130101; G06Q 30/0633 20130101; G16B 30/00 20190201; G16B
45/00 20190201; G01N 35/00663 20130101; G06Q 50/22 20130101; G16B
30/20 20190201; G16B 25/00 20190201; G16B 30/10 20190201; G16B
35/00 20190201; G16C 20/60 20190201; G16B 50/00 20190201; B01L
3/5453 20130101; G16H 10/40 20180101; G06Q 30/0621 20130101; C12Q
1/6858 20130101; C12Q 2547/101 20130101 |
Class at
Publication: |
435/006 ;
702/020 |
International
Class: |
C12Q 001/68; G06F
019/00 |
Claims
1-34. (canceled)
35. A library that contains a plurality of data sets, corresponding
to one or more respective oligonucleotides that can function as a
respective assay to hybridize with at least one respective Single
Nucleotide Polymorphism (SNP) in a nucleic acid sequence, wherein
the library is compiled using a method comprising the steps of:
providing a representation of a nucleic acid sequence; designating
a region within at least a part of the nucleic acid sequence;
determining the locations of known SNPs, if any, within the region;
determining the locations of newly identified SNPs, if any, within
the region; and selecting for the library a collection of data sets
that corresponds to a collection of the known SNPs and the newly
identified SNPs, wherein all of the SNPs of collection are: (1)
spaced less than a specified maximum distance apart from one
another along the nucleic acid sequence; (2) spaced more than a
minimum distance apart from one another along the nucleic acid
sequence; (3) spaced apart from one another such that each of the
distances between adjacent selected SNPs is from about 75% to about
125% of an average of the distances between the adjacent selected
SNPs; or (4) a combination of (1), (2), and (3).
36. The library of claim 35, wherein the method further comprises
the step of designating a specified maximum distance between
adjacent selected SNPs that correspond to respective data sets.
37. The library of claim 35, wherein the method further comprises
the step of designating a specified minimum distance between
adjacent selected SNPs that correspond to respective data sets.
38. The library of claim 35, wherein the method further comprises
the step of designating both a specified minimum distance between
adjacent selected SNPs that correspond to respective data sets and
a specified maximum distance between adjacent selected SNPs that
correspond to respective data sets.
39. The library of claim 35, wherein each of the distances between
adjacent selected SNPs is from about 75% to about 125% of an
average of the distances between the adjacent selected SNPs.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] This application claims the benefit under 35 U.S.C. .sctn.
119(e) of prior U.S. Provisional Patent Applications No.
60/352,039, filed Jan. 25, 2002; 60/352,356, filed Jan. 28, 2002;
60/369,127, filed Apr. 1, 2002; 60/369,657, filed Apr. 3, 2002;
60/370,921, filed Apr. 9, 2002; 60/376,171, filed Apr. 26, 2002;
60/380,057, filed May 6, 2002; 60/383,627, filed May 28, 2002;
60/383,954, filed May 29, 2002; 60/390,708, filed Jun. 21, 2002;
60/394,115, filed Jul. 5, 2002; and 60/399,860, filed Jul. 31,
2002; U.S. Non-Provisional patent application Ser. No. 10/335,707,
filed Jan. 2, 2003; U.S. Non-Provisional patent application Ser.
No. 10/335,690, filed Jan. 2, 2003; and U.S. Non-Provisional patent
application Ser. No. 10/334,793, filed Jan. 2, 2003; and
International Patent Application No. PCT/US03/00128, filed Jan. 2,
2003; all of which are incorporated herein in their entireties by
reference.
BACKGROUND
[0002] Assays can include probes and/or primers that hybridize with
a target nucleic acid sequence. These probes and primers can be
useful for visualizing or amplifying the target nucleic acid
sequence. The target nucleic acid sequence can have a Single
Nucleotide Polymorphism (SNP) contained therein. A relative
indication of concentration of the target nucleic acid sequence can
be obtained based on relative fluorescence of the probes.
[0003] Single nucleotide polymorphisms (SNPs) are an abundant form
of genetic variation. These single nucleotide changes are found
approximately every 500 bp in the human genome. Almost all SNPs are
bi-allelic, that is, only two different alleles exist. Typically,
one allele is present in the majority of the chromosomes of a
population, and the alternative variant, that is, the minor allele,
is present with less frequency. Only alleles that are present at a
frequency greater than 1% are considered polymorphisms.
[0004] SNPs are promising tools for mapping susceptibility
mutations that contribute to complex diseases. Although most SNPs
are neutral and do not affect phenotype, they can be used as
surrogate markers for positional cloning of genetic loci, because
of the allelic association, known as linkage disequilibrium (LD),
that can be shared by groups of adjacent SNPs. LD is eroded by gene
conversion and recombination, and the amount of LD depends on the
age of the mutations and on the demographic history of the
population. The extent of LD across a genomic region dictates the
density of SNP markers necessary to ensure association between a
marker and the causative allele sought.
[0005] Early attempts to model the extent of LD on theoretical
grounds predicted very short regions of LD, extending only a few
kilobases (Kb). However, empirical surveys reported average LD
distances between 5 Kb and 60 Kb, with the upper range extending up
to hundreds of Kb.
[0006] Previous efforts for typing the specific SNP alleles present
in a DNA sample produce unphased genotypes (i.e., the alleles
detected cannot be assigned to either the maternal or the paternal
chromosome). Although there are a few cumbersome methods to
directly determine haplotypes, previous algorithms are widely used
to infer the haplotypes from genotypes using maximum-likelihood or
Bayesian principles.
SUMMARY
[0007] According to various embodiments, methods are provided for
SNP validation that take into consideration a number of findings
and statistics. Studies have reported a discontinuous structure in
the patterns of LD across a set of regions sampled from the human
genome, where long stretches of strong LD are punctuated by
recombination hot-spots. These LD "blocks" show little evidence of
historical recombination. According to various embodiments, these
results are deconvoluted to predict that a reduced set of
contiguous chromosomal segments, or haplotypes, exist in specific
populations. For example, for a block spanning tens of Kbs for
which 10 SNPs exist, instead of the 2 theoretically possible
haplotypes, it has been found that 95% of the haplotype diversity
is made up on only 4 to 6 so-called common haplotypes.
[0008] It is noteworthy that these LD block patterns change
depending on the population sampled because of historical
differences; for example, populations that have experienced
bottlenecks (e.g., Caucasians) show longer LD blocks and less
evidence of historical recombination events, than other
populations. The haplotype diversity in a given population is
typically constant in a given region irrespective of the number of
SNPs sampled; therefore typing an arbitrarily large number of SNPs
within a LD block is unnecessary. Selecting the minimum subset of
SNPs within LD blocks, or any other discrete genetic locus, that
enable discrimination of the common haplotypes present in a block
without loss of information can be used to validate SNPs and/or to
compile a concise library of assays useful for genetic
analysis.
[0009] According to various embodiments, a method of compiling a
library of polynucleotide data sets is provided. The data sets can
correspond to polynucleotides that each can function as (A) a
primer for producing a nucleic acid sequence that is complementary
to at least one target nucleic acid sequence including a target
SNP, (B) a probe for rendering detectable the at least one target
nucleic acid sequence including a target SNP, or (C) both (A) and
(B). The method can include the step of selecting for the library
polynucleotide data sets that each correspond to a respective
polynucleotide that contains a sequence that is complementary to a
respective first allele included in each of the at least one target
nucleic acid sequences, if, under a set of reaction conditions a
number of parameters are met by each polynucleotide corresponding
to the data sets included in the library. The parameters can
include: (1) the respective polynucleotide has a background signal
value less than or equal to a first defined value, where the
background signal value is a first normalized ratio of a
fluorescence intensity of the respective polynucleotide reacted
with first assay reactants in the absence of the target nucleic
acid sequence, and under first conditions of fluorescence
excitation, to a dye fluorescence intensity of a passive-reference
dye under the first conditions; (2) the respective polynucleotide
has a signal generation value of greater than or equal to a second
defined value, wherein the signal generation value is the
difference between (i) a second normalized ratio of the
fluorescence intensity of the respective polynucleotide reacted
with the first assay reactants in the presence of the target
nucleic acid sequence, to the dye florescence intensity and (ii)
the background signal value; (3) the respective polynucleotide has
a specificity value of less than or equal to a third defined value,
wherein the specificity value is the difference between (i) a third
normalized ratio of the fluorescence intensity of the respective
polynucleotide reacted with second assay reactants that contain a
second allele included in the at least one target nucleic acid
sequence to the dye fluorescence intensity, wherein the second
allele differs from the first allele, and (ii) the background
signal value; (4) at least one individual from a population of
individuals has a genotype identifiable under the first conditions,
that results from reacting the respective polynucleotide with the
first assay reactants and in the presence of the target nucleic
acid sequence, wherein the population includes at least one
individual that has the identifiable genotype and at least one
individual that does not have the identifiable genotype; and (5) at
least one individual from the population has an identifiable minor
allele of the identifiable genotype, under the first conditions
that results from reacting the respective polynucleotide with the
first assay reactants in the presence of the target nucleic acid
sequence, wherein the population includes at least one individual
that has the identifiable minor allele, and at least one individual
that does not have the identifiable minor allele.
[0010] According to various embodiments, a method of compiling a
library of polynucleotide data sets is provided. The data sets can
correspond to polynucleotides that each can function as (A) a
primer for producing a nucleic acid sequence that is complementary
to at least one target nucleic acid sequence including a target
SNP, (B) a probe for rendering detectable the at least one target
nucleic acid sequence including a target SNP, or (C) both (A) and
(B). The method can include the step of determining a background
signal value by calculating a first normalized ratio of a
fluorescence intensity of a respective polynucleotide that contains
a sequence that is complementary to a first allele included in the
at least one target nucleic acid sequence, reacted with first assay
reactants in the absence of the target nucleic acid sequence, and
under first conditions of fluorescence excitation, to a dye
fluorescence intensity of a passive-reference dye under the first
conditions. The method can include the step of comparing a
difference between (i) a second normalized ratio of the
fluorescence intensity of the respective polynucleotide reacted
with the first assay reactants in the presence of the target
nucleic acid sequence, to the dye fluorescence intensity, and (ii)
the background signal value. The method can include the step of
comparing a difference between (i) a third normalized ratio of the
fluorescence intensity of the respective polynucleotide reacted
with second assay reactants that contain a second allele included
in the at least one target nucleic acid sequence to the dye
fluorescence intensity, wherein the second allele differs from the
first allele, and (ii) the background signal value. The method can
include the step of determining whether at least one individual
from a population of individuals has a genotype identifiable under
the first conditions that results from reacting the respective
polynucleotide with the first assay reactants and in the presence
of the target nucleic acid sequence, wherein the population
includes at least one individual that has the identifiable genotype
and at least one individual that does not have the identifiable
genotype. The method can include the step of determining whether at
least one individual from the population has an identifiable minor
allele of the identifiable genotype, under the first conditions
that results from reacting the respective polynucleotide with the
first assay reactants in the presence of the target nucleic acid
sequence. Various combinations of the herein described method steps
and/or parameters can be used.
[0011] According to various embodiments, a method of confirming the
existence of a SNP is provided. The method can include the step of
identifying a location corresponding to a possible SNP in a
polynucleotide ir a first collection of data sets containing
information on genomic deoxyribonucleic acid (DNA) samples in the
form of data sets corresponding to polynucleotides. The method can
include the step of confirming the existence of the SNP if at least
one condition is met. A condition can be met if a second collection
of data sets containing information on genomic deoxyribonucleic
acid (DNA) samples contains information that identifies the
location as containing the possible SNP. A condition can be met,
for example, if at least two data sets from the first collection of
data sets contain information corresponding to a minor allele of
the possible SNP at the location, wherein the at least two data
sets represent genomic deoxyribonucleic acid (DNA) samples obtained
from two independent sources. A condition can be met if a data set
that corresponds to a consensus sequence of genomic
deoxyribonucleic acid (DNA) samples that contains the minor allele
of the possible SNP in a third collection of data sets. The source
of the consensus sequence of genomic deoxyribonucleic acid (DNA)
samples, and the sources of the genomic deoxyribonucleic acid (DNA)
samples from the first collection of data sets, can be
independent.
[0012] According to various embodiments, a library is provided that
contains data corresponding to respective oligonucleotides that can
function as assays to detect Single Nucleotide Polymorphisms
(SNPs). The library can have a number of data sets corresponding to
not more than a sufficient number of oligonucleotides necessary to
provide a collection of assays that provides a maximum statistical
loss of a defined percentage of haplotype diversity across a human
genome.
[0013] According to various embodiments, an algorithm to select the
minimal subset of SNPs required for capturing the diversity of
haplotype blocks or other genetic loci is provided. The algorithm
can be used to quickly select the minimum SNP subset with no loss
of haplotype information. In addition, the algorithm can be used in
a more aggressive mode to further reduce the original SNP set, with
minimal loss of information.
BRIEF DESCRIPTION OF THE DRAWINGS
[0014] FIGS. 1a-b are graphs of SNPs per LD block v. minimum
information SNP subset for African-American and Caucasian
populations, respectively;
[0015] FIGS. 2a-2e are schematic diagrams of quenchable dyes that
can be part of a mixture of reagents provided and/or used according
to various embodiments;
[0016] FIG. 3 is a workflow diagram according to various
embodiments;
[0017] FIG. 4 is a graph showing visualized assay results,
according to various embodiments;
[0018] FIG. 5 is a flowchart showing an algorithm according to
various embodiments;
[0019] FIG. 6 is a flowchart showing an algorithm according to
various embodiments;
[0020] FIG. 7 is an illustration of SNPs selected by a method
according to various embodiments;
[0021] FIG. 8 shows a table of SNPs and genes on a hypothetical
chromosome;
[0022] FIG. 9 is a histogram of gene lengths of the hypothetical
chromosome of FIG. 8;
[0023] FIG. 10 is a histogram of the specified maximum distance
between adjacent SNPs;
[0024] FIG. 11 is a histogram of actual maximum distance between
adjacent SNPs;
[0025] FIG. 12 is a histogram of total selected SNPs per gene;
and
[0026] FIG. 13 is a histogram of the number of newly identified
SNPs selected per gene.
DESCRIPTION OF VARIOUS EMBODIMENTS
[0027] According to various embodiments, nucleic acid analogs can
be used in addition to or instead of nucleic acids. Examples of
nucleic acid analogs can include the family of peptide nucleic
acids (PNA), wherein the sugar/phosphate backbone of DNA or RNA has
been replaced with acyclic, achiral, and neutral polyamide
linkages. For example, a probe or primer can have a PNA polymer
instead of a DNA polymer. The 2-aminoethylglycine polyamide linkage
with nucleobases attached to the linkage through an amide bond can
be used as a PNA and shown to possess exceptional hybridization
specificity and affinity. An example of a PNA is as shown below in
a partial structure with a carboxyl-terminal amide: 1
[0028] "Nucleobase" as used herein means any nitrogen-containing
heterocyclic moiety capable of forming Watson-Crick hydrogen bonds
in pairing with a complementary nucleobase or nucleobase analog,
e.g. a purine, a 7-deazapurine, or a pyrimidine. Typical
nucleobases are the naturally occurring nucleobases such as, for
example, adenine, guanine, cytosine, uracil, thymine, and analogs
of the naturally occurring nucleobases, e.g. 7-deazaadenine,
7-deazaguanine, 7-deaza-8-azaguanine, 7-deaza-8-azaadenine,
inosine, nebularine, nitropyrrole, nitroindole, 2-aminopurine,
2-amino-6-chloropurine, 2,6-diaminopurine, hypoxanthine,
pseudouridine, pseudocytosine, pseudoisocytosine,
5-propynylcytosine, isocytosine, isoguanine, 7-deazaguanine,
2-azapurine, 2-thiopyrimidine, 6-thioguanine, 4-thiothymine,
4-thiouracil, O.sup.6-methylguanine, N.sup.6-methyladenine,
O.sup.4-methylthymine, 5,6-dihydrothymine, 5,6-dihydrouracil,
4-methylindole, pyrazolo[3,4-D]pyrimidines, "PPG", and
ethenoadenine.
[0029] "Nucleoside" as used herein refers to a compound consisting
of a nucleobase linked to the C-1' carbon of a sugar, such as, for
example, ribose, arabinose, xylose, and pyranose, in the natural
.beta. or the .alpha. anomeric configuration. The sugar can be
substituted or unsubstituted. Substituted ribose sugars can
include, but are not limited to, those riboses having one or more
of the carbon atoms, for example, the 2'-carbon atom, substituted
with one or more of the same or different Cl, F, --R, --OR,
--NR.sub.2 or halogen groups, where each R is independently H,
C.sub.1-C.sub.6 alkyl or C.sub.5-C.sub.14 aryl. Ribose examples can
include ribose, 2'-deoxyribose, 2',3'-dideoxyribose, 2'-haloribose,
2'-fluororibose, 2'-chlororibose, and 2'-alkylribose, e.g.
2'-O-methyl, 4'-.alpha.-anomeric nucleotides, 1'-.alpha.-anomeric
nucleotides, 2'-4'- and 3'-4'-linked and other "locked" or "LNA",
bicyclic sugar modifications. Exemplary LNA sugar analogs within a
polynucleotide can include the following structures: 2
[0030] where B is any nucleobase.
[0031] Sugars can have modifications at the 2'- or 3'-position such
as methoxy, ethoxy, allyloxy, isopropoxy, butoxy, isobutoxy,
methoxyethyl, alkoxy, phenoxy, azido, amino, alkylamino, fluoro,
chloro and bromo. Nucleosides and nucleotides can have the natural
D configurational isomer (D-form) or the L configurational isomer
(L-form). When the nucleobase is a purine, e.g. adenine or guanine,
the ribose sugar is attached to the N.sup.9-position of the
nucleobase. When the nucleobase is a pyrimidine, e.g. cytosine,
uracil, or thymine, the pentose sugar is attached to the
N.sup.1-position of the nucleobase.
[0032] "Nucleotide" as used herein refers to a phosphate ester of a
nucleoside and can be in the form of a monomer unit or within a
nucleic acid. "Nucleotide 5'-triphosphate" as used herein refers to
a nucleotide with a triphosphate ester group at the 5' position,
and can be denoted as "NTP", or "dNTP" and "ddNTP" to particularly
point out the structural features of the ribose sugar. The
triphosphate ester group can include sulfur substitutions for the
various oxygens, e.g. .alpha.-thio-nucleotide 5'-triphosphates.
[0033] As used herein, the terms "polynucleotide" and
"oligonucleotide" mean single-stranded and double-stranded polymers
of, for example, nucleotide monomers, including
2'-deoxyribonucleotides (DNA) and ribonucleotides (RNA) linked by
internucleotide phosphodiester bond linkages, e.g. 3'-5' and 2'-5',
inverted linkages, e.g. 3'-3' and 5'-5', branched structures, or
internucleotide analogs. Polynucleotides can have associated
counter ions, such as H.sup.+, NH.sup.4+, trialkylammonium,
Mg.sup.2+, Na.sup.+ and the like. A polynucleotide can be composed
entirely of deoxyribonucleotides, entirely of ribonucleotides, or
chimeric mixtures thereof. Polynucleotides can be comprised of
internucleotide, nucleobase and sugar analogs. For example, a
polynucleotide or oligonucleotide can be a PNA polymer.
Polynucleotides can range in size from a few monomeric units, e.g.
5-40 when they are more commonly frequently referred to in the alt
as oligonucleotides, to several thousands of monomeric nucleotide
units. Unless otherwise denoted, whenever a polynucleotide sequence
is represented, it will be understood that the nucleotides are in
5' to 3' order from left to right and that "A" denotes
deoxyadenosine, "C" denotes deoxycytidine, "G" denotes
deoxyguanosine, and "T" denotes thymidine, unless otherwise
noted.
[0034] "Internucleotide analog" as used herein means a phosphate
ester analog or a non-phosphate analog of a polynucleotide.
Phosphate ester analogs can include: (i) C.sub.1-C.sub.4
alkylphosphonate, e.g. methylphosphonate; (ii) phosphoramidate;
(iii) C.sub.1-C.sub.6 alkyl-phosphotriester; (iv) phosphorothioate;
and (v) phosphorodithioate. Non-phosphate analogs can include
compounds wherein the sugar/phosphate moieties are replaced by an
amide linkage, such as a 2-aminoethylglycine unit, commonly
referred to as PNA.
[0035] "Heteroygous" as used herein means both members of a pair of
alleles of a gene are present in a sample obtained from a single
source, wherein a gene call have two alleles due to, for example,
the fusion of two dissimilar gametes with respect to the gene.
[0036] "Heterozygous assay" as used herein means an assay adapted
to identify the allelic state of a gene having one or both members
of a pair of alleles.
[0037] "Homozygous" as used herein means one member of a pair of
alleles is present in a sample obtained from a single source,
wherein a gene can have one allele due to, for example, the fusion
of two identical gametes with respect to the gene.
[0038] "Homozygous assay" as used herein means an assay adapted to
identify only one of two possible allelic states of a gene having
one or both members of a pair of alleles.
[0039] "Lossy" as used herein means the loss of haplotype diversity
in a linkage disequilibrium block.
[0040] "Lossless" as used herein means that there is no loss of
haplotype diversity in a linkage disequilibrium block.
[0041] According to various embodiments, a library of assays can be
provided. The library of assays can have from about 100,000 to
about 500,000 polynucleotides, for example, about 150,000 to about
250,000 polynucleotides. According to various embodiments, a
library of data sets can be provided. The library of data sets can
have from about 100,000 to about 500,000 data sets, for example,
about 150,000 to about 250,000 data sets.
[0042] According to various embodiments, an algorithm is provided
that can select a minimum subset of SNPs without loss of haplotype
information, or an even smaller subset with some acceptable loss of
information. In an example, a SNP set was reduced by 18% for an
African American population and by 32% for a Caucasian population
with no loss of haplotype distribution information. The algorithm
can produce optimal results in a reasonable time. The algorithm can
allow for the real-time calculation of minimum SNP subsets for
haplotype blocks.
[0043] According to various embodiments, an algorithm to select the
minimal subset of SNPs required for capturing the diversity of
haplotype blocks or other genetic loci is provided. The algorithm
can be used to quickly select the minimum SNP subset with no loss
of haplotype information. In addition, the algorithm can be used in
a more aggressive mode to further reduce the original SNP set, with
minimal loss of information.
[0044] When SNPs are initially selected for typing, often not much
is known about the existence or location of LD blocks, or about the
number and relative frequencies of haplotypes within the blocks. It
is therefore typical in previous efforts to "over-sample" the
chromosomal region, (i.e., select SNPs as densely as one's budget
permits). Since there may be large costs associated with detecting
the genotype for each SNP, it can be practical to minimize the
number of SNPs used in a study. When beginning with a population
sample large enough to allow for accurate inference of the
haplotype distributions, various embodiments can reduce the set of
SNPs to the minimum number required for adequate coverage with no
loss of haplotype information. Furthermore, various embodiments can
be used to eliminate additional SNPs while minimizing loss of
haplotype information.
[0045] According to various embodiments, family relationships of
the DNA donors, if available, can be used to increase haplotype
inference accuracy. In the absence of family information, the
Expectation-Maximization algorithm introduced by Excoffier and
Slatkin can be accurate, especially in regions of low diversity.
According to various embodiments, the analysis of haplotype
distributions in genetic studies aimed to find susceptibility
mutations in case-control populations can be useful in finding
associations. Therefore, haplotype interference can be used in
disease and pharmacogenomic studies.
[0046] According to various embodiments, given a block containing N
SNPs and M haplotypes, a probability vector P of length M can be
defined where P.sub.i is the relative frequency of the i.sup.th
haplotype. According to various embodiments, A, a haplotype/SNP
allele state matrix of N columns and M rows is defined, wherein
A.sub.ij (the i.sup.th row of the j.sup.th column of the matrix)
indicates the allele state (`1` or `2`) of the j.sup.th SNP for the
i.sup.th haplotype. The algorithm can eliminate columns of A while
preserving as much of the information in P as possible. Quantifying
the information in P can be defined using the Shannon Entropy
equation: 1 H = - i = 1 M P i ln ( P i )
[0047] According to various embodiments, it can be useful to use an
algorithm in lossless mode. According to such embodiments, it can
be irrelevant which information measure is used for the haplotype
distribution. The algorithm can use other measures of
information.
[0048] According to various embodiments, the algorithm can consist
of two phases (phases I and II). These phases can be performed
sequentially. The operations can be outlined in lossless mode as
follows below.
[0049] In the first phase (Phase I), any column that is identical
to another column, or is the exact opposite of another column, can
be eliminated. A column in a matrix that is identical to another
column can represent a SNP that behaves identically to another SNP
for all tested samples. Thus, when the number of DNA samples is
large enough to infer the major haplotypes, the redundant SNP will
not provide any additional information. Similarly, when a column is
the exact opposite of another column in the matrix, this represents
a SNP where the behavior can always be predicted from the behavior
of another SNP simply by inverting it. Therefore, according to such
embodiments, this SNP will not provide new information. According
to various embodiments, after phase I, it can be assumed that N
columns of matrix A have been reduced to N' unique columns where
N'.ltoreq.N.
[0050] According to various embodiments, in the second phase (Phase
II) any column whose elimination does not reduce the number of
unique rows, can be eliminated. Each row in a matrix can represent
the allelic states of the SNPs for a specific haplotype. Removing a
"useful" SNP can eliminate the ability to detect at least one
haplotype. According to such an embodiment, two or more haplotypes
can register the same allelic state at the remaining SNPs, thereby
reducing the number of unique rows. Therefore, if the elimination
of a column does not reduce the number of unique rows, it can be
omitted.
[0051] Phase I can be a "sub-set" of phase II, in the sense that if
phase I is skipped, phase II can eliminate the SNPs that phase I
would have eliminated. Phase I can be computationally easier to
perform than phase II, for example, in lossy mode. Therefore it can
be more efficient to begin with phase I.
EXAMPLE 1
[0052] Example 1 illustrates a method according to various
embodiments, in lossless mode. Four SNPs and four haplotypes that
yield the following allelic responses are illustrated in Table
1.
1TABLE 1 Haplotype/SNP Allele State Matrix SNP.sub.1 SNP.sub.2
SNP.sub.3 SNP.sub.4 Haplotype 1 1 1 1 2 Haplotype 2 2 2 1 1
Haplotype 3 2 2 2 1 Haplotype 4 1 2 2 2
[0053] The fourth column is the exact opposite of the first column.
This implies that either SNP.sub.4 or SNP.sub.1 is redundant. If
SNP.sub.4 is removed from the SNP set, no information is lost. When
SNP.sub.1 registers allele "1", the state of SNP.sub.4 is known as
allele "2", and conversely, when SNP.sub.1 registers allele "2",
the state of SNP.sub.4 is known as allele "1". Removing SNP.sub.4
leaves the matrix seen in Table 2.
2TABLE 2 Haplotype/SNP Allele State Matrix after Phase I SNP.sub.1
SNP.sub.2 SNP.sub.3 Haplotype 1 1 1 1 Haplotype 2 2 2 1 Haplotype 3
2 2 2 Haplotype 4 1 2 2
[0054] The three columns are unique (including accounting for
opposites), thus phase I is complete. N=4 has been reduced to N'=3,
as phase II is entered.
[0055] Table 3 depicts the three remaining matrices, following the
removal of SNP.sub.1, SNP.sub.2, or SNP.sub.3, respectively. The
first and the third matrices only have three unique rows, whereas
the second matrix has four unique rows. Thus, if the haplotype list
is exhaustive, SNP.sub.2 can be eliminated with no loss of
haplotype detection.
3TABLE 3 Three Possible Haplotype/SNP Allele State Matrices
SNP.sub.2 SNP.sub.3 SNP.sub.1 SNP.sub.3 SNP.sub.1 SNP.sub.2
Haplotype 1 1 1 1 1 1 1 Haplotype 2 2 1 2 1 2 2 Haplotype 3 2 2 2 2
2 2 Haplotype 4 2 2 1 2 1 2
[0056] According to various embodiments, the set {SNP.sub.1,
SNP.sub.3} can provide the same haplotype detection ability as the
full set {SNP.sub.1, SNP.sub.2, SNP.sub.3, SNP.sub.4}. In Example
1, each phase can cause the elimination of exactly one SNP.
However, according to various embodiments, each phase can result in
the elimination of multiple SNPs or no SNPs.
EXAMPLE 2
[0057] Further elimination of SNPs, beyond the lossless elimination
shown in Example 1 above, can be implemented. According to such
embodiments, the retained SNP set can be optimized to minimize the
loss of haplotype detection. Phase I can remain unchanged and phase
II can select the optimal SNPs to eliminate.
[0058] According to various embodiments, for the 2 ( N ' k )
[0059] possible selections of k SNPs, the entropy H for the
resulting P is computed. The selection with the highest H can be
chosen as the best selection. When only k out of N' SNPs are
selected, N'-k columns can be eliminated. The resulting matrix
(with k columns) can have fewer unique rows than the full matrix
(with N' columns). Several "minor" haplotypes can be measured as a
single "major" haplotype when a row is repeated more than once.
This can occur because with fewer SNPs the ability to make the
finer distinction between them is lost. According to various
embodiments, the relative frequency (probability) of a "major"
haplotype is equal to the sum of the frequencies of the "minor"
haplotypes. According to such embodiments, when elimination of
columns results in repeating rows, the repeating rows can be
combined into a single row, and their respective probabilities can
be summed to form a new probability. The vector P can be shorter
and can have larger numbers. This can reduce the value of the
entropy, H. According to various embodiments, the combination with
the smallest reduction of entropy can be deemed the optimal
selection. According to various embodiments, if all the rows are
unique after elimination of N'-k columns, the entropy is not
reduced, and k SNPs can be used with no loss of information, as in
Example 1.
[0060] Example 2 uses an LD block that was discovered using the
Caucasian population panel, in Chromosome 6, overlapping the Human
gene TTK (RefSeq ID NM.sub.--003318, Celera ID hCG401205) in lossy
mode. The block consists of 17 SNPs, and the EM algorithm inferred
8 haplotypes, with two major ones: haplotype 2 and haplotype 7 with
frequencies of approximately 43% and 33%, respectively. The
remaining 24% of the diversity is spread among the remaining 6
haplotypes. Table 4 summarizes the allelic states of the 17 SNPs,
as well as the respective probability, for each of the 8
haplotypes.
4TABLE 4 Original Haplotype/SNP Allele State Matrix Haplotype SNP
No. Number P 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1 0.1136 1 1
1 1 1 2 1 2 1 2 1 1 2 2 2 1 2 2 0.4318 1 1 1 1 1 2 1 2 1 2 1 1 2 2
2 2 2 3 0.0114 1 1 1 2 2 1 2 1 2 1 2 2 1 1 1 1 1 4 0.0454 1 2 2 1 1
2 1 2 1 2 1 1 2 2 2 1 1 5 0.0454 2 1 1 1 1 2 1 2 1 2 1 1 2 2 2 1 2
6 0.0118 2 2 2 2 2 1 2 1 2 1 2 1 1 1 1 1 1 7 0.3292 2 2 2 2 2 1 2 1
2 1 2 2 1 1 1 1 1 8 0.0114 2 2 2 2 2 1 2 1 2 2 1 2 1 1 1 1 1
[0061] After running phase I of the algorithm, the number of SNPs
is reduced almost immediately to 7, with the remaining SNP set
being {SNP.sub.1, SNP.sub.2, SNP.sub.4, SNP.sub.10, SNP.sub.12,
SNP.sub.16, SNP.sub.17}. All the haplotype information is
preserved, including haplotype distribution, as shown in Table 5.
The entropy of the original distribution of haplotypes is
H(P)=2.0351 bits. Phase II of the algorithm was then performed.
5TABLE 5 Haplotype/SNP Allele State Matrix After Phase I Haplotype
Number P SNP.sub.1 SNP.sub.2 SNP.sub.4 SNP.sub.10 SNP.sub.12
SNP.sub.16 SNP.sub.17 1 0.1136 1 1 1 2 1 1 2 2 0.4318 1 1 1 2 1 2 2
3 0.0114 1 1 2 1 2 1 1 4 0.0454 1 2 1 2 1 1 1 5 0.0454 2 1 1 2 1 1
2 6 0.0118 2 2 2 1 1 1 1 7 0.3292 2 2 2 1 2 1 1 8 0.0114 2 2 2 2 2
1 1
[0062] Table 6 shows the optimum SNP subset for k SNPs out of the 8
SNPs that survived phase I. Note that the haplotype information is
fully preserved all the way down to k=5, therefore if a lossless
version of Phase II were run, the minimum SNP size with no loss of
information would be 5. At a SNP subset size of 4, some information
is lost. In fact, haplotype 8, the rarest haplotype, was merged
into haplotype 7, and the ability to distinguish between the two is
lost. As a result, 3.5% of the entropy of the resultant haplotype
distribution is lost. The optimal selection of 3 SNPs cause
haplotypes 3 and 4 to merge and cause haplotypes 6, 7, and 8 to
merge, with a total loss of 9.2% of original entropy. The optimal
single SNP is SNP.sub.16. With single SNP.sub.16, the detection
ability is reduced to: "haplotype2" or "other." Since haplotype 2
is the most common, with 43.2% of the frequency, if only a single
SNP was chosen, SNP.sub.16 would be the most useful choice.
6TABLE 6 Lossy Min. SNP Set Example No. of SNPs (k) 3 No. of
Combinations ( 8 k ) Optimal Set of k SNPs Haplotype Distribution
Resulting from the Optimal SNP Set Resulting Entropy (H) (bits) 7 8
{SNP.sub.1, SNP.sub.2, (0.114, 0.432, 2.0351 SNP.sub.4, SNP.sub.10,
0.011, 0.045, 0.045, SNP.sub.12, SNP.sub.16, 0.012, 0.329, 0.011)
SNP.sub.17} 6 28 {SNP.sub.1, SNP.sub.4, (0.114, 0.432, 2.0351
SNP.sub.10, SNP.sub.12, 0.011, 0.045, 0.045, SNP.sub.16,
SNP.sub.17} 0.012, 0.329, 0.011) 5 56 {SNP.sub.1, SNP.sub.10,
(0.114, 0.432, 2.0351 SNP.sub.12, SNP.sub.16, 0.011, 0.045, 0.045,
SNP.sub.17} 0.012, 0.329, 0.011) 4 70 {SNP.sub.1, SNP.sub.12,
(0.114, 0.432, 1.9631 SNP.sub.16, SNP.sub.17} 0.011, 0.045, 0.045,
0.012, 0.341)) 3 56 {SNP.sub.1SNP.sub.16, (0.114, 0.432, 1.8475
SNP.sub.17} 0.057, 0.045, 0.352)
[0063] To validate and assess the utility of the algorithm,
genotyping data was used from 11,160 SNPs distributed in a
gene-centric fashion across chromosomes 6, 21, and 22, with
intragenic spacing averaging 12 Kb, 8 Kb, and 9 Kb, respectively.
The SNPs were scored with 5.alpha.nuclease assays including
TAQMAN-MGB probes from Applied Biosystems' Assays-on-Demand.TM. SNP
Genotyping Products (Foster City, Calif., USA). The samples typed
included 45 African-American and 45 Caucasian DNAs from the Coriell
Human Diversity Collection available from Coriell Institute for
Medical Research, Camden, N.J., USA. LD blocks and haplotypes were
computed independently for each population using methods described
in Abecasis, et al., Merlin--rapid analysis of dense genetic maps
using sparse gene flow trees. Nat Genet 30:97-101 (2002) and
Gabriel et al., The structure of haplotype blocks in the human
genome Science 296:2225-2229 (2002), both of which are herein
incorporated in their entireties by reference. Only blocks of 3 or
more SNPs were considered. Therefore, only 4,864 SNPs were used for
the African-American population and 7,347 SNPs were used for the
Caucasian population. The Caucasian population is known, in
general, to have more and longer LD blocks. The algorithm was
implemented in MATLAB v 6.1, available from The MathWorks Inc.,
Natick, Mass., USA, without further optimization. The computations
were completed on a 700 MHz PC in less than 1 minute.
[0064] Table 5 summarizes the results after applying the algorithm
to the haplotype blocks detected in data for chromosomes 6, 21, and
22. The African-American population panel is denoted by `A` and the
Caucasian population panel is denoted by `C`.
7TABLE 7 Results Summary Mean Mean Total Spacing Spacing Mean Mean
Mean Min. No. Between Between No. of Block SNPs SNP per Block of
SNPs SNPs in Haplotype Size per <10% Chr. Pop. SNPs (bp) Genes
(bp) Blocks (bp) Block Lossless Loss 6 A 2,504 24,386 10,840 646
23,000 3.88 2.94 2.44 C 4,009 23,694 10,630 883 34,000 4.54 2.86
2.33 21 A 955 12,424 7,382 242 14,933 3.95 2.92 2.39 C 1,555 11,921
7,031 336 21,032 4.63 2.88 2.32 22 A 1,405 10,041 6,035 350 13,714
4.01 2.99 2.47 C 1,783 9,080 7,760 417 17,505 4.28 2.81 2.27
[0065] FIG. 1 illustrates the relationship between the original
number of SNPs in an LD block (horizontal axis) and the minimum
number of SNPs required to genotype the LD block with no loss of
information (vertical axis). The thickness of the `x` corresponds
to the number of different blocks found in chromosome 6 with the
same properties.
[0066] Previous algorithms for finding the minimum SNP sub-set have
been concerned with complete gene or randomly selected loci, as
opposed to LD blocks, in such previous efforts, the number of
haplotypes, and more importantly, the amount of information in the
haplotype distribution, was expected to be much higher than those
presented in the above examples. As a result, these previous
efforts were challenged by the numerical complexity, and thus
locally optimal solutions were sought. In contrast, algorithms
according to various ones of the present embodiments can compute
the global optimum, whether in lossless or lossy mode, making it
superior to previous efforts.
[0067] The method described by Judson et al., How many SNPs does a
genome-wide haplotype map require? Pharmacogenomics 3:379 (2002) is
essentially equivalent to phase II of the lossy version of various
embodiments except that the algorithm is limited to k.ltoreq.11.
This is expected, because without the efficient pruning of SNPs
performed by phase I, the exponential nature of phase II can result
in practically infinite execution time. For example, the largest
block found in the above examples consisted of 22 SNPs. In previous
efforts, comparing sub-sets of 1 to 11 out of 22 SNPs required
examining over 2.4 million combinations. Even after that
computation, the optimal solution is not assured since it is only a
local optimum. In general, sub-sets of 12 to 22 would also need to
be examined in order to assure a global optimum, bringing the total
number of combinations to almost 4.2 million. Various embodiments
can use phase I to quickly reduce the 22 SNPs to a subset of 4
SNPs. As a result, phase II can find the global optimum (3 SNPs in
lossless mode or 2 SNPs with less than 10% loss) in 15 comparisons
only.
[0068] The method described in the on-line supplement to the paper
by Johnson et al., Haplotype tagging for the identification of
common disease genies Nat Genet 29:233-237 (2001) compromises the
maximization of the information detected by the SNP set with other
considerations, e.g. maximization of the individual SNPs'
properties. There is little explanation of the method details. One
example provided tries to show that a haplotype matrix with full
rank cannot be pruned. Note, however, that the haplotype matrix
used in Example 1 and shown in Table I herein is of full rank, yet
was pruned with no loss of information. A counter example is the 2
by 2 identity matrix, which is know to be full rank, but has the
second column as the perfect inverse of the first column, thus
providing no new information. The Johnson et al. on-line supplement
also provides executable programs, but the maximum subset size is
set to k.ltoreq.5, thereby guaranteeing suboptimal results based on
the finding that the global optimum is greater than 5 in some
blocks.
[0069] Previous reports have suggested that the Caucasian
population has longer LD blocks than the African-American
population, which is considered more diverse. Examples 1 and 2
herein used the same SNPs for both populations, but kept only SNPs
which formed LD blocks of 3 or more SNPs. This left more SNPs for
the Caucasian population. As Table 5 shows, the Caucasian
population yielded more blocks, and more SNPs per block. However,
after "compression" of the SNP sets of each block into the minimum
required to represent the information (with no loss), the two
populations are almost the same. This can reflect the arbitrary
criteria that defines an LD block, and can reflect that the
criteria was applied uniformly to both populations.
[0070] The premise of the algorithm, according to various
embodiments, is that the DNA sample size for each population is
large enough so that the inferred haplotypes adequately represent
reality. There can be a risk that a SNP whose behavior is identical
to another SNP (and thus deemed worthless in terms of new
information) for the sample size used, could differentiate an
additional haplotype inferred with a larger sample size. This risk
can be lower for common haplotypes. Rare haplotypes can harbor a
causative mutation and can be present in higher frequency in some
cases. Experimental errors can eliminate data points and thus can
render suboptimal the minimum SNP subset. Therefore, additional
SNPs can enhance the minimum SNP subset to enhance robustness.
[0071] Data was provided from over 11,000 SNPs with an average
spacing of 6 to 11 Kb, across all the genes of chromosomes 6, 21,
and 22, and typed on DNA samples of 45 unrelated African-Americans
and 45 Caucasians from the Coriell Human Diversity Collection. With
no loss of information, the number of SNPs required to capture the
haplotype block diversity by 25% for the African-American
population and 36% for the Caucasian population was reduced. With a
maximum loss of 10% of haplotype distribution information, the SNP
reduction was 38% and 49%, respectively, for the two populations.
All computations were performed in less than 1 minute for the
dataset used.
[0072] The availability of human genome data, including putative
single nucleotide polymorphisms (SNPs), can enable new clinical and
research methods. According to various embodiments, databases
containing information on SNPs can be used for conducting genetic
studies. Putative SNPs can be validated and can be assembled into a
standardized SNP marker map or a database containing data sets
corresponding to the standardized SNP marker map. SNP information
can be easily accessible and standardized assay reagents can be
developed, validated, and made available to enable high throughput
and automation, for example, to screen many SNPs on many
individuals.
[0073] According to various embodiments, a reference SNP database
can be produced from SNP and genomic information from both
proprietary and public databases. The database can be used for
linkage disequilibrium (LD) mapping and can be used to provide, for
example, validated, ready-to-use assays and reagents. The database
can provide high-density coverage of any known gene regions and can
enable easier and more affordable candidate gene association
studies and candidate region association studies. Researchers can
select SNPs across candidate genes or chromosomal regions that are
most suitable for a given study, and can quickly translate that
information into practice by, for example, directly obtaining assay
protocols and reagents for those SNPs. A "core" set of SNPs and the
associated assay reagents can be compiled into a database or
library and expanded or refined as additional information become
available, such as haplotype definition for some or all of the
genome. The SNP database can also be used to compile a fixed set of
chromosome-based assays for cost-efficient whole genome association
(WGA) studies using, for example, oligonucleotide ligation assay
(OLA) PCR Bead Array systems for ultra-high throughput
genotyping.
[0074] Linkage disequilibrium (LD) is the non-random association of
alleles in a chromosomal segment, and can be the basis of all
genetic mapping. Selecting SNPs as genetic markers for LD studies
involve considering all genetic and assay-specific technical
factors that affect the ability to find association between a
marker and the susceptibility mutations being mapped.
[0075] According to various embodiments, the extent of LD across a
genomic region can dictate the SNP density necessary to ensure
association between a marker and the allele sought. Early attempts
to model the extent of LD predicted very short LD of only a few
kilobases (kb). However, recent empirical surveys report average LD
levels between 5 kb and 60 kb, and extending up to hundreds of kb,
which implies that the number of SNPs required for WGA studies
could range from 50,000 to 250,000, and that markers spaced by tens
of kb will suffice for candidate gene studies. Common SNPs are the
most likely to be useful for LD studies across more than one
population since they represent ancient mutations that arose before
ethnic group segregation. Simulation studies suggest that common
SNPs are more likely than coding SNPs (cSNPs) to be in LD with a
given causative allele regardless of whether the allele is present
at low or high frequency.
[0076] According to various embodiments, common SNPs can be used to
assemble a database in a hybrid gene-based approach. SNPs can be
considered "common" when the minor allele frequency is, for
example, less than 15% in at least one of the populations used for
validation. A gene list can include 25,083 gene regions derived by
Celera Genomics. A gene region can be defined as bounded by the
first and last transcribed base, including untranslated regions,
plus 10 kilobases (kb) upstream and downstream to account, for
example, for uncharacterized exons and regulatory regions. SNPs can
be selected within gene regions at an average density of, for
example, one SNP per 10 kb, such that the map can resemble a
gene-focused picket fence. Density for specific regions can be
adjusted as data on recombination and LD extent emerges. Additional
SNPs in intergenic regions, such as, for example, non-coding
regions of homology between mouse and human, can be added to a
database or library.
[0077] Over four million unique SNPs have been reported. However,
literature reports state that as few as 50% of SNPs randomly
selected from public databases are polymorphic and can yield
working assays. According to various embodiments, obtaining a
validated SNP assay at the end of a process of validating possible
SNPs can be enhanced by defining prioritization criteria One
criterion that can be used to validate a possible SNP is evidence
of independent discovery of the minor allele. A data set
corresponding to a possible SNP can be cross-referenced against
data available in public sources. When a data set corresponding to
a possible SNP has no equivalent in the public domain, observation
of the minor allele in genotypes of two independent donors can be
used to validate the SNP. When a data set corresponding to a
possible SNP has a minor allele that is found in only one donor, an
independent instance of the minor allele found by searching the
consensus assembly of the public Human Genome Project can be used
to validate the SNP. A SNP database obtained using the above
criteria can include about one million data sets corresponding to
SNPs.
[0078] According to various embodiments, methods of confirming the
existence of a SNP are provided. One step can include identifying a
location corresponding to a possible SNP in a polynucleotide in a
first collection of data sets. The first collection can contain
information on genomic deoxyribonucleic acid (DNA) samples in the
form of data sets corresponding to polynucleotides. Another step
can include confirming the existence of the SNP if at least one of
a number of conditions is present or met. A condition can be that a
second collection of data sets containing information on genomic
DNA samples contains information that identifies the location as
containing the possible SNP. A condition can be that at least two
data sets from the first collection of data sets contain
information corresponding to a minor allele of the possible SNP at
the location. According to various embodiments, the at least two
data sets representing genomic DNA samples are obtained from two
independent sources. A condition can be that a data set that
corresponds to a consensus sequence of genomic DNA samples in a
third collection of data sets has the minor allele of the possible
SNP. According to various embodiments, the source of the genomic
DNA of the consensus sequence and the sources of the genomic
deoxyribonucleic acid (DNA) samples from the first collection of
data sets are independent. The third database of genomic DNA
samples can be, for example, a public database of the Human Genome
Project. The first database of at least one genomic DNA sample can
be, for example, a proprietary database of the Human Genome
Project.
[0079] According to various embodiments, a multi-step,
high-throughput assay design pipeline can be provided to ensure
optimum performance of assays. The methods provided can enable
automation, minimize assay failure, and ensure compatibility of the
SNP sequence with, for example, TAQMAN probe-based 5' nuclease
chemistry, available from Applied Biosystems, Foster City, Calif.,
and/or other assay formats. A stringent scoring system can be used
to select only those SNP context sequences with the highest
probability of success.
[0080] According to various embodiments, a bioinformatics process
can be used to design assays. A step can involve masking SNPs
adjacent to the target polymorphism and/or any sequence discrepancy
between the Celera and the HGP human genome assembly, within the
600 bases of a context sequence. This can prevent primers and
probes from being placed on top of other SNPs and can maximize the
chance that the probes will hybridize to the correct genomic
sequence.
[0081] TAQMAN 5' nuclease primers and probes can be designed using,
for example, the ASSAYS-BY-DESIGN custom oligonucleotide reagent
service (Applied Biosystems, Foster City, Calif.). Oligonucleotides
can be designed in batch mode without manual intervention, and a
scoring scheme can select the best sequences for a given SNP. The
design algorithm can implement thermodynamic and heuristic rules
and additional empirically-derived factors can increase
manufacturability and assay performance. According to various
embodiments, probes can be designed successfully for, for example,
97% of SNPs. After this step, a further computational
quality-control step can also be performed in the context of the
genome that can allow the elimination of potentially problematic
SNP targets that may arise from repeated genomic regions,
pseudo-SNPs, and/or other possible assembly artifacts.
[0082] Finally, the primers and probes can be synthesized, and
additional quality-control steps can occur. For example,
oligonucleotide integrity can be tested. For further example, assay
performance can be tested against a panel of 10 DNA samples.
According to various embodiments, assays that pass
post-manufacturing quality control can be validated in the
population panels.
[0083] Assay validation in population panels can ensure that the
locus is polymorphic and that the allele frequency is adequate for
association studies in a variety of populations. For example,
ninety (90) samples from the Coriell Human Variatiori Collection
were obtained. By obtaining individual genotypes from a panel of 45
African Americans, a panel of 45 Caucasians, and a chimp DNA sample
(to provide insight into ancestral alleles), sufficient information
was obtained to estimate linkage disequilibrium between the SNPs in
the LD map and to computationally infer common haplotypes. Assay
validation in population panels can provide additional information
on the usefulness of the markers, the coverage provided for a given
study, and/or provide an independent assessment of assay
performance.
[0084] According to various embodiments, the performance of each
assay that can be comprised of at least one polynucleotide can be
benchmarked against criteria, such as, for example: background
signal (e.g., low signal in the control experiments run without
template); signal generation (e.g., good separation between control
experiments run without template and allele clusters); and
specificity. A criterion can be that a maximum of three clusters of
fluorescing sample and a minimum of two clusters of fluorescing
sample must be observed. Another criterion can be that at least 90%
of samples yield callable genotypes.
[0085] According to various embodiments, a method is provided for
compiling a library of polynucleotide data sets that each
correspond to polynucleotides that can function as (A) a primer for
producing a nucleic acid sequence that is complementary to at least
one target nucleic acid sequence including a target SNP, (B) a
probe for rendering detectable the at least one target nucleic acid
sequence including a target SNP, or (C) both (A) and (B). The
method can include the step of selecting for the library
polynucleotide data sets that each correspond to a respective
polynucleotide that contains a sequence that is complementary to a
respective first allele included in each of the at least one target
nucleic acid sequences, if, under a set of reaction conditions, a
number of parameters are met by each polynucleotide corresponding
to the data sets included in the library. These parameters can
include: (1) the respective polynucleotide has a background signal
value less than or equal to a first defined value, where the
background signal value is a first normalized ratio of a
fluorescence intensity of the respective polynucleotide reacted
with first assay reactants in the absence of the target nucleic
acid sequence, and under first conditions of fluorescence
excitation, to a dye fluorescence intensity of a passive-reference
dye under the first conditions; (2) the respective polynucleotide
has a signal generation value of greater than or equal to a second
defined value, wherein the signal generation value is the
difference between (i) a second normalized ratio of the
fluorescence intensity of the respective polynucleotide reacted
with the first assay reactants in the presence of the target
nucleic acid sequence, to the dye fluorescence intensity and (ii)
the background signal value; (3) the respective polynucleotide has
a specificity value of less than or equal to a third defined value,
wherein the specificity value is the difference between (i) a third
normalized ratio of the fluorescence intensity of the respective
polynucleotide reacted with second assay reactants that contain a
second allele included in the at least one target nucleic acid
sequence to the dye fluorescence intensity, wherein the second
allele differs from the first allele, and (ii) the background
signal value; (4) at least one individual from a population of
individuals has a genotype identifiable under the first conditions,
that results from reacting the respective polynucleotide with the
first assay reactants and in the presence of the target nucleic
acid sequence, wherein the population includes at least one
individual that has the identifiable genotype and at least one
individual that does not have the identifiable genotype; and (5) at
least one individual from the population has an identifiable minor
allele of the identifiable genotype, under the first conditions
that results from reacting the respective polynucleotide with the
first assay reactants in the presence of the target nucleic acid
sequence; wherein the population includes at least one individual
that has the identifiable minor allele, and at least one individual
that does not have the identifiable minor allele. Various
embodiments of these and/or other parameters can be used in
deciding whether to select a polynucleotide or a polynucleotide
data set for a library.
[0086] According to various embodiments, the first reaction
conditions can comprise a 900 nM final primer concentration and a
250 nM final probe concentration under thermal cycling conditions.
The first defined value can be about 2.0, the second defined value
can be about 1.0, and the third defined value can be about 2.0. At
least, for example, 0.01% of individuals from the population can
have the identifiable genotype. At least 10% of individuals from
the population can have the identifiable genotype. At least 20% of
individuals from the population can have the identifiable genotype.
The identifiable genotype can result from reacting the respective
polynucleotide with the first assay reactants in the presence of
the target nucleic acid sequence. The reaction can occur under the
first conditions. The population can have a frequency of the minor
allele of greater than or equal to about 5%. For example, the minor
allele frequency can be greater than or equal to about 10%. The
minor allele frequency can be greater than or equal to about 15%.
According to various embodiments, methods can include not selecting
a second polynucleotide data set that corresponds to a second
polynucleotide if one or more of parameters (1)-(5), above, is not
met by the second polynucleotide.
[0087] A library of polynucleotide data sets can be compiled using
methods according to various embodiments. A library of assays can
be compiled using methods according to various embodiments. The
method can include manufacturing a library of assays wherein each
assay can be made using a polynucleotide data set compiled in the
library. A library of polynucleotides can be compiled by
manufacturing polynucleotides corresponding to polynucleotide data
sets compiled using methods according to various embodiments. A
library of assays can be compiled using methods according to
various embodiments.
[0088] According to various embodiments, a method of detecting a
SNP can be provided. A step of the method can be reacting a sample
containing a target nucleic acid sequence that has a target SNP
with an assay selected from the library of assays compiled
according methods described herein. A step can be determining the
genotype of the target nucleic acid sequence that has the target
SNP by detecting a characteristic attributable to the genotype of
the target SNP in the sample.
[0089] According to various embodiments, a method is provided for
compiling a library of polynucleotide data sets that correspond to
polynucleotides that each can function as (A) a primer for
producing a nucleic acid sequence that is complementary to at least
one target nucleic acid sequence including a target SNP, (B) a
probe for rendering detectable the at least one target nucleic acid
sequence including a target SNP, or (C) both (A) and (B). The
method can include the step of determining a background signal
value by calculating a first normalized ratio of a fluorescence
intensity of a respective polynucleotide that contains a sequence
that is complementary to a first allele included in the at least
one target nucleic acid sequence, reacted with first assay
reactants in the absence of the target nucleic acid sequence, and
under first conditions of fluorescence excitation, to a dye
fluorescence intensity of a passive-reference dye under the first
conditions. The method can include the step of comparing a
difference between (i) a second normalized ratio of the
fluorescence intensity of the respective polynucleotide reacted
with the first assay reactants in the presence of the target
nucleic acid sequence, to the dye fluorescence intensity, and (ii)
the background signal value. The method can include the step of
comparing a difference between (i) a third normalized ratio of the
fluorescence intensity of the respective polynucleotide reacted
with second assay reactants that contain a second allele included
in the at least one target nucleic acid sequence to the dye
fluorescence intensity, wherein the second allele differs from the
first allele, and (ii) the background signal value. The method can
include the step of determining whether at least one individual
from a population of individuals has a genotype identifiable under
the first conditions that results from reacting the respective
polynucleotide with the first assay reactants and in the presence
of the target nucleic acid sequence, wherein the population
includes at least one individual that has the identifiable genotype
and at least one individual that does not have the identifiable
genotype. The method can include the step of determining whether at
least one individual from the population has an identifiable minor
allele of the identifiable genotype, under the first conditions,
that results from reacting the respective polynucleotide with the
first assay reactants in the presence of the target nucleic acid
sequence. The method can include a combination of some of or all of
these steps.
[0090] According to various embodiments, the polynucleotide data
set corresponding to the respective polynucleotide can be selected
for the library if, for example, the background signal value in
parameter (1) is less than or equal to about two, if the ratio from
the comparison in parameter (2) is greater than or equal to about
one, if the ratio from the comparison in parameter (3) is less than
or equal to about two, if the at least one individual of parameter
(4) has the identifiable genotype, and if the at least one
individual of parameter (5) has the identifiable minor allele.
[0091] A library of polynucleotide data sets can be compiled using
methods according to various embodiments. A library of
polynucleotides can be compiled by manufacturing polynucleotides
corresponding to polynucleotide data sets compiled using a method
or methods according to various embodiments.
[0092] A method of compiling a library of assays can be provided.
The method can include manufacturing a library of assays, wherein
each assay is manufactured using a polynucleotide data set compiled
in a library according to various embodiments.
[0093] According to various embodiments, methods of detecting a SNP
can be provided. The method can include the step of reacting a
sample containing a target nucleic acid sequence that has a target
SNP with an assay selected from the library of assays compiled
using a method or methods according to various embodiments. A step
can include determining the genotype of the target nucleic acid
sequence that has the target SNP by detecting a characteristic
attributable to the genotype of the target SNP in the sample.
EXAMPLE 3
[0094] Using chromosome 22 as a pilot project, 2,260 SNP assays
were validated. Of those assays tested, 94% of the SNPs tested with
population panels were polymorphic and 90% of the assays passed
performance criteria. When a minor allele frequency cut-off of 15%
or greater was used, 85% of the samples from the African-American
panel, and 90% from the Caucasian panel, met the performance
criteria. Ninety-nine percent (99%) of SNPs had a minor allele
frequency of 15% or greater in at least one of the population
panels. Ninety-six percent (96%) of samples had a minor allele
frequency of 20% or greater in at least one population tested, and
62% of samples had a minor allele frequency of 20% or greater in
both populations tested.
[0095] According to various embodiments, an automatic allele
calling software can be used to automatically analyze validated
assay data without user intervention. According to various
embodiments, at least, for example, 90% of the assay data can be
processed automatically to identify an allele. According to various
embodiments, an automated validation process can be used for high
volume commercial or research purposes.
[0096] FIG. 4 is a plot 100 of fluorescence data from many SNP
assays according to various embodiments. The x-axis represents
relative fluorescence of a 6-FAM dye label and the y-axis
represents relative fluorescence of a VIC dye label. Cluster 110
represents the relative fluorescence of control samples having
probes labeled with 6-FAM and VIC, respectively. The control
samples did not contain a target nucleic acid sequence. The
background signal value is the average of the relative fluorescence
of the control samples as represented by cluster 110 in FIG. 4. The
background signal value in FIG. 4 is less than about 2.0. Cluster
120 represents the relative fluorescence of samples having
homozygous alleles (allele 2) that hybridized with probes labeled
with 6-FAM. Cluster 130 represents the relative fluorescence of
samples having heterozygous alleles (alleles 1 and 2) that
hybridized with probes labeled with VIC and 6-FAM, respectively.
Cluster 140 represents the relative fluorescence of samples having
homozygous alleles (allele 1) that hybridized with probes labeled
with VIC. The signal generation value for the assays is based, at
least in part, on the average of the relative fluorescence of at
least one of clusters 120, 130, and 140.
[0097] FIG. 5 is a flowchart showing a comparison of reference
values determined by experimental assays according to various
embodiments against target values. As used herein TBV is a target
background value; TsigV is the target signal value; TSpV is the
target specificity value; TIP is the target identifiable
percentage, or the minimum frequency that the assay produces an
identifiable genotype; and TMI is the target minor allele
frequency, or the minimum frequency that the minor allele appears
in the population.
[0098] FIG. 6 is a flowchart that illustrates the selection of a
data set into a library of data sets, according to various
embodiments. Experimental or reference data is obtained by
performing an assay against individuals in a population.
Experimental data can be obtained at least until results are
statistically significant for a given population. When a
statistically significant sample size has been determined, the
experimental data can be processed according to various embodiments
as shown, for example, in FIG. 5. If the selection criteria are
met, the data set corresponding to polynucleotides used in the
experimental assays can be added into the library. If the selection
criteria are not met, the polynucleotide can be redesigned or
discarded, where the expendable assays can be repeated.
[0099] According to various embodiments, SNP assays and reagents
can use easily automated chemistry and can be compatible with
readily available high-throughput instrumentation and software
systems. According to various embodiments, SNP assays and reagents
can have few enzymatic steps, no post-reaction transfer of liquids,
and/or universal reaction conditions that can facilitate robotic
liquid handling automation. The assays, reagents, and/or
high-throughput workflow can be easy to implement and automate
and/or can use components that are ready-to-use out-of-the-box and
require no optimization.
[0100] TAQMAN probe-based 5' nuclease assay chemistry, available
from Applied Biosystems, Foster City, Calif., can meet almost any
assay requirement and can unite PCR amplification and signal
generation into a single step, thereby simplifying automation of
both reaction set-up and data collection. In the TAQMAN system, a
hybridization probe with fluorogenic and quencher tags is cleaved
by the 5' nuclease activity of thermos aquaticus (Taq) DNA
polymerase during PCR amplification. Cleavage produces fluorescence
by freeing the fluorogenic molecule from the quencher. By using two
probes, one specific to each allele of the SNP and labeled with
distinct fluorogenic tags, both alleles can be specifically
detected in a single tube. In addition, the fluorescent 5' nuclease
assays can be part of an easy-to-use, automated system for SNP
genotyping. FIGS. 2a-2e provide an overview of the TAQMAN
probe-based 5' nuclease assay chemistry for SNP genotyping.
[0101] The TAQMAN system is adapted to provide allelic
discrimination and high-throughput SNP genotyping. Chemistry
improvements have increased assay design flexibility, enabled easy
protocol standardization, enabled the use of universal reagents,
and reduced background fluorescence, all of which can be desirable
for high throughput SNP processing and allelic discrimination.
[0102] For example, previous probes of up to 30 bases were required
to achieve the specificity required for scoring SNPs. The
conjugation of a DNA minor groove binder (MGB) to a probe can
significantly stabilize probe-template complexes, enabling the use
of probes in the 13-mer to 20-mer size range. Such conjugated
probes can have better mismatch discrimination, can be easier to
design for challenging genetic regions such as those high in GC
content or those in variable context sequences, and/or can increase
the signal-to-noise ratio by bringing the quencher closer to the
fluorescent tag. For high-throughput SNP scoring, conjugated probes
can increase the melting temperature window that can be used in
reaction protocols, thereby allowing all the SNP assays to run
under identical conditions.
[0103] Previous 5' nuclease quenchers emitted their own
fluorescence, therefore signal detection was complicated. New
non-fluorescent quenchers can provide improved signal detection and
can facilitate automated allele calling, another desirable feature
useful for high-throughput SNP scoring.
[0104] In a pharmacogenomic study, the most precious reagent can be
the DNA sample itself. High-quality DNA quantification using
real-time analysis on the ABI PRISM 7900HT Sequence Detection
System with the TAQMAN RNase P Control Reagents Kit, both available
from Applied Biosystems, Foster City, Calif., permits optimal
amounts of DNA per reaction to maximize study efficiency. The
7900HT system can use a 5 microliter reaction volume and can
consume one nanogram of DNA per genotyping reaction, thereby
minimizing reagent costs and conserving DNA template samples.
[0105] The 5' nuclease assay can be suited for automation because
of its easy, three-step workflow. A universal master mix, including
probes and primers, can be added directly to plates of dry or fresh
DNA using standard robotics. Plates can be sealed and cycled using
standard thermal cyclers, such as, for example, the Applied
Biosystems Dual 394-Well GENEAMP PCR System 9700 thermal cycler,
available from Applied Biosystems, Foster City, Calif. Following
cycling, plates can be automatically read on the 7900HT that can
support the collection of more than 250,000 genotypes per day. In
addition, the availability of thermal cyclers with automated lid
handling can increase throughput by enabling robotics integration
for 24-hour unattended operation. Automation software can also
increase both quality and throughput. For example, automation of
allele calling can remove inter-technician variability, increasing
confidence in data quality and reducing the time spent on data
analysis by 8.5 person-hours per day. FIG. 3 provides an example of
an automated workflow system.
[0106] According to various embodiments, an assay that uses two
different types of probes can be provided wherein the
polynucleotide and the reporter dyes differ. For example, the first
type of probe can have a first polynucleotide with a VIC reporter
dye attached to the 5' end of the first polynucleotide, and the
second type of probe can have a second polynucleotide with a 6-FAM
reporter dye attached to the 5' end of the second polynucleotide,
and the first and second polynucleotides can differ by at least one
nucleic acid residue at the same location in the polynucleotide
when the polynucleotides are aligned 5' to 3'. The dye-labeled
probes can be adopted to perform a heterozygous assay or a
homozygous assay.
[0107] The probe can anneal to a complementary sequence between the
forward and reverse primer sites. At the time of annealing, the
probe is intact and the proximity of the reporter dye to the
quencher can result in suppression of fluorescence of the reporter
dye. A polymerase can cleave a reporter dye only when the probe has
completely, mostly, or substantially hybridized to the target DNA
sequence. When the reporter dye is cleaved from the probe, the
relative flourescence of the reporter dye increases. The increase
in relative fluorescence can be caused to only occur if the
amplified target DNA sequence is complementary, mostly
complementary, or substantially complementary to the probe.
Therefore, the fluorescent signal generated by PCR amplification
can indicate which alleles are present in a sample. Mismatches
between a probe and a target DNA sequence can reduce efficiency of
probe hybridization and/or a polymerase can be more likely to
displace a mismatched probe without cleaving it and therefore not
produce a fluorescent signal. For example, if one of two possible
reporter dyes fluoresce during an assay, then the presence of a
homozygous gene is indicated. For further example, if both possible
reporter dyes fluoresce during an assay, then the presence of a
heterozgous gene is indicated.
[0108] According to various embodiments, at least one primer can be
provided, wherein the primer can be a sequence that is shorter than
the target DNA sequence. The primer can have a polynucleotide
and/or a minor groove binder. The primer can be a sequence that is
complimentary to, or mostly complimentary to, the target DNA
sequence. The primer can be at least 90% homologous to a
corresponding length of the target DNA sequence, at least 80%
homologous to a corresponding length of the target DNA sequence, at
least 70% homologous to a corresponding length of the target DNA
sequence, or at least 50% homologous to a corresponding length of
the target DNA sequence.
[0109] According to various embodiments, a thermostable DNA
polymerase, such as, for example, thermus aquaticus (Taq), and at
least 4 embodiments of a deoxyribonucleic acid (e.g., adenosine,
tyrosine, cytosine, and guanine) can be provided. The polymerase
can be, for example, AMPLITAQ GOLD, available from Applied
Biosystems, Foster City, Calif. According to various embodiments,
components of a fluorogenic 5' nuclease assay or other assay
reagents that utilize 5' nuclease chemistry, for example, TAQMAN
minor groove binder probes, available from Applied Biosystems,
Foster City, Calif., can be provided. Some or all of the
above-listed components can be replaced by or used with
commercially-available products, for example, buffers or AMPLITAQ
GOLD PCR MASTER MIX (Applied Biosystems, Foster City, Calif.).
[0110] According to various embodiments, a high-quality LD map of
validated SNPs was created by integrating information from both
public and private human genome efforts. A set of over 200,000
validated, easy-to-use, individual SNP assays and TAQMAN
ready-to-use assay reagents created by using methods according to
various embodiments can be provided. A minor groove binder and a
non-fluorescent quencher, and the integration of the 5' nuclease
chemistry with an automated detection system, such as, for example,
the 7900HT, can be used. According to various embodiments. A
web-based bioinformatics and ordering system can be provided where
a customer can search for SNPs and order assay reagents, thus
reducing the time and costs associated with candidate-gene and
candidate-region association studies. According to various
embodiments, and LD map can, for example, enable candidate-gene and
candidate-region association studies using 5' nuclease chemistry
and/or be implemented on an ultra-high throughput SNP genotyping
platform to enable WGA studies. According to such embodiments, the
5' nuclease chemistry system can leverage the specificity of the
OLA-PCR assay chemistry and the highly parallel detection of, for
example, BEADARRAY technology, available from Illumina, Inc., San
Diego, Calif. According, to such embodiments, the system can enable
the generation of about 2,100,000 genotypes per day and all
components of the assay can be universal except for, for example,
the SNP-specific OLA probes.
[0111] According to various embodiments, assays for over 4,000,000
SNPs from the Celera database can cover every gene in the human
genome. According to various embodiments, many SNPs can have the
necessary variability for genetic association studies and assays
for the SNPs can be provided. According to various embodiments,
assays can be grouped together into convenient SNP sets optimized
for specific assays such as, for example, p450 genotyping and
disease-specific gene studies.
[0112] FIGS. 2a-2e are schematic diagrams showing the interaction
of components that can be part of a mixture of reagents according
to various embodiments. In FIG. 2a, primer 52 has annealed to
template strand 54. Replication of the template strand from primer
52 will occur in the 5' to 3' direction. Probe 50, including a
generic reporter dye R, quencher Q, and minor groove binder MGB,
has annealed to the template strand 54. Arrow 53 shows that as the
complementary strand (not shown) is produced from the template
strand 54 starting at the forward primer 52, the complementary
strand will meet probe 50. FIG. 2b shows the complementary strand
55 as it meets probe 50a. Polymerase 60 cleaves VIC reporter dye V
during the production of complementary strand 55 given that probe
50a has annealed to the target strand 54 because the target strand
54 and the probe 50a are completely complementary. FIG. 2c shows
the complementary strand 55 as it meets probe 50b. Polymerase 60
does not cleave FAM reporter dye F during the production of
complementary strand 55 given that probe 50b has not hybridized
with the target strand 54 because of a mismatched base pair at
location 64. FIG. 2d shows the complementary strand 55 as it meets
probe 50b. Polymerase 60 cleaves FAM reporter dye F during the
production of complementary strand 55 given that probe 50b has
annealed to the target strand 54 because the target strand 54 and
the probe 50b are completely complementary. FIG. 2e shows the
complementary strand 55 as it meets probe 50a Polymerase 60 does
not cleave VIC reporter dye V during the production of
complementary strand 55 given that the probe 50a has not hybridized
with the target strand 54 because of a mismatched base pair at
location 66.
[0113] FIG. 7 is an illustration of SNPs selected by a method
according to various embodiments.
[0114] According to various embodiments: a library is provided that
contains a plurality of data sets, corresponding to one or more
respective oligonucleotides that can function as a respective assay
to hybridize with at least one respective Single Nucleotide
Polymorphism (SNP) in a nucleic acid sequence. The nucleic acid
sequence can include three or more adjacent SNPs and the data sets
can correspond to at least the three or more adjacent SNPs,
respectively. Each adjacent SNP can be spaced a distance from at
least one other adjacent SNP and each of the distances between
adjacent SNPs can be from about 75% to about 125% of an average of
the distances between the adjacent SNPs. The corresponding,
adjacent SNPs can be spaced apart along at least a region of a
chromosome. The corresponding, adjacent SNPs can be spaced apart
along at least a region of a gene. The distances between all
corresponding, adjacent SNPs can be equal, plus or minus 30%. The
distances between all corresponding, adjacent SNPs can be equal,
plus or minus 20%. Each of the distances between corresponding,
adjacent SNPs can be from about 90% to about 110% of an average of
the distances between the adjacent SNPs. Each of the distances
between corresponding, adjacent SNPs can be from about 95% to about
105% of an average of the distances between the adjacent SNPs. The
nucleic acid sequence can be a consensus sequence. The distances
between all corresponding, adjacent SNPs can be less than a
specified maximum distance. The specified maximum distance can be,
for example, 10 kilobases, 15 kilobases, 20 kilobases, or 30
kilobases. According to various embodiments, the algorithm can
select a minimum number of SNPs per region. For example, three SNPs
per gene can be selected. The distances between all corresponding,
adjacent SNPs can be greater than a specified minimum distance. The
nucleic acid sequence can be a consensus sequence corresponding to
the human genome. The nucleic acid sequence can be a nucleic acid
sequence data set.
[0115] The library can comprise a number of data sets corresponding
to not more than a sufficient number of oligonucleotides necessary
to provide a collection of assays that can provide a maximum
statistical loss of haplotype diversity, across the region, of less
than ten (10) percent. The sufficient number of oligonucleotides
can be obtained by providing a matrix comprised of data
representing haplotype blocks and SNP locations. The columns of the
matrix can contain data representing existence of respective SNPs
within a haplotype block and the rows of the matrix can contain
data representing respective haplotype blocks. At least one column
can be eliminated, wherein elimination of the at least one column
may not reduce the number of rows in the matrix that contains
non-duplicative information. At least one column of the matrix that
is identical to a second column of the matrix and/or completely
opposite to a second column of the matrix can be eliminated.
[0116] FIG. 7 illustrates SNPs that can be selected according to
various embodiments. SNPs 710, 720, 730, and 740 that are present
in region 702 of nucleic acid sequence 700 have been selected based
on prioritization criteria. SNPs 710, 720, 730, and 740 are
separated from adjacent SNPs by a distance no greater than distance
760. The effective range of usefulness of each SNP is equal to plus
or minus one half of distance 760. Therefore, there are no gaps in
coverage of region 702 of nucleic acid sequence 700. SNPs 750, 752,
754, 756 were not selected using the same selection criteria. The
selection criteria can be used to select data sets for the library,
where the data sets correspond to one or more respective
oligonucleotides that can function as a respective assay to
hybridize with at least one respective SNP. According to various
embodiments, at least one selection criterion of the selection
criteria can be used. One criterion can be a specified maximum
distance between adjacent SNPs. Another criterion can be a
specified minimum distance between adjacent SNPs. A criterion can
be a target distance between adjacent SNPs. The actual distance
between adjacent SNPs can vary from the target distance by, for
example, 5%, 10%, 20%, or 30%. An algorithm to prioritize SNPs can
select as few SNPs as possible to achieve a coverage of the region
where the largest gap between adjacent. SNPs is less than or equal
to the specified maximum distance. An algorithm to prioritize SNPs
may not select a SNP if the distance between one of two adjacent
SNPs is less than the specified minimum distance.
[0117] Another prioritization criterion can involve using known
SNPs. Known SNPs can be preferentially selected to include in the
library because, for example, known SNPs are well-characterized,
and may have assays that have previously been designed to target
such known SNPs, they may have little or no additional cost
associated with providing an oligonucleotide or a data set
corresponding to an oligonucleotide for an assay directed to the
SNP. An algorithm to prioritize SNPs can consider known SNPs as
"must use" SNPs or as "prefer to use" SNPs.
[0118] According to various embodiments, a known SNP is a
previously characterized SNP marker that has low or no cost
associated with designing and/or manufacturing one or more
oligonucleotides. Known SNPs can be preferentially used as a
selection criterion. A newly identified SNP is a SNP marker that is
relatively uncharacterized and therefore there is a higher relative
cost associated with designing and/or manufacturing one or more
oligonucleotides. A selection criterion can be newly identified
SNP. Newly identified SNPs can be selected after all or some of the
known SNPs have been selected because, for example, newly
identified SNPs require investigation into whether a functional
assays can be produced for the newly identified SNPs and therefore
newly identified SNPs have greater costs associated with them than
known SNPs. An algorithm to prioritize SNPs can consider newly
identified SNPs as "must never use" SNPs, "use only after using all
known" SNPs, or as "try not to use" SNPs.
[0119] For example, according to various embodiments, using at
least one selection criterion, data sets that correspond to one or
more respective oligonucleotides that function as a respective
assay to hybridize with at least one selected SNP, were selected.
Selection criteria can be used to select data sets corresponding to
the largest number of known SNPs and the smallest number of newly
identified SNPs that provide a coverage of the region, where the
distance between adjacent, selected SNPs is approximately
equal.
[0120] According to various embodiments, a specified maximum
distance can be allowed between two adjacent selected SNPs, as well
as between the two outermost selected SNPs and the boundaries of
the region, gene, chromosome, or other boundary denoting the
beginning and ending of the nucleic acid sequence. One requirement
can be, for example, to never have a distance between adjacent
selected SNPs that is greater than a maximum required distance,
unless maintaining the maximum required distance is impossible
because there are no adjacent SNPs that fall within the maximum
required distance. For further example, the largest distance
between adjacent selected SNPs can be as small as possible.
[0121] According to various embodiments, an algorithm can be
provided that selects, based on prioritization criteria, a
plurality of data sets that corresponds to one or more respective
oligonucleotides that can function as a respective assay to
hybridize with at least one selected SNP. A region can be at least
a part of the nucleic acid sequence. The region can be separated
into sub-regions. The sub-regions can, for example, be separated by
known SNPs, if any. Each sub-region can be solved and optimized
independently according to various embodiments of the algorithm.
The sub-region can be bound by known SNPs. If there are no known
SNPs, then the region may not be separated into sub-regions.
Various embodiments of the algorithm can be repeated for each
sub-region. Each sub-region can begin, for example, with the
region's 5' end or a known SNP and can end with the 3' end or a
known SNP.
[0122] According to various embodiments, a step of the algorithm
can be a locally optimal solution ("local step") that can determine
the number of newly identified SNPs that can be utilized. The step
can include selecting all the newly identified SNPs. The step can
include selecting the known SNPs closest to one or both ends of the
region or sub-region of the nucleic acid sequence. The step can
include selecting the newly identified SNPs closest to one or both
ends of the region or sub-region of the nucleic acid sequence. The
step can iteratively eliminate newly identified SNPs until, for
example, no newly identified SNPs can be eliminated. The step can
iteratively eliminate newly identified SNPs until, for example, the
remaining number of known SNPs and newly identified SNPs is at or
below a minimum number of SNPs, if any, that can be selected.
[0123] For the local step, the distance between all adjacent SNPs
can be calculated. The first and/or last distances between adjacent
SNPs in a region or sub-region can be doubled if the distances are
between a SNP and the edge of the region or sub-region. From the 5'
end of the region or sub-region, the consecutive distances between
adjacent SNPs can be added until the cumulative sum of consecutive
distances exceeds the specified maximum distance. If the cumulative
sum of consecutive distances exceeds the specified maximum distance
after more than one consecutive distance, i.e., more than just two
adjacent SNPs, or more than just one SNP and an edge of a region or
sub-region, then one SNP in the set of newly identified SNPs can be
removed.
[0124] According to various embodiments, the local step can include
identifying the smallest distance between adjacent SNPs in the
sequence of adjacent SNPs that makes up the cumulative sum, where
the cumulative sum of adjacent distances is greater than the
specified maximum distance. If there is only one distance between
adjacent SNPs having a distance equal to the smallest distance,
then one of the two adjacent SNPs bounding the smallest distance
can be eliminated if one of the SNPs is a newly identified SNP. For
example, if both SNPs in the pair of adjacent SNPs having a
distance equal to the smallest distance are newly identified SNPs,
one of the SNPs can be eliminated at random or the SNP nearest the
5' end or the SNP nearest the 3' end can be eliminated by
convention. Because each distance value is the distance between two
adjacent SNPs, each SNP can have two distance values associated
with it. The distance value (5') is the distance to the adjacent
SNP on the 5' end of the nucleic acid sequence, and the distance
value (3') is the distance to the adjacent SNP on the 3' end. If
there is more than one distance between adjacent SNPs having a
distance equal to the smallest distance, e.g. a "tie," then the tie
can be broken by, for example, choosing the smallest distance on
the 3' end of the nucleic acid sequence. For another example, the
smallest distance can be chosen arbitrarily (e.g. the smallest
distance closest to the 5' end, the smallest distance closest to
the 3' end, or the smallest distance that falls in the middle of
the other smallest distances). After removing a SNP adjacent to the
smallest distance, the cumulative sum can be recomputed and the
process can be reiterated using the remaining SNPs. According to
various embodiments, if no newly identified SNPs can be removed,
the process of the local step can be stopped. If the total number
of SNPs is at or below a specified minimum number of SNPs, the
process of the local step can be stopped.
[0125] According to various embodiments, a step of the algorithm
can be a globally optimal solution ("global step") that can
determine the optimum selection of newly identified SNPs given the
number of newly identified SNPs that can be utilized. For example,
the number of newly identified SNPs that can be utilized can be
provided by the local step. According to various embodiments, at
least one step of the algorithm is performed.
[0126] According to various embodiments, if there are K newly
identified SNPs in the region or sub-region, and N SNPs will be
selected, then there can be P=K!/[N!(K-N)!] possible selected SNPs.
N can be a value from 1 to K. The value K can be selected by
determining the specified minimum and/or specified maximum
distances of the region or sub-region. According to various
embodiments, the maximum amount of time T that is dedicated to
calculating the global step can be specified. If a computer can
compute one selection in time t, then if P<T/t, the global step
can be performed. If P>=T/t, then the global step may not be
performed. The global step can include calculating all possible
selections P of the number of newly identified SNPs to be selected
N out of the total number of newly identified SNPs in the region or
sub-region K. The global step can include calculating the largest
distance between adjacent, selected SNPs for each selection P. The
global step can include choosing the selection P with the
"smallest" largest distance between adjacent, selected SNPs. The
global step can be illustrated by the following operation:
8 Min { Max [ SNP.sub.i - SNP.sub.i-1 ] } Over All i=1,...,N
Selections Of N out of K
[0127] The smallest value of N can be selected where Min {Max
[SNP.sub.1-SNP.sub.i-1]} (where i=1, . . . , N), that is less than
T.
[0128] According to various embodiments, a specified minimum
distance between adjacent selected SNPs can be specified. A minimum
number of total markers can be specified. Prioritization criterion
can be assigned to different, respective newly identified SNPs that
can assign preference to some newly identified SNPs over other
newly identified SNPs. To assign preference to some newly
identified SNPs, the distance value associated with a "high
priority" newly identified SNP can be, for example, marked or
changed so that the "high priority" newly identified SNP is always
preferred or selected over a "low priority" newly identified
SNP.
[0129] FIG. 8 details a hypothetical chromosome having 969 genes on
the chromosome. Of those 969 genes, 1,639 known SNPs are present on
the chromosome and are well characterized. The chromosome contains
11,095 newly identified SNPs that are not well characterized. Of
the 969 genes, 611 genes contain SNPs and 358 genes do not contain
SNPs. Of the 611 genes containing SNPs, FIG. 8 lists the average
gene length, in bases, the number of newly identified SNPs per
gene, the number of known SNPs per gene, the total number of
selected SNPs per gene that were selected using various
embodiments, and the number of newly identified selected SNPs per
gene that were selected using various embodiments. FIG. 9 is a
histogram of gene lengths of the genes found on the hypothetical
chromosome of FIG. 8. FIG. 10 is a histogram of the specified
maximum distance between adjacent SNPs, according to various
embodiments, of the selected SNPs from the hypothetical chromosome
of FIG. 8. FIG. 11 is a histogram of the actual maximum distance
between adjacent SNPs, according to various embodiments, of
selected SNPs of the hypothetical chromosome of FIG. 8. FIG. 12 is
a histogram of total selected SNPs per gene, according to various
embodiments, from the hypothetical chromosome of FIG. 8. FIG. 13 is
a histogram of the number of newly identified SNPs per gene that
were selected from the hypothetical chromosome of FIG. 8 using
various embodiments.
[0130] The present invention relates to the foregoing and other
embodiments as will be apparent to those skilled in the art from
consideration of the present specification and practice of the
present invention disclosed herein. It is intended that the present
specification and examples be considered as exemplary only with a
true scope and spirit of the invention being indicated by the
following claims and equivalents thereof.
* * * * *