|
|
 |
 |
 |
|
| |
DNA Sequence Analysis Method
Correlagen uses two different DNA sequencing methods in
its assays, known as Sanger sequencing and next-generation sequencing-by-synthesis.
Each requires a different method for preparing the patient DNA for sequencing,
and each has its own strengths and limitations (see below). Sanger sequencing
is currently considered the “gold standard” for sequence analysis because
of its high sensitivity for detecting most small variations and its low
false positive rate. However, Sanger sequencing is best suited for analyzing
small numbers of genes, for a “narrow and deep” genetic analysis. In contrast,
next-generation sequencing offers the capacity for analyzing very large
numbers of genes simultaneously, thereby increasing the clinical sensitivity
of a test, although with currently somewhat lower analytical sensitivity–
for a “broader and more shallow” genetic analysis. Next-generation sequencing,
at the moment, shows lower senstivity for detecting variants such as deletion
and/or insertion variants spanning several nucleotides or variants that
occur in duplicated gene regions. For substitution variants, however, which
account for the vast majority of variants found in clinical sequencing tests,
sensitivity of next-generation sequencing is close to that of Sanger sequencing.
In addition, sensitivity limitations of next-generation sequencing are mostly
based on current data analysis methods and may be overcome by re-analysis
of existing data sets once the analysis methods have further evolved. Improved
data analysis methods may then also allow detection of variants that cannot
be detected by Sanger sequencing, such as very large deletion and/or insertion
variants (copy-number variants) or variants present in only a subset of
gene copies tested (mosaic or somatic variants).
For the index patient, sequence is typically determined for all coding exons
of a gene that are represented in the most prevalent mRNA isoform or, if
known, the mRNA isoform most relevant to the expression of the disease phenotype.
For genes with more than one exon, flanking intronic sequences containing
the two highly conserved splice sites are also analyzed. Additional intronic
sequences and untranslated (UTR) sequences upstream and downstream of the
first and last, respectively, coding exon are generally, but not always
also analyzed, unless they are known to contain pathogenic variants When
confirming presence or absence of certain variants, eg, for the purpose
of family testing, sequencing may be limited to specific amplicons.
Sanger Sequencing
In Sanger sequencing, the region of interest is broken
up into small units of 200 to 600 base pairs (usually corresponding to exons),
and each unit is enriched and sequenced separately from all other units.
Enrichment is achieved by employing a technique known as polymerase chain
reaction (PCR) that leads to exponential amplification of DNA sequences
located between two specific oligonucleotides (primers). The PCR products
(amplicons) are then sequenced bi-directionally by extending primers bound
to the end of the PCR product, randomly terminating the extension process
at every possible position in the extension product through using a mixture
of normal nucleotides and fluorescently labeled “chain-terminator” nucleotides,
and determining the length of each extension product and the nature (A,
G, T, or C) of the terminal base at that length. Occasionally, the sequence
can only be determined in one direction, since small deletions or insertions
at the beginning of an extension product render the downstream sequence
uninterpretable.
Return to top
Limitations of Sanger Sequencing
- The method does not allow any conclusion as to whether
a heterozygous variant is present on the maternal or the paternal chromosome
copy. For this reason, the DNA sequence analysis performed here cannot
determine if two different heterozygous variants are located on the same
or on different chromosome copies. For the purposes of this report, two
different heterozygous variants are "by default" assumed to be located
on separate chromosome copies. Parent testing can be used to determine
if two different heterozygous variants are located on the same or on different
chromosome copies. If both variants are inherited from the same parent,
they are likely to be located on the same chromosome copy. If each variant
is inherited from a different parent, they are likely to be located on
different chromosome copies.
- The method does not reliably detect mosaic variants.
The sequencing output reflects the sum of all sequence versions present
in the PCR product. Presence of a variant in only some of the templates
will lead to a mixed-base signal at the variant position. While heterozygous
variants, which are present in about half of all templates, can be detected
with 99% reliability by the software algorithm used, mosaic variants,
which could be present in only a small proportion of templates, may or
may not be detected.
- The method cannot detect large deletions. If
one or both of the primer binding sites for an amplicon are deleted from
a template, eg, as part of a large deletion, the amplicon cannot be generated
from this template. If all template versions carry the deletion, no PCR
product will be generated. If, for example, only the template derived
from the maternal chromosome copy carries the deletion, the amplicon can
still be generated from the paternal chromosome copy. The only indication
that sequence was derived only from the paternal chromosome copy would
be that all variants detected on that amplicon would appear homozygous
in the final sequence output. The prevalence of large deletions varies
widely between genes.
- The method cannot detect large duplications, inversions,
or other re-arrangements. Re-arrangements that disrupt an amplicon
will not serve as a template during PCR. Re-arrangements,such as inversions,
that preserve an amplicon will not affect generation of the PCR product
in any way and will therefore not be detectable through the sequencing
output. Duplications will also not be detected, since the sequence output
does not allow any conclusion about the number of template copies present.
- The method is affected by allele-dropout. If
a template contains a variant in a primer binding site for an amplicon,
the amplicon cannot be generated from this template. If, for example,
the template derived from the maternal chromosome copy carries a variant
in a primer binding site, the amplicon can only be generated from the
paternal chromosome copy. In this case, the only indication that the PCR
product and the sequence derived from it reflect only the paternal chromosome
copy would be that all variants detected on that amplicon would appear
homozygous in the final sequence output. Allele dropout should be a rare
event, since primer binding sites are specifically chosen not to cover
any known variant location.
- The method may not be able to determine the exact
numbers of T/A or microsatellite repeats. During PCR and/or during
the sequencing reaction, the polymerase may slip on a long stretch of
T’s (or A’s) or a microsatellite (such as a CA repeat) in the template,
leading to a variable number of T’s (or A’s) or a variable number of microsatellite-repeats
in the sequence output. Such slippage may prevent accurate determination
of the number of T’s (or A’s) or of microsatellite repeats actually present
in the template.
Return to top
Next Generation Sequencing-by-synthesis
In next-generation sequencing, which is also known as
“massively parallel sequencing,” all regions of interest are sequenced together,
and the origin of each sequence read is determined by comparison (alignment)
to a reference sequence. The regions of interest can be enriched together
in one reaction, or they can be enriched separately and then combined before
sequencing. Correlagen’s next-generation sequencing assay uses two different
methods for enriching DNA regions of interest. All DNA sequences deriving
from coding exons of genes included in the assay are enriched by bulk hybridization
of randomly fragmented genomic DNA to specific RNA probes. The same adapter
sequences are attached to the ends of all fragments, allowing enrichment
of all hybridization-captured fragments by PCR with one primer pair in one
reaction. Regions that are only inefficiently captured by hybridization
are amplified by PCR with specific primers. In addition, PCR with specific
primers is also used to amplify exons for which similar sequences (“pseudo
exons”) exist elsewhere in the genome. This is necessary because hybridization
does not discriminate against pseudo exon-derived sequences, but will instead
capture all sequences that can hybridize to the specific RNA probes.
PCR products are concatenated to form long stretches
of DNA, which are sheared into short fragments by accoustic energy. This
step ensures that the fragment ends are distributed throughout the regions
of interest. Subsequently, a stretch of dA nucleotides is added to the 3’
end of each fragment, which allows the fragments to bind to a planar surface
coated with oligo(dT) primers (the “flow cell”). Each fragment is then sequenced
by extending the oligo(dT) primer with fluorescently-labeled nucleotides.
During each sequencing cycle, only one type of nucleotide (A, G, T, or C)
is added, and only one nucleotide is allowed to be incorporated through
use of chain terminating nucleotides. For example, during the 1st sequencing
cycle, a fluorescently labeled dCTP could be added. This nucleotide will
only be incorporated into those growing complementary DNA strands that need
a C as the next nucleotide. After each sequencing cycle, an image of the
flow cell is taken to determine which fragment was extended. DNA strands
that have incorporated a C will emit light, while DNA strands that have
not incorporated a C will appear dark. Chain termination is reversed to
make the growing DNA strands extendible again, and the process is repeated
for a total of 120 cycles. The images are converted into strings of bases,
commonly referred to as “reads,” which recapitulate the 3’ terminal 25 to
60 bases of each fragment. The reads are then compared to the reference
sequence for the DNA that was analyzed (eg, the human genome sequence if
human genomic DNA was used). Since any given string of 25 bases typically
only occurs once in the human genome, most reads can be “aligned” to one
specific place in the human genome. Finally, a consensus sequence of each
genomic region is built from the available reads and compared to the exact
sequence of the reference at that position. Any differences between the
consensus sequence and the reference are called as sequence variants.
Sensitivity of variant detection:
Sensitivity of variant detection largely depends on the
number of reads covering this position (known as “depth” or “coverage”),
ie, the amount of sequence information available for that particular position.
Since both the enrichment methods and the sequencing step are influenced
by the sequence context, coverage varies from region to region. In addition,
sensitivity of variant detection also differs by type of variant (substitution
versus deletion and/or insertion). At high coverages (≥30x), sensitivity
is currently 99% for detecting substitution variants, 90% for detecting
deletions and/or insertions spanning ≤ 5 bases, and approximately 30% for
detecting deletions and/or insertions spanning from 6 to about 40 bases.
Based on Correlagen’s data , insertions and/or deletions spanning ≤5 bases
or ≥6 bases account for about 10% and 1%, respectively, of all variant occurrences
and for about 16% and 2.6%, respectively, of all pathogenic variant occurrences.
Taking into account coverage at each base position within the sequenced
regions, length of the sequenced regions, and variant-type specific sensitivity,
Correlagen calculates and reports an overall sensitivity of variant detection
for each gene included in the assay. For example, if 80% of the analyzed
bases in a gene have a coverage corresponding to 97% sensitivity, 15% have
a coverage corresponding to 92% sensitivity, and 5% have a coverage corresponding
to 80% sensitivity, the overall sensitivity for that gene would be calculated
as 95%. Of note, exons with a sensitivity of less than 50% are not included
in the overall sensitivity estimate per gene, but are reported separately
as “segments not sequenced.”
False-positive rate of variant detection:
All sequence variants detected by next-generation sequencing
that are known or predicted to be pathogenic as well as all sequence variants
that are novel (ie, not previously described in the literature or a database)
are confirmed by uni-directional Sanger sequencing before reporting. Therefore,
the false positive rate of reported variants is expected to be very low.
Return to top
Limitations of Correlagen's Next-Generation Sequencing
Assay
- The method does not allow any conclusion as to
whether a heterozygous variant is present on the maternal or the paternal
chromosome copy. For this reason, the DNA sequence analysis performed
here cannot determine if two different heterozygous variants are located
on the same or on different chromosome copies. For the purposes of this
report, two different heterozygous variants are "by default" assumed
to be located on separate chromosome copies. Parent testing can be used
to determine if two different heterozygous variants are located on the
same or on different chromosome copies. If both variants are inherited
from the same parent, they are likely to be located on the same chromosome
copy. If each variant is inherited from a different parent, they are
likely to be located on different chromosome copies
- The method has limited sensitivity in duplicated
gene regions. Due to the relatively short length of the sequencing
reads, reads derived from duplicated gene regions cannot be unambiguously
mapped to the regions they were derived from, but may erroneously be
mapped to the duplicates of these regions.
- The method currently does not detect large deletions,
duplications, inversions, or other re-arrangements. In principle,
next-generation sequencing can detect many of these variants. Correlagen’s
current assay and analysis procedures, however, do not yet detect these
types of variations.
- The method currently does not reliably detect
mosaic variants. In principle, next-generation sequencing can detect
variants even if they are present only in a small percentage of template
molecules. Correlagen’s current assay and analysis procedures, however,
are primarily designed for detecting homozygous (present on all templates)
or heterozygous (present on 50% of templates) variants and may or may
not detect mosaic or somatic variants (present on <50% of templates).
- The method is affected by allele-dropout.
If a template contains a variant in a primer binding site for a PCR
product, the product cannot be generated from this template. If, for
example, the template derived from the maternal chromosome copy carries
a variant in a primer binding site, the PCR product can only be generated
from the paternal chromosome copy. In this case, the only indication
that the PCR product and the sequence derived from it reflect only the
paternal chromosome copy would be that all variants detected on that
amplicon would appear homozygous in the final sequence output. Allele
dropout should be a rare event, since primer binding sites are specifically
chosen not to cover any known variant location.
Of note, enrichment of DNA regions by hybridization capture and thus
Correlagen's next-generation sequencing assay overall is much less susceptible
to allele drop-out than enrichment by PCR.
- The method may not be able to determine the exact
numbers of T/A or microsatellite repeats. During PCR and/or during
the sequencing reaction, the polymerase may slip on a long stretch of
T’s (or A’s) or a microsatellite (such as a CA repeat) in the template,
leading to a variable number of T’s (or A’s) or a variable number of
microsatellite-repeats in the sequence output. Such slippage may prevent
accurate determination of the number of T’s (or A’s) or of microsatellite
repeats actually present in the template.
Sequence Variant Naming
Correlagen numbers and names all variants relative
to the human reference sequence published by http://genome.ucsc.edu in March
of 2006 (hg18) and according to the system suggested by the Human Genome
Variation Society (http://www.genomic.unimelb.edu.au/mdi/mutnomen),
regardless of whether the cited publication does or does not adhere to this
convention. According to the HGVS system, the start of the coding sequence
(ie, the "A" of the start codon ATG) is designated as +1. All
coding nucleotides, ie, all exonic nucleotides, in the designated mRNA isoform
are numbered consecutively. Intronic nucleotides are numbered relative to
the nearest exonic nucleotide.
| Variant
Numbering |
| Exon 1 |
Intron 1 |
Exon 2 |
Intron 2 |
Exon 3 |
| 5'UTR |
|
5'UTR |
Met |
Glu |
|
Val |
stop |
3'UTR |
| G |
A |
G |
G |
T |
A |
G |
G |
T |
A |
T |
G |
G |
A |
G |
G |
T |
A |
G |
G |
T |
A |
T |
G |
A |
G |
A |
| -5 |
-4 |
-3 |
-3+1 |
-3+2 |
-2-2 |
-2-1 |
-2 |
-1 |
1 |
2 |
3 |
4 |
5 |
6 |
6+1 |
6+2 |
7-2 |
7-1 |
7 |
8 |
9 |
10 |
11 |
12 |
13 |
14 |
Sequence variants are named according to the change
they cause in the DNA sequence. The most common types of changes are:
- Substitutions of one nucleotide for another nucleotide
(eg, c.3G>C).
- Deletions of one or more nucleotides (eg, c.4_6delGAG).
- Insertions of one or more nucleotides (eg, c.4_5insT).
- Substitutions of a group of nucleotides for a group
of different nucleotides, where the number of deleted and inserted nucleotides
can be different (eg, c.4_6delinsT).
Please click here for a more
detailed description of the numbering and naming rules used by Correlagen.
Mutation types reflect the predicted effect of a variant
on the mRNA or the protein level. The most common mutation types are:
- Splice-site mutations destroy an existing splice
site or create a new splice site. Both types of variations can lead to
altered mRNA processing and a dramatically different mature mRNA sequence,
which translates into a dramatically different protein sequence.
- Nonsense mutations introduce a stop codon in
the middle of the coding region, leading to truncation of the protein.
Nonsense mutations are commonly caused by a single-nucleotide substitution,
as shown in the example below:
| G |
G |
G |
T |
T |
G |
A |
A |
A |
A |
C |
A |
G |
C |
G |
| Glycine |
Leucine |
Lysine |
Threonine |
Alanine |
| G |
G |
G |
T |
A |
G |
A |
A |
A |
A |
C |
A |
G |
C |
C |
| Glycine |
stop |
|
|
|
- Missense mutations change one amino acid in
the protein into another. Missense mutations are commonly caused by a
single-nucleotide substitution, as shown in the example below:
| G |
G |
G |
C |
T |
T |
A |
A |
A |
A |
C |
A |
G |
C |
G |
| Glycine |
Leucine |
Lysine |
Threonine |
Alanine |
| G |
G |
G |
C |
C |
T |
A |
A |
A |
A |
C |
A |
G |
C |
C |
| Glycine |
Proline |
Lysine |
Threonine |
Alanine |
- Synonymous mutations do not cause a change in
the amino acid sequence of the protein:
| G |
G |
G |
C |
T |
T |
A |
A |
A |
A |
C |
A |
G |
C |
G |
| Glycine |
Leucine |
Lysine |
Threonine |
Alanine |
| G |
G |
G |
C |
T |
C |
A |
A |
A |
A |
C |
A |
G |
C |
C |
| Glycine |
Leucine |
Lysine |
Threonine |
Alanine |
- Frameshift mutations cause a shift in reading
frame, leading to a complete change of the amino acid sequence downstream
of the frameshift site. Since stop codons tend to be enriched in the two
unused reading frames, frameshift mutations often lead to truncation of
the protein. A frameshift mutation is caused by a net deletion or net
insertion of a number of nucleotides not divisible by 3. Of note, the
amino acid sequence may not change until several amino acids downstream
of the actual frameshift site, as shown in the example below:
| G |
G |
G |
C |
T |
T |
A |
A |
A |
A |
C |
A |
G |
C |
G |
| Glycine |
Leucine |
Lysine |
Threonine |
Alanine |
| G |
G |
|
C |
T |
T |
A |
A |
A |
A |
C |
A |
G |
C |
C |
| Glycine |
Leucine |
Lysine |
Glutamine |
Arg... |
- In-frame deletions and/or insertions lead to
deletion and/or insertion of one or more amino acids from/into the protein.
In-frame deletions and/or insertions do not alter the reading frame and
therefore do not change the amino acid sequence downstream of the deletion
and/or insertion site. In-frame deletions and/or insertions may or may
not lead to a missense mutation, as shown in the example of a 3-nucleotide
insertion (GCA) below:
| G |
G |
G |
C |
T |
T |
A |
A |
A |
A |
C |
A |
G |
C |
G |
| Glycine |
Leucine |
Lysine |
Threonine |
Alanine |
| G |
G |
G |
C |
G |
C |
A |
T |
T |
A |
A |
A |
A |
C |
A |
| Glycine |
Arginine |
Isoleucine |
Lysine |
Threonine |
Return to top
Correlagen’s Variant Scoring Method
Meaning of Correlagen’s Variant Scores:
Correlagen’s variant scores reflect the probability of
association with monogenic disease as well as the strength of the supporting
data, ie, the confidence that the score is correct (see Figure 1).
The variant scores do not reflect severity of disease.
The variant scores also do not reflect the probability of association with
disease in an oligogenic or polygenic rather than monogenic manner. In other
words, a variant score of "unlikely to be associated" does not
exclude the possibility that a variant may "weakly" contribute
to a disease in association with several other variants in the same or different
genes.
Correlagen’s variant scores are based on Correlagen’s
scoring algorithms and may differ from the variant scores proposed by the
authors of a publication. To request detailed information on how the score
for a specific variant found in your patient’s sample was derived, please
call Correlagen at 1-866-647-0735.
Return to top
How Are Correlagen’s Variant Scores Determined:
Correlagen’s variant scores are based on the following
considerations (summarized in Figure 2):
Has the variant been observed in the general population
or normal controls? If a variant is observed more frequently in the
general population than is compatible with the prevalence and mode of inheritance
of the disease, then this variant is assumed to be non-pathogenic. Data
for variant frequency in the general population are derived from dbSNP (NCBI
EntrezGene), from publications,2 or from prevalence studies conducted
at Correlagen. (2 peer-reviewed English-language publications
listed in PubMed
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi)
Has the variant been observed in affected individuals?
If a variant is observed only in diseased individuals and not in the general
(healthy) population, it is assumed to be at least possibly associated with
disease. The probability of association depends on such parameters as the
number of diseased individuals with the variant and the consistency of co-migration
of variant and disease within families. Data for variant frequency in affected
individuals are derived from the peer-reviewed published literature.
What effect does the variant have in a controlled experimental
system? A significant effect of a variant on the synthesis, cellular
location, and/or the function of the encoded protein in an experimental
system suggests that the variant is pathogenic. While experimental systems
can provide powerful information, the results must also be seen with caution,
since an experimental environment lacks many of the complexities of the
actual in-vivo environment. Data for variant effect in an experimental
system are derived from the peer-reviewed published literature.
What is the predicted effect of the variant on synthesis
and/or function of the encoded protein3? If the variant leads
to truncation of the gene product due a nonsense mutation or a frameshift
mutation, it is assumed to be pathogenic (for diseases related to loss-of-function
mutations). If the variant affects one of the highly conserved donor or
acceptor splice sites, it is predicted to lead to exon skipping and is assumed
to be pathogenic. If the variant leads to a missense variant, it is considered
to be possibly pathogenic, in absence of other information. If the variant
is located in the coding region away from exon/intron junctions and does
not lead to a change in the amino acid sequence (synonymous variant), it
is considered unlikely to be pathogenic, although pathogenicity cannot be
excluded. If a variant is located in the coding sequence close to an exon/intron
junction or in an intron away from the exon/intron junction, its effect
cannot be predicted, and it is classified as a variant of unknown significance.
3 Predictions are based on the mRNA isoform
chosen for reporting. Often, the same gene sequence can give rise to several
mRNA isoforms. For genes with many exons, different mRNA isoforms may contain
sequence from different permutations of exons. The exons reflected in a
particular mRNA isoform define the actual coding region and thus the predicted
effect of a variant on protein synthesis and/or function.
A number algorithms (eg, SIFT, PolyPhen, Align-GVGD)4
have been developed in an attempt to predict the impact of a missense variant
on protein function. Prediction algorithms are typically based on evolutionary
conservation, structure of the protein at the site of the variant, and/or
amino acid properties. While Correlagen routinely uses such algorithms for
variant evaluation, it does not base the variant score on a prediction from
any single one of the algorithms, since their specificity and sensitivity
are limited and their predictions frequently contradict each other.
4 http://blocks.fhcrc.org/sift/SIFT.html
http://coot.embl.de/PolyPhen/
http://agvgd.iarc.fr/agvgd_input.php
Additional considerations for scoring include co-occurrence
of a variant with known pathogenic variants, occurrence of a variant in
mutually exclusive disease phenotypes, predicted effect of synonymous variants
on splicing, and certain gene-specific and/or disease-specific properties.
Return to top
What is the Difference Between the Variant Score and
the Result Interpretation?
The variant score reflects the relationship of an individual
variant to a disease phenotype. The result interpretation considers the
variant scores of the two most significant variants in the context of gene/disease-specific-parameters
(eg, mode of inheritance) and patient-specific parameters (eg, variant zygosity
and patient sex). Two variants are considered, since an autosomal recessive
disease may be caused by a combination of two different heterozygous variants.
Significance of a variant is determined by the probability of its association
with the test phenotype. Eg, a variant scored as associated is more significant
than a variant scored as possibly associated.
The patient symptoms are assumed to be consistent with
the test indication. Consequently, if a variant is found that has not
previously been reported in association with any disease but that is predicted
to be pathogenic, this variant is assumed to be associated with the test
phenotype. If a patient’s symptoms are not consistent with the test indication,
the interpretation given in the report may be wrong.
The patient’s sex, if not specified, is assumed to
be consistent with the primary test indication, unless heterozygous
variants in an X-linked gene indicate female sex. This assumption is important
in the context of tests that are indicated primarily for one sex only. If
this assumption is wrong, the interpretation given in the report may be
wrong.
Can a negative result exclude disease in the patient?
Unless the patient was tested for a known familial variant, a negative result
(absence of pathogenic variants) cannot exclude disease. Instead, testing
may have failed to detect the pathogenic sequence variant causing the disease,
because
- Of the limitations inherent to the test method used.
- The variant is mosaic. Depending on which cell lineage
is affected by mosaicism and the ratio of cells containing and not containing
the variation, Correlagen’s test methodology may or may not be able to
detect a mosaic variant.
- The variant is located in a gene region not included
in the test.
- The variant is located in a gene not included in the
test.
Return to top
How Can Parent Testing Help To Interpret the Sequencing
Results for a Child?
Parent testing can help to determine if two heterozygous
variants are located on the same or on different chromosome copies in the
child. If each variant is inherited from a different parent, the variants
are likely to be located on different chromosome copies. If both variants
are inherited from the same parent, the variants are likely to be located
on the same chromosome copy. This information is often important since a
recessively inherited disease is only expressed if the variants are located
on different chromosome copies, and a dominantly inherited disease is often
more severe if the variants are located on different chromosome copies.
Information about whether two variants are located on the same or on different
chromosome copies is also relevant for genetic counseling or in a situation
where expression of the disease is influenced by whether a pathogenic variant
was inherited from the father or from the mother.
Return to top
How Can Family Testing Help To Interpret the Sequencing
Results for a Patient?
If a variant is suspected of being associated with disease,
this variant interpretation can be strengthened by detecting the variant
in all blood relatives who are affected with the disease but not, or only
very rarely, in blood relatives not affected with disease.
Return to top
Copyright © 2009, Correlagen Diagnostics, Inc. All rights reserved.
|
| |
|
 |
|
 | |
|
