- Open Access
E-Predict: a computational strategy for species identification based on observed DNA microarray hybridization patterns
© Urisman et al.; licensee BioMed Central Ltd. 2005
- Received: 26 April 2005
- Accepted: 26 July 2005
- Published: 30 August 2005
DNA microarrays may be used to identify microbial species present in environmental and clinical samples. However, automated tools for reliable species identification based on observed microarray hybridization patterns are lacking. We present an algorithm, E-Predict, for microarray-based species identification. E-Predict compares observed hybridization patterns with theoretical energy profiles representing different species. We demonstrate the application of the algorithm to viral detection in a set of clinical samples and discuss its relevance to other metagenomic applications.
- Respiratory Syncytial Virus
- Similarity Score
- Additional Data File
- Severe Acute Respiratory Syndrome
- Energy Profile
Metagenomics, an emerging field of biology, utilizes DNA sequence data to study unculturable microorganisms found in the natural environment. Metagenomic applications include studies of diversity and ecology in microbial communities, detection and identification of representative species in environmental and clinically relevant samples, and discovery of genes or organisms with novel or useful functional properties (for recent reviews, see [1–4]).
Common to all of these applications is the task of identifying (and often quantifying the abundance of) individual genes, species, or even groups of species from the large and often complex sequence space being explored. In the most general approach, shotgun sequencing is used to both identify and quantify individual sequences in a sample of interest [5–8]. In a more targeted approach, polymerase chain reaction (PCR) is used to amplify a particular subset of sequences, which can then be cloned and analyzed. For example, 16S rRNA sequences are frequently used to identify bacterial and archaeal species [9–12]. Another approach is based on functional screening of shotgun expression libraries to identify DNA fragments that encode proteins with desirable activities [13–15].
DNA microarrays are also emerging as an important tool in metagenomics [2, 16–18]. Particularly in applications concerned with real-time identification of known or related species, microarrays provide a practical high-throughput alternative to costly and time-consuming cloning and repetitive sequencing. For example, as previously reported, DNA microarrays have successfully been used to detect known viruses [19–22] and to discover a novel human viral pathogen . Other metagenomic applications in which microarrays have great potential include monitoring food and water quality , tracking bioremediation progress [2, 25], and assessment of biologic threat .
Use of DNA microarrays in metagenomics introduces a series of analytical challenges. First, the sequence space to explore may be very large, especially in the case of environmental samples. Given the technologic constraints on the total number of probes that can be placed on a microarray, improved algorithms are required for optimal probe selection to maximize coverage. Second, microarray data generated in metagenomic studies can be very complex. In the case of viral diagnostics, nucleic acid extracted from clinical specimens usually contains host and bacterial contaminants in addition to viral RNA and DNA. As a result, hybridization patterns are complicated by substantial amounts of noise introduced by specific and nonspecific cross-hybridization that cannot be anticipated or controlled. Third, multiple and potentially closely related species may be present in a single sample, resulting in complex or even overlapping hybridization patterns. Finally, a species identification strategy based on the use of experimentally derived patterns alone is not feasible, because such empirical controls can be obtained only for a limited number of species available as pure cultures or genomic clones. New analytical tools capable of overcoming these challenges are acutely needed.
We have previously reported the development of a DNA microarray-based platform for viral detection and discovery  (NCBI GEO , accession GPL366). Briefly, the platform employs a spotted 70-mer oligonucleotide microarray containing approximately 11,000 oligonucleotides, which represent the most conserved sequences from 954 distinct viruses corresponding to every NCBI reference viral genome available at the time of design. Nucleic acids are extracted from a sample of interest, typically a clinical specimen, and are amplified and labeled using random-primed reverse transcription, second strand synthesis, and PCR. The labeled DNA is then hybridized to the microarray, and hybridization patterns are analyzed to identify particular viruses that are present in the sample.
The E-Predict algorithm
Theoretical hybridization energy profiles were computed for every completely sequenced reference viral genome available in GenBank as of July 2004 (1,229 distinct viruses). This set of profiles included all viruses represented on the microarray and many viruses whose genomes became available after the array design had been completed. All microarray oligonucleotides expected to hybridize to a given viral genome were identified using nucleotide BLAST (basic local alignment search tool) alignment . Free energy of hybridization (ΔG) was then computed for each alignment using the nearest neighbor method [29, 30]. Oligonucleotides that failed to produce a BLAST alignment were assumed to have hybridization energies equal to zero. Thus, a given theoretical energy profile consists of the non-zero hybridization energies calculated for the subset of oligonucleotides producing a BLAST alignment to the corresponding genome. Collectively, the energy profiles of all the viruses constitute a sparsely populated energy matrix, in which each row corresponds to a viral species and each column corresponds to an oligonucleotide from the microarray (Figure 1b).
The general E-Predict algorithm for interpreting observed hybridization patterns is shown in Figure 1b. A vector of oligonucleotide intensities is normalized and compared with every normalized profile in the energy matrix using a simple similarity metric, resulting in a vector of raw similarity scores. Each element in this vector denotes the similarity between the observed pattern and one of the predicted profiles for a species represented in the energy matrix. The statistical significance of the raw similarity scores is estimated using a set of experimentally obtained null probability distributions. Profiles associated with statistically significant similarity scores suggest the presence of the corresponding viral species in the sample.
Normalization and similarity metric choice
In order to optimize the ability of E-Predict to discriminate between true positive and true negative predictions, we first evaluated the performance of several commonly used normalizations and similarity metrics. For this purpose we constructed a training dataset of 32 microarrays obtained from samples known to be infected by specific viruses. Fifteen microarrays represented independent hybridizations of RNA extracted from HeLa cells - a human cell line that is permanently infected with human papillomavirus (HPV) type 18. The remaining microarrays were obtained from 17 independent clinical specimens from children with respiratory tract infections. Ten specimens contained respiratory syncytial virus (RSV) and seven contained influenza A virus (FluA), as determined by direct fluorescent antibody (DFA) test.
With the exception of Spearman rank correlation, all considered metrics assigned the highest similarity scores to the match profiles on all 32 microarrays, independent of normalization choice. Not surprisingly, separation between interfamily profiles was greater than that between intrafamily profiles. In addition, changes in normalization and similarity metric had greater impact on intrafamily than on interfamily separation. The best overall separation was determined by calculating the product of the means of the intrafamily and interfamily separations divided by the corresponding standard deviations. Sum normalization of the intensity vectors, quadratic normalization of the energy vectors, and uncentered Pearson correlation as the similarity metric achieved the highest overall separation, producing a mean intrafamily separation of 0.69 (standard deviation 0.17) and a mean interfamily separation of 0.93 (standard deviation 0.08). Therefore, we settled on this combination of normalization and similarity metric parameters as our method of choice.
Raw similarity scores, as described above, provide an effective means of ranking viral energy profiles based on similarity to an observed hybridization pattern. However, such ranking provides no explicit information regarding the likelihood that viruses corresponding to the best scoring profiles are actually present in a sample under investigation. For example, two profiles may have identical high scores, but one of the scores may reflect a true positive whereas the other may be the result of over-representation of cross-hybridizing oligonucleotides in a profile.
To facilitate the interpretation of individual raw similarity scores, we sought to develop a test of their statistical significance. For this purpose, we obtained empirical distributions of the scores for every virus profile in the energy matrix. The distributions were based on 1,009 independent microarray experiments collected from a wide range of clinical and nonclinical samples representing different tissues, cell types, and nucleic acid complexities. Given such sample diversity, we assumed that any given virus was present in only a small fraction of all samples. Therefore, the empirical distributions are essentially distributions of true negative scores. The loge-transformed similarity scores were approximately normally distributed. Outliers on the right tails of the distributions, assumed to be true positives, were removed (see Materials and methods, below), and parameters of the null distributions were estimated as the mean and standard deviation of the remaining observations. These parameters were used to calculate the probability associated with any observed similarity score. Probabilities obtained this way should be interpreted as one-tail P values for the null hypothesis, that the virus represented by the profile is not present in the sample.
Our laboratory is conducting a series of studies focused on human diseases suspected of having viral etiologies. The E-Predict algorithm was developed to assist in the analysis of samples obtained as part of these investigations. As an illustration of its versatility we present four example applications of E-Predict, as it is used in our laboratory.
Example 1: Hepatitis microarray - predicted virus profiles
Hepatitis B virus
Woolly monkey hepatitis B virus
Woodchuck hepatitis B virus
Arctic ground squirrel hepatitis B virus
Ground squirrel hepatitis virus
Human herpesvirus 7
Example 1: hepatitis microarray - oligonucleotides contributing to hepatitis B virus profile prediction
Parental virus genome
Hepatitis B virus
Hepatitis B virus
Hepatitis B virus
Woolly monkey hepatitis B virus
Hepatitis B virus
Hepatitis B virus
Hepatitis B virus
Example 2 - FluA, RSV double infection
Influenza A virus
Influenza A virus
Influenza A virus
Human respiratory syncytial virus
Respiratory syncytial virus
Bovine respiratory Syncytial virus
Example 3: SARS microarray
Avian infectious bronchitis virus
Avian nephritis virus
Equine rhinitis B virus
Simian type D virus 1
Human coronavirus OC43
Simian type D virus 1
Mason-Pfizer monkey virus
Human herpesvirus 5
Human rhinovirus B
Tomato leaf curl Vietnam virus
Tomato yellow leaf curl Thailand virus
Tobacco leaf curl Kochi virus
Pepper leaf curl virus
In addition to the SARS CoV prediction mentioned above, several astrovirus and picornavirus profiles had similarity scores with significant P values. However, these predictions were based on oligonucleotides corresponding to a conserved 3'-untranslated region shared by these viruses with the SARS CoV [23, 33]. To identify incorrect predictions, such as these, resulting from partial profile overlaps with a match virus, we implemented an iterative version of E-Predict in which oligonucleotide intensities corresponding to the top scoring profile from one iteration are set to zero before running the next iteration. As a consequence, misleading predictions resulting from oligonucleotides shared with the top scoring profile fail to attain significant similarity scores in subsequent iterations. Conversely, only those predictions that are based on alternative oligonucleotides, namely predictions representing distinct species, remain. When iterative E-Predict was used on the SARS microarray, no astrovirus or picornavirus profile attained a statistically significant score (P > 0.04) in the second iteration, effectively removing these profiles from consideration. Complete E-Predict output for this example is available as Additional data file 3.
Identifying individual species present in a complex environmental or clinical sample is an essential component of many current and proposed metagenomic applications. Given a foundation of genomic sequence information, DNA microarrays are a high-throughput and cost-effective methodology for detecting species in an unbiased and highly parallel manner. Metagenomic applications employing DNA microarrays include characterization of microbial communities from environmental samples such as soil and water [2, 17], pathogen detection in clinical specimens and field isolates , monitoring of bacterial contamination of food and water , and detection of agents involved in potential cases of bioterrorism .
Despite the increasing use of DNA microarrays for species detection and identification, bioinformatics tools for interpreting hybridization patterns associated with complex clinical and environmental samples are lacking. Existing methods have utilized direct visual inspection of hybridizing oligonucleotides [23, 37] or inspection following clustering [19, 38]. Such methods are intractable for interpreting complex hybridization patterns, are time consuming, and suffer from user bias. Improved data interpretation tools must address several challenges. First, hybridization patterns may represent signal from dozens or even hundreds of species. Also, several closely related species may be present in a sample, giving rise to overlapping hybridization signals. A likely additional source of noise is unanticipated cross-hybridization, because many of the genomes present in a complex sample may be uncharacterized. Finally, obtaining pure samples of each possible species for the purpose of generating reference hybridization patterns is impractical or impossible in most cases.
When challenged with each of these problems, E-Predict proved to be a useful tool for interpreting hybridization patterns, correctly identifying viruses from diverse viral families present in a variety of clinical samples. In particular, E-Predict does not rely on the use of empirically generated reference hybridization patterns, because species identification is based instead on theoretical hybridization energy profiles. The energy profile matrix currently represents over 1,200 distinct viruses whose complete genomic sequences are known. As new viral genomes are sequenced, profiles are added to the matrix to broaden the range of species detection. For example, addition of the SARS CoV profile enabled accurate identification of the virus, even though no oligonucleotides derived from its genome were present on the microarray. Conversely, even when a perfectly matching profile is not available because of limited sequence coverage, E-Predict will identify the closest related species, as long as such species are represented on the microarray. This feature is particularly useful for detecting novel viruses as well as for discriminating between closely related viruses such as HRV serotypes. Naturally, maximum range and precision of detection is achieved through addition of new profiles and periodic microarray updates to include specific oligonucleotides from newly sequenced species.
E-Predict is also useful in overcoming problems related to nucleic acid complexity frequently encountered in clinical samples. For example, E-Predict correctly identified hepatitis B virus in a serum sample, despite the fact that the hybridization pattern was complicated by a low signal-to-noise ratio. In another example, E-Predict deconvoluted a complex hybridization pattern, correctly suggesting the presence of two viruses (FluA and RSV) in a nasopharyngeal aspirate sample. In yet another example, iterative application of E-Predict (see Materials and methods, below) to a hybridization pattern involving oligonucleotides derived from seemingly unrelated families (coronaviridae and astroviridae) premitted objective recognition that the pattern represented the presence of only one virus (SARS CoV).
Using a training dataset of 32 microarrays derived from samples known to contain specific viral species, we identified a set of normalization and similarity metric parameters, which yielded the best discrimination between true positive and true negative species predictions. The combination of sum normalization of the intensity vectors, quadratic normalization of the energy vectors, and uncentered Pearson correlation as the similarity metric was the optimal choice for our data. However, a different set of parameters may be required for applications that use a different nucleic acid amplification or detection strategy. An independent evaluation of potentially useful normalization and similarity metric parameters is therefore recommended for each specific application of the algorithm.
Using our best combination of normalization and similarity metric parameters, we obtained a set of null distributions representing true negative scores. These distributions were based on over 1,000 independent hybridizations and the assumption that the majority of samples were negative for the presence of any given virus. Although valid for our data, this assumption will not hold for all cases. For example, in applications concerned with bacterial species detection, some species may be present in most or even all samples and others encountered only rarely. In this case, a more complicated model will be required to assess whether a specific distribution represents negative, positive, or both negative and positive scores. For example, in cases in which distributions appear bimodal, one mode may represent true negatives and the other true positives. In some cases, targeted experimental verification of a subset of representative scores may be necessary. If both positive and negative score distributions are available, then P values can be calculated for each distribution.
Several modifications to the algorithm may potentially result in improved prediction accuracy. First, in the current implementation oligonucleotides exhibiting nonspecific cross-hybridization are filtered and the remaining oligonucleotides are weighted equally. Because oligonucleotides exhibit a continuous range of nonspecific hybridization [20, 30], a more sophisticated system of oligonucleotide weights may result in better performance. For example, using a procedure similar to that used to generate null distributions for the virus profile scores, empirical distributions can be obtained for individual oligonucleotide intensities, and individual oligonucleotide contributions may be weighted by the probabilities associated with the corresponding observed intensities. Such weighting may allow a more accurate assessment of significance.
Second, no attempt was made to normalize nucleic acid abundances of individual species, which may vary widely in different samples depending on factors such as target-to-background ratio, number of species present, and efficiency of nucleic acid extraction and amplification. Although individual nucleic acid abundances are difficult or impossible to estimate in most metagenomic applications, particularly before the corresponding species have been identified, in applications in which such estimates can be made, either experimentally or on theoretical grounds, the use of correction factors for calculating similarity scores or stratification of P value estimation may be needed. In addition, for highly abundant species, care should be taken to avoid saturation of individual oligonucleotides, because E-predict performance drops sharply after 20-25% of oligonucleotides in a given profile are saturated (data not shown).
Third, even though viral genomes were used as the basis for calculating energy profiles, the concept can easily be extended to other taxonomy nodes such as genera or families of viruses. This requires every sequence element to be classified at the appropriate node in the taxonomy hierarchy.
Finally, iterative use of E-Predict was intended for identification of multiple species that may be present in a sample. In this setting, it is important to distinguish between true predictions representing unique species present in the sample and misleading predictions arising from partially overlapping profiles. In each iteration it is assumed that the profile attaining the highest score corresponds to the species most likely to be among those present in the sample. When a novel species is present, this assumption may not hold because of limited oligonucleotide coverage. For instance, in the SARS CoV example, although SARS CoV attained a higher similarity score than mink astrovirus, the corresponding P values were comparable. However, even if mink astrovirus were the top prediction in the first iteration, SARS CoV would be the top prediction in the second iteration (P = 2 × 10-6; data not shown) and therefore would not be missed as a true positive. In our current studies P values in all iterations are estimated using the same set of null probability distributions. In addition, we use two iterations as our default, and essentially never need to run more than three iterations, because detection of more than two or three viruses is rare. However, iterative resolution of hundreds or thousands of species present in a sample may necessitate other normalization methods or adjustments to the null distributions for P value estimation. As an alternative, noniterative algorithms for analyzing overlapping profile signatures are also being explored.
In conclusion, E-Predict is a novel computational approach for species identification, which is generally applicable to a wide range of metagenomic applications using DNA microarrays. In particular, as more sequencing efforts are being directed at natural microbial communities, DNA microarrays are bound to become a central tool for various downstream applications such as identification of microbial species or detection of genes and biochemical pathways in such communities. E-Predict addresses an acute need for computational tools that are capable of interpreting the highly complex microarray data obtained through such studies. E-Predict was developed for viral species identification and therefore has immediate implications for medical diagnostics and viral discovery. In addition, the concept of theoretical energy profiles can be extended to represent other microorganisms, particular genes, or biochemical pathways.
Sample preparation and hybridization to microarrays
All patient samples were collected according to protocols approved by the University of California San Francisco Committee on Human Research.
HeLa cells were grown to confluence in a T150 tissue culture flask in Dulbeco's modified Eagle medium supplemented with 10% fetal bovine serum and antibiotics. The cells were harvested by adding 10 ml Trizol reagent (Invitrogen, Carlsbad, CA, USA), and total RNA was isolated according to the manufacturer's protocol. A quantity of HeLa total RNA (50 ng) was used for each amplification and hybridization.
With respect to pediatric respiratory samples, frozen nasopharyngeal aspirate samples were thawed and 200 μl aliquots were used to extract RNA using RNeasy Mini Kit (Qiagen USA, Valencia, CA,USA) as follows. RLT buffer (750 μl) containing 1% 2-mercaptoethanol was added to each sample and mixed. Then, 1 ml of 100% ethanol was added, and the resulting mixture was applied to the columns in three 650 μl aliquots. The remaining steps were carried out in accordance with the manufacturer's protocol, including on-column DNase digest. RNA was eluted from the columns with 30 μl nuclease-free water, and 9 μl was used for amplification and hybridization. For the hepatitis sample, frozen serum sample was thawed and a 150 μl aliquot was used to extract total nucleic acid using MagNA Pure LC Total Nucleic Acid Isolation Kit (Roche Molecular Systems, Alameda, CA, USA), in accordance with the manufacturer's protocol. RNA was eluted in 50 μl nuclease-free water, and 9 μl was used for amplification and hybridization.
For HRV serotypes, frozen samples of low passage viral culture supernatants were thawed on ice and pre-filtered with a 0.2 μm syringe filter. Aliquots (200 μl) of the pre-filtered supernatants were treated with 600 U micrococcal nuclease (Fermentas USA, Hanover, MD, USA) in the presence of 10 mmol/l CaCl2 for 3 hours at 37°C. RNA was then extracted using Trizol reagent (Invitrogen), in accordance with the manufacturer's protocol. Linearized polyacrylamide (20 μg; Ambion, Austin, TX, USA) was used as the carrier during the 2-propanol precipitation. RNA was resuspended in 30 μl nuclease-free water, and 9 μl was used for amplification and hybridization.
Microarrays used in the study were essentially identical to those previously described . Detailed description of the microarray platform, including oligonucleotide sequences, can be found in the NCBI GEO database  (accession GPL 1834). Briefly, 70-mer oligonucleotides representing the most conserved viral genomic elements were selected as 70-mers having sequence similarity (determined by nucleotide alignment) to the highest number of viral genomes . Oligonucleotides were resuspended in 3 × SSC (0.45 M sodium chloride, 0.045 M sodium citrate, pH 7.0) at 50 μmol/l concentration and spotted onto poly-lysine coated glass slides . Each spot on the microarray also contained a unique 'alien' sequence 70-mer (Spike70: 5'-ACC TCG CTA ACC TCT GTA TTG CTT GCC GGA CGC GAG ACA AAC CTG AAC ATT GAG AGT CAC CCT CGT TGT T-3'), spotted at a 1:50 ratio with the viral oligonucleotide to facilitate gridding of the microarrays (see below).
RNA extracted from the samples was amplified using a modified Round A-B random PCR method , as previously described (protocol S1 in ). Briefly, random-primed reverse transcription and second strand synthesis were carried out using primer A (5'-GTT TCC CAG TCA CGA TCN NNN NNN NN-3'). The resulting material was then amplified with 40 cycles of PCR using primer B (5'-GTT TCC CAG TCA CGA TC-3'). This was followed by an additional 20 cycles of PCR with primer B to incorporate aminoallyl-dUTP. The amplified material was then labeled with Cy5, and 0.1-1.0 pmol Probe70 (an oligonucleotide complementary to Spike70 containing five amino-modified bases for dye coupling: 5'-AAC AAC GAG GG[AmC6-dT] GAC TCT CAA [AmC6-dT]GT TCA GGT TTG TC[AmC6-dT] CGC GTC CGG CAA GCA A[AmC6-dT]A CAG AGG T[AmC6-dT]A GCG AGG T-3', Operon Biotechnologies, Huntsville, AL, USA) was labeled with Cy3. The Cy5 and Cy3 probes were pooled and hybridized to the microarray in 3 × SSC at 65°C overnight . The Cy3 channel was used to facilitate gridding but otherwise was ignored in the data analysis. Microarrays were scanned with an Axon 4000B scanner (Axon Instruments, Union City, CA, USA) and gridded using the bundled GenePix 3.0 software.
Fifteen HeLa microarrays were chosen randomly from a set of 43 HeLa hybridizations having at least five papillomavirus oligonucleotides with sum-normalized intensities greater or equal to 0.005. Ten RSV microarrays were chosen randomly from a set of 22 clinical hybridizations having at least five paramyxovirus oligonucleotides with sum-normalized intensities greater than or equal to 0.005 and confirmed to be RSV-positive by DFA. Seven FluA microarrays were chosen from eight available clinical hybridizations having at least five orthomyxovirus oligonucleotides with sum-normalized intensities greater than or equal to 0.005 and confirmed to be FluA-positive by DFA. The eighth FluA microarray was excluded because it was also positive for RSV by visual inspection.
Theoretical energy profiles
The energy profile matrix used in this study included all NCBI reference viral genomes (1,229) available as of July 2004 . Nucleotide BLAST (blastall version 2.2.8  with the default settings) was used to align microarray oligonucleotides with the viral genomes. Energies of hybridization were computed from the alignments using a program distributed with ArrayOligoSelector [30, 43]. In cases in which an oligonucleotide had multiple alignments to the same genome, energy calculations were based on the highest scoring alignment. The energy profile matrix is available as Additional data file 5.
Control oligonucleotides and oligonucleotides known to result in nonspecific hybridization were removed from consideration by setting their intensities and energies to zero. The list of these oligonucleotides (Additional data file 6) was obtained by including 129 oligonucleotides with unnormalized median intensity greater than 500, calculated from 1,009 independent hybridizations described below. The list also included 137 oligonucleotides obtained by clustering of distributions of sum-normalized intensity, based on the same set of 1,009 hybridizations, and visual identification of an outlier cluster with median sum-normalized intensities significantly higher than those observed for most oligonucleotides. Energy vectors were further filtered to exclude terms with energy predictions higher than -30 kcal/mol (again by setting their values to zero), because such predictions on our platform do not correspond to detectable array intensities . A profile was considered only if it had at least three oligonucleotides with non-zero energy predictions. The resulting intensity and energy vectors were normalized using appropriate normalization methods (no normalization, sum, quadratic, and unit-vector). Similarity scores were computed using an appropriate similarity metric (dot product, Pearson correlation, uncentered Pearson correlation, Spearman rank correlation, and similarity based on Euclidean distance).
Null distributions of similarity scores were obtained using a set of 1,009 hybridizations, which included all hybridizations performed on our platform to date. Similarity scores were calculated as described above using uncentered Pearson correlation as the similarity metric, and sum and quadratic normalizations for intensity and energy vectors, respectively. Scores were log-transformed. Right tail outliers corresponding to positive cases were excluded by iterative trimming of the top scores in 1% increments until the best normality fit was obtained, as judged by the Shapiro-Wilk normality test  (implemented in R ). Trimming was allowed to involve 0-25% of all scores. Over one-third of virus profiles required no trimming at all. Only a small number of profiles (34) required trimming beyond 10%, all of which corresponded to viruses frequently present in our samples. No profile required trimming of more than 17% of the scores. The resulting trimmed distributions were assumed to be normal, and their parameters were estimated as the mean and standard deviation of the included scores (Additional data file 7). Obtained parameters were used to estimate significance of individual scores as probabilities associated with observing values equal or greater than the scores. For this purpose, only profiles with at least three oligonucleotides with raw intensity greater than 100 (about two to four times background) were considered.
The first iteration was carried out as described above. For each additional iteration, oligonucleotide intensities of the profile attaining the highest similarity score in the previous iteration were set to zero. The resulting intensity vector was normalized, and similarity scores and P values were calculated using the same normalization method, similarity metric, and null distributions as in the initial iteration.
Clustering of human rhinovirus serotypes
Similarity scores were calculated as described above using uncentered Pearson correlation as the similarity metric, and sum and quadratic normalizations for intensity and energy vectors, respectively. Scores corresponding to picornavirus profiles were clustered using Cluster (version 2.0) [46, 47] by hierarchical average linkage clustering with Pearson correlation as the similarity metric. Cluster images were obtained using Java TreeView (version 1.0.8) [48, 49].
The phylogenetic tree based on nucleotide sequences of VP1 capsid protein was constructed using data from the report by Jonassen and coworkers . Sequence alignment of relevant serotypes and the resulting tree were obtained using ClustalX (version 1.81 for Windows [50, 51]) with default settings.
Polymerase chain reaction
The presence of hepatitis B virus in the hepatitis sample was confirmed using primers Hep_1F (5'-GAC TCG TGG TGG ACT TCT CTC AA-3') and Hep_4R (5'-GAA AGC CCT GCG AAC CAC TGA A-3') with amplified cDNA (Round B material; see  for amplification details) as the template. The presence of RSV in the FluA/RSV double-infected sample was confirmed by PCR using primers AU_041 (5'-GAT GAA AAA TTA AGT GAA ATA TTA GG-3') and AU_042 (5'-GTT CAC GTA TGT TTC CAT ATT TG-3') with cDNA (Round A material; see  for amplification details) as the template. In both cases, amplified PCR fragments were sequenced and had at least 99% nucleotide identity to the genomes of Hepatitis B virus (GenBank: NC_003977) and RSV (GenBank: NC_001803).
The E-Predict software is available for download by any interested party .
The following additional data are available with the online version of this paper: a text file of E-Predict output for the hepatitis example (example 1) (Additional data file 1); a text file of E-Predict output for the FluA/RSV double infection example (example 2) (Additional data file 2); a text file of E-Predict output for the SARS CoV example (example 3) (Additional data file 3); a text file of E-Predict output for the HRV serotypes example (example 4) (Additional data file 4); a tab delimited text file containing the energy profile matrix (Additional data file 5); a text file containing the list of nonspecific oligonucleotides ignored during E-Predict (Additional data file 6); a tab delimited text file containing the list of profile parameters used to estimate P values (Additional data file 7). A text file of E-Predict output used to evaluate normalization and similarity metric parameters (Additional data file 8).
We thank Dr Hao Li, Christina Chaivorapol, and Amir Najmi for helpful discussions. We thank Dr Yu-Tsueng Liu for performing microarray hybridization and PCR follow up of the hepatitis sample. The hepatitis sample was graciously provided as part of an ongoing study by Dr Tim Davern (UCSF). Pediatric respiratory samples were graciously provided as part of an ongoing study by Dr Tara Greenhow, Dr Peggy Weintrub, Dr Lawrence Drew, and Carolyn Wright (UCSF). Cultures of HRV serotypes were graciously provided as part of an ongoing study by Dr David Schnurr and Dr Shigeo Yagi (California Viral and Rickettsial Disease Laboratory, Richmond, CA, USA). This work was supported by a Genentech Graduate Fellowship (A.U.) and grants from the Sandler Program for Asthma Research, and Doris Duke Charitable Foundation (J.D.).
- Riesenfeld CS, Schloss PD, Handelsman J: Metagenomics: genomic analysis of microbial communities. Annu Rev Genet. 2004, 38: 525-552. 10.1146/annurev.genet.38.072902.091216.PubMedView ArticleGoogle Scholar
- Eyers L, George I, Schuler L, Stenuit B, Agathos SN, El Fantroussi S: Environmental genomics: exploring the unmined richness of microbes to degrade xenobiotics. Appl Microbiol Biotechnol. 2004, 66: 123-130. 10.1007/s00253-004-1703-6.PubMedView ArticleGoogle Scholar
- Rodriguez-Valera F: Environmental genomics, the big picture?. FEMS Microbiol Lett. 2004, 231: 153-158. 10.1016/S0378-1097(04)00006-0.PubMedView ArticleGoogle Scholar
- Schloss PD, Handelsman J: Biotechnological prospects from metagenomics. Curr Opin Biotechnol. 2003, 14: 303-310. 10.1016/S0958-1669(03)00067-3.PubMedView ArticleGoogle Scholar
- Venter JC, Remington K, Heidelberg JF, Halpern AL, Rusch D, Eisen JA, Wu D, Paulsen I, Nelson KE, Nelson W, et al: Environmental genome shotgun sequencing of the Sargasso Sea. Science. 2004, 304: 66-74. 10.1126/science.1093857.PubMedView ArticleGoogle Scholar
- Tyson GW, Chapman J, Hugenholtz P, Allen EE, Ram RJ, Richardson PM, Solovyev VV, Rubin EM, Rokhsar DS, Banfield JF: Community structure and metabolism through reconstruction of microbial genomes from the environment. Nature. 2004, 428: 37-43. 10.1038/nature02340.PubMedView ArticleGoogle Scholar
- Breitbart M, Hewson I, Felts B, Mahaffy JM, Nulton J, Salamon P, Rohwer F: Metagenomic analyses of an uncultured viral community from human feces. J Bacteriol. 2003, 185: 6220-6223. 10.1128/JB.185.20.6220-6223.2003.PubMedPubMed CentralView ArticleGoogle Scholar
- Breitbart M, Felts B, Kelley S, Mahaffy JM, Nulton J, Salamon P, Rohwer F: Diversity and population structure of a near-shore marine-sediment viral community. Proc Biol Sci. 2004, 271: 565-574. 10.1098/rspb.2003.2628.PubMedPubMed CentralView ArticleGoogle Scholar
- Acinas SG, Klepac-Ceraj V, Hunt DE, Pharino C, Ceraj I, Distel DL, Polz MF: Fine-scale phylogenetic architecture of a complex bacterial community. Nature. 2004, 430: 551-554. 10.1038/nature02649.PubMedView ArticleGoogle Scholar
- van der Wielen PW, Bolhuis H, Borin S, Daffonchio D, Corselli C, Giuliano L, D'Auria G, de Lange GJ, Huebner A, Varnavas SP, et al: The enigma of prokaryotic life in deep hypersaline anoxic basins. Science. 2005, 307: 121-123. 10.1126/science.1103569.PubMedView ArticleGoogle Scholar
- Liles MR, Manske BF, Bintrim SB, Handelsman J, Goodman RM: A census of rRNA genes and linked genomic sequences within a soil metagenomic library. Appl Environ Microbiol. 2003, 69: 2684-2691. 10.1128/AEM.69.5.2684-2691.2003.PubMedPubMed CentralView ArticleGoogle Scholar
- Woese CR: Bacterial evolution. Microbiol Rev. 1987, 51: 221-271.PubMedPubMed CentralGoogle Scholar
- Brady SF, Chao CJ, Clardy J: New natural product families from an environmental DNA (eDNA) gene cluster. J Am Chem Soc. 2002, 124: 9968-9969. 10.1021/ja0268985.PubMedView ArticleGoogle Scholar
- Rondon MR, August PR, Bettermann AD, Brady SF, Grossman TH, Liles MR, Loiacono KA, Lynch BA, MacNeil IA, Minor C, et al: Cloning the soil metagenome: a strategy for accessing the genetic and functional diversity of uncultured microorganisms. Appl Environ Microbiol. 2000, 66: 2541-2547. 10.1128/AEM.66.6.2541-2547.2000.PubMedPubMed CentralView ArticleGoogle Scholar
- Henne A, Schmitz RA, Bomeke M, Gottschalk G, Daniel R: Screening of environmental DNA libraries for the presence of genes conferring lipolytic activity on Escherichia coli. Appl Environ Microbiol. 2000, 66: 3113-3116. 10.1128/AEM.66.7.3113-3116.2000.PubMedPubMed CentralView ArticleGoogle Scholar
- Bodrossy L, Sessitsch A: Oligonucleotide microarrays in microbial diagnostics. Curr Opin Microbiol. 2004, 7: 245-254. 10.1016/j.mib.2004.04.005.PubMedView ArticleGoogle Scholar
- Zhou J: Microarrays for bacterial detection and microbial community analysis. Curr Opin Microbiol. 2003, 6: 288-294. 10.1016/S1369-5274(03)00052-3.PubMedView ArticleGoogle Scholar
- Cook KL, Sayler GS: Environmental application of array technology: promise, problems and practicalities. Curr Opin Biotechnol. 2003, 14: 311-318. 10.1016/S0958-1669(03)00057-0.PubMedView ArticleGoogle Scholar
- Wang D, Coscoy L, Zylberberg M, Avila PC, Boushey HA, Ganem D, DeRisi JL: Microarray-based detection and genotyping of viral pathogens. Proc Natl Acad Sci USA. 2002, 99: 15687-15692. 10.1073/pnas.242579699.PubMedPubMed CentralView ArticleGoogle Scholar
- Sengupta S, Onodera K, Lai A, Melcher U: Molecular detection and identification of influenza viruses by oligonucleotide microarray hybridization. J Clin Microbiol. 2003, 41: 4542-4550. 10.1128/JCM.41.10.4542-4550.2003.PubMedPubMed CentralView ArticleGoogle Scholar
- Klaassen CH, Prinsen CF, de Valk HA, Horrevorts AM, Jeunink MA, Thunnissen FB: DNA microarray format for detection and subtyping of human papillomavirus. J Clin Microbiol. 2004, 42: 2152-2160. 10.1128/JCM.42.5.2152-2160.2004.PubMedPubMed CentralView ArticleGoogle Scholar
- Lin B, Vora GJ, Thach D, Walter E, Metzgar D, Tibbetts C, Stenger DA: Use of oligonucleotide microarrays for rapid detection and serotyping of acute respiratory disease-associated adenoviruses. J Clin Microbiol. 2004, 42: 3232-3239. 10.1128/JCM.42.7.3232-3239.2004.PubMedPubMed CentralView ArticleGoogle Scholar
- Wang D, Urisman A, Liu YT, Springer M, Ksiazek TG, Erdman DD, Mardis ER, Hickenbotham M, Magrini V, Eldred J, et al: Viral discovery and sequence recovery using DNA microarrays. PLoS Biol. 2003, 1: E2-10.1371/journal.pbio.0000002.PubMedPubMed CentralView ArticleGoogle Scholar
- Lemarchand K, Masson L, Brousseau R: Molecular biology and DNA microarray technology for microbial quality monitoring of water. Crit Rev Microbiol. 2004, 30: 145-172. 10.1080/10408410490435142.PubMedView ArticleGoogle Scholar
- Rhee SK, Liu X, Wu L, Chong SC, Wan X, Zhou J: Detection of genes involved in biodegradation and biotransformation in microbial communities by using 50-mer oligonucleotide microarrays. Appl Environ Microbiol. 2004, 70: 4303-4317. 10.1128/AEM.70.7.4303-4317.2004.PubMedPubMed CentralView ArticleGoogle Scholar
- Ivnitski D, O'Neil DJ, Gattuso A, Schlicht R, Calidonna M, Fisher R: Nucleic acid approaches for detection and identification of biological warfare and infectious disease agents. Biotechniques. 2003, 35: 862-869.PubMedGoogle Scholar
- NCBI Gene Expression Omnibus. [http://www.ncbi.nlm.nih.gov/geo/]
- Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ: Basic local alignment search tool. J Mol Biol. 1990, 215: 403-410. 10.1006/jmbi.1990.9999.PubMedView ArticleGoogle Scholar
- Peyret N, Seneviratne PA, Allawi HT, SantaLucia J: Nearest-neighbor thermodynamics and NMR of DNA sequences with internal A.A, C.C, G.G, and T.T mismatches. Biochemistry. 1999, 38: 3468-3477. 10.1021/bi9825091.PubMedView ArticleGoogle Scholar
- Bozdech Z, Zhu J, Joachimiak MP, Cohen FE, Pulliam B, DeRisi JL: Expression profiling of the schizont and trophozoite stages of Plasmodium falciparum with a long-oligonucleotide microarray. Genome Biol. 2003, 4: R9-10.1186/gb-2003-4-2-r9.PubMedPubMed CentralView ArticleGoogle Scholar
- Ksiazek TG, Erdman D, Goldsmith CS, Zaki SR, Peret T, Emery S, Tong S, Urbani C, Comer JA, Lim W, et al: A novel coronavirus associated with severe acute respiratory syndrome. N Engl J Med. 2003, 348: 1953-1966. 10.1056/NEJMoa030781.PubMedView ArticleGoogle Scholar
- Rota PA, Oberste MS, Monroe SS, Nix WA, Campagnoli R, Icenogle JP, Penaranda S, Bankamp B, Maher K, Chen MH, et al: Characterization of a novel coronavirus associated with severe acute respiratory syndrome. Science. 2003, 300: 1394-1399. 10.1126/science.1085952.PubMedView ArticleGoogle Scholar
- Jonassen CM, Jonassen TO, Grinde B: A common RNA motif in the 3' end of the genomes of astroviruses, avian infectious bronchitis virus and an equine rhinovirus. J Gen Virol. 1998, 79: 715-718.PubMedView ArticleGoogle Scholar
- Ledford RM, Patel NR, Demenczuk TM, Watanyar A, Herbertz T, Collett MS, Pevear DC: VP1 sequencing of all human rhinovirus serotypes: insights into genus phylogeny and susceptibility to antiviral capsid-binding compounds. J Virol. 2004, 78: 3663-3674. 10.1128/JVI.78.7.3663-3674.2004.PubMedPubMed CentralView ArticleGoogle Scholar
- Blomqvist S, Savolainen C, Raman L, Roivainen M, Hovi T: Human rhinovirus 87 and enterovirus 68 represent a unique serotype with rhinovirus and enterovirus features. J Clin Microbiol. 2002, 40: 4218-4223. 10.1128/JCM.40.11.4218-4223.2002.PubMedPubMed CentralView ArticleGoogle Scholar
- Savolainen C, Blomqvist S, Mulders MN, Hovi T: Genetic clustering of all 102 human rhinovirus prototype strains: serotype 87 is close to human enterovirus 70. J Gen Virol. 2002, 83: 333-340.PubMedView ArticleGoogle Scholar
- Loy A, Lehner A, Lee N, Adamczyk J, Meier H, Ernst J, Schleifer KH, Wagner M: Oligonucleotide microarray for 16S rRNA gene-based detection of all recognized lineages of sulfate-reducing prokaryotes in the environment. Appl Environ Microbiol. 2002, 68: 5064-5081. 10.1128/AEM.68.10.5064-5081.2002.PubMedPubMed CentralView ArticleGoogle Scholar
- Cho JC, Tiedje JM: Bacterial species determination from DNA-DNA hybridization by using genome fragments and DNA microarrays. Appl Environ Microbiol. 2001, 67: 3677-3682. 10.1128/AEM.67.8.3677-3682.2001.PubMedPubMed CentralView ArticleGoogle Scholar
- Eisen MB, Brown PO: DNA arrays for analysis of gene expression. Methods Enzymol. 1999, 303: 179-205.PubMedView ArticleGoogle Scholar
- Bohlander SK, Espinosa R, Le Beau MM, Rowley JD, Diaz MO: A method for the rapid sequence-independent amplification of microdissected chromosomal material. Genomics. 1992, 13: 1322-1324. 10.1016/0888-7543(92)90057-Y.PubMedView ArticleGoogle Scholar
- NCBI Reference Viral Genomes. [http://www.ncbi.nlm.nih.gov/genomes/VIRUSES/viruses.html]
- NCBI BLAST. [http://www.ncbi.nlm.nih.gov/BLAST/]
- ArrayOligoSelector. [http://arrayoligosel.sourceforge.net/]
- Shapiro SS, Wilk MB: An analysis of variance test for normality (complete samples). Biometrika. 1965, 52: 591-611.View ArticleGoogle Scholar
- The R Project. [http://www.r-project.org/]
- Eisen MB, Spellman PT, Brown PO, Botstein D: Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA. 1998, 95: 14863-14868. 10.1073/pnas.95.25.14863.PubMedPubMed CentralView ArticleGoogle Scholar
- Eisen Lab, Software. [http://rana.lbl.gov/EisenSoftware.htm]
- Saldanha AJ: Java Treeview: extensible visualization of microarray data. Bioinformatics. 2004, 20: 3246-3248. 10.1093/bioinformatics/bth349.PubMedView ArticleGoogle Scholar
- Java TreeView. [http://jtreeview.sourceforge.net/]
- Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, Higgins DG: The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 1997, 25: 4876-4882. 10.1093/nar/25.24.4876.PubMedPubMed CentralView ArticleGoogle Scholar
- ClustalX. [ftp://ftp-igbmc.u-strasbg.fr/pub/ClustalX/]
- DeRisi Lab: E-Predict Download. [http://derisilab.ucsf.edu/epredict/]
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.