A MICROARRAY-BASED APPROACH TO INFLUENZA A DIAGNOSIS AND SURVEILLANCE

Human Influenza A causes significant morbidity in humans despite the best efforts of prophylaxis. The virus is able to improve its virulence and to evade immune recognition through an evolutionary process that continually generates novel sequences of its surface glycoproteins hemagglutinin (HA) and neuraminidase (NA). An understanding of this virus’s phylodynamics is essential to a meaningful surveillance effort. In addition, this type of approach can lead to a diagnostic device that would allow the all-important distinction between mild, seasonal isolates and the deadly emerging forms that claim many lives each year.

During an outbreak of influenza, it is essential that patients be diagnosed and classified as accurately and as quickly as possible. Quarantine and treatment decisions when made correctly are critical to successful containment. We have developed a microarray-based device that can subtype patient isolates within 4 hours of sample collection. In addition, this system is able classify patient isolates into groups based on the underlying sequence diversity. After subtyping, further classification can be accurately performed by using microarray probe intensity hierarchical clustering. This process uses a novel probe choice algorithm, and analysis method that exploits the descriptive data provided by a microarray. The probe choice strategy creates overlapping sets of probes that define emerging subclusters of influenza through the associated sequences that they match. Because the probe set often includes probes from many parts of each subcluster, the system is much more sensitive to the type of mutational drift that takes place in these populations.  In addition, it allows isolate hybridizations to be clustered based on their probe intensities. Since multiple differentiating probes are used, the assay becomes much more sensitive to subtle mutational drift, allowing the discovery of novel sequence clusters well before sequence data becomes available. We have used this method to cluster hybridization data from human influenza samples provided by a number of surveillance laboratories, and we present this data below.