Literature review

Task 1 Topic: VIPR: A probabilistic algorithm for analysis of microbial detection microarrays (BMC Bioinformatics 2010, 11:384)

Task 2 Summary

Today, the clinical diagnostic assays in use focus on detecting the presence of a single, well defined target agent or a set of agents. In recent years, microarray-based diagnostics have been developed that greatly apply the highly parallel detection of multiple micro-organisms that may be present in a clinical specimen.

A novel interpretive algorithm, VIPR (Viral Identification using a Probabilistic algorithm), which uses Bayesian inference to optimize detection sensitivity by capitalizing on empirical training data. To illustrate this approach, the detection of viruses that cause hemorrhagic fever using a custom HF-virus microarray had been used to demonstrate. HF viruses belong to four virus families: Arenaviridae, Bunyaviridae, Flaviviridae and Filoviridae. A custom microarray was designed to detect all known HF viruses and closely related viruses. Furthermore, VIPR’s performance was compared with that of the existing interpretive algorithms that are not capable of utilizing training data in this fashion.

The probes were selected 35, 45 or 60 nucleotides in length which were designed to bind to viral genomes from the four families that contain all viruses known to cause HF. A total of 51 strains of 33 distinct virus species were obtained. In total, 110 hybridizations were performed including 102 positive controls, 4 Vero negative controls and 4 c6/36 negative controls. All raw microarray data are available in NCBI GEO.

Normalization was preformed to account for variation in reagent concentration or fluorescence across the microarray. Log transformation of the data was desirable for the estimation of normal distributions. Candidate genomes to be scored in the VIPR algorithm were limited to all complete genomes in the NCBI virus RefSeq database.  The entire set of oligonucleotide probe on the microarray was aligned using BLASTN against each of the RefSeq viral genomes.

The comparison of VIPR with the existing algorithm for analyzing diagnostic microarrays, the results of accuracy are 94%, 83%, 61% and 49% to VIPR, DetectiV, E-Predict and PhyloDetect respectively. The high accuracy represent a proof of VIPR will be useful in clinical laboratory settings to analyze microarrays.

Task 3

VIPR focus on detecting the presence of known virus, is it possible to apply to detect a set of genes in the bacteria?

Objectives: I would like to use the microarray technology to detect the multiple antibiotic resistant genes in the pathogens. It can provide a guide to the clinician for selecting an appropriated antimicrobial therapy.