Rapid and reliable diagnostics for infectious diseases are essential for saving human lives and prev

1
Calculating Ultraspecific Signatures
Our Polymerase Chain Reaction (PCR) based assay
consists of 20 primer-pairs. The 8 primer-pairs
cover the most strains are enough to cover 90 of
all HIV strains. Our microarray bases assay
consists of 7 probes. The 4 probes cover the most
strains are enough to cover over 90 of all HIV
strains.
Host/Background Sequence
Rapid and reliable diagnostics for infectious
diseases are essential for saving human lives and
preventing epidemics. The majority of new
pathogen identification and diagnosis assays have
been developed using nucleic acid-based
technologies. The key challenge for all of these
technologies is that primers and probes must be
computationally designed to be specific for the
intended target and not misprime (interfere) with
non-target DNA/RNA which will cause a false
positive outcome of the test. Because the vast
majority of the currently available primer/probe
design algorithms rely on heuristics,
computationally designed diagnostic assays
require extensive experimental validation.
Computational tools have been developed within
the University of Houston Bioinformatics
Laboratory that make it possible to exhaustively
rather than heuristically design highly specific
(ultraspecific) candidate genomic signatures
for the identification of any organism of
interest in the presence of a significantly
greater amount of known genomic background
material, e.g. the detection of viral genomic DNA
in the presence of human DNA. The human
immunodeficiency virus (HIV) was chosen as one of
the first applications for developing diagnostic
assays using this approach due to the urgent
need, especially in developing countries, for a
fast and inexpensive diagnostic. Herein, we
present the result of a primer-based and a
probe-based assay for the detection of HIV-1 in
which each primer/probe is 2 mutations
(insertions, deletions and substitutions at any
position) away from any subsequence present in
the human genome (including SNPs and ESTs) as
well as several opportunistic organisms which
typically co-infect immunocompromised
individuals.
Pathogen Sequences 1-Change Away
Average Number of Ultraspecifc N-mers 2-Change
Away In HIV Strains
Criteria for Primer/Probe Design
2-Change Away
PCR and Microarray Based Assays
A set of ultraspecifc signatures were computed
using 857 strands of HIV as the pathogen against
the background sequences including the human
genome, the single nucleotide polymophisms (SNPs)
and the expressed sequence tags (ESTs) of human,
and organisms which cause common co-infections
including Candidasis, Hepatitis C, Solmonella,
Turberculosis, HPV, Pneumonia, Herpes, and
Kaposis sarcoma.
Next, we screened the ultraspecific signatures
for suitable assay candidates. While all nucleic
acid-based assays are founded on the principles
of hybridization, the different platforms vary in
their abilities to accommodate certain properties
of signature sequences. For instance,
amplification-based assays necessitate the use of
two or more signatures that hybridize within a
certain physical distance of each other. Other
constraints include signature length, melting
temperature (Tm), difference in Tm between
signatures, and length of the longest
polynucleotide stretch. In any nucleic
acid-based assay design, particular signatures
(despite their specificity to the target) will
not hybridize to the target, but will instead
hybridize with themselves or with other
signatures in the assay (primer dimer formation).

Finally, a set of primers/probes were selected
to be included into our assays to detect all of
the HIV strains such that the number of
primers/probes is minimized. When all signatures
which are both ultraspecific and accommodating of
assay constraints are considered it is often the
case that some do not contribute to the
informativity of the assay. For example, two
primer pairs may amplify the same subset of
target strains. As the cost (monetary, labor,
and time) of a nucleic acid-based assay is
directly proportional to the number of signatures
employed, it is usually preferable to design
assays that contain as few signatures as
possible. This problem can be formulated as an
instance of the set covering problem. Computing
the optimal (minimal) set of signatures required
for the final assay in a realistic amount of time
is nontrivial (NP-complete). Thus, we use a
greedy approach for identifying the solution to
the minimum set cover problem. This is made
possible by the development of several new
algorithms and data structures.