PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of cviPermissionsJournals.ASM.orgJournalAEM ArticleJournal InfoAuthorsReviewers
 
Clin Vaccine Immunol. Dec 2011; 18(12): 2148–2153.
PMCID: PMC3232686
Plasma Antibody Profiles as Diagnostic Biomarkers for Tuberculosis[down-pointing small open triangle]
Imran H. Khan,1,2* Resmi Ravindran,1 Viswanathan V. Krishnan,1,2,3 Irum Nawaz Awan,4 Syed Kumail Rizvi,4 Muhammad Arif Saqib,4 Mirza Imran Shahzad,4 Sabira Tahseen,5 Greg Ireton,6 Celia W. Goulding,7 Phil Felgner,8 Kathy DeRiemer,9 Azra Khanum,4 and Paul A. Luciw1,2
1Center for Comparative Medicine
2Department of Pathology and Laboratory Medicine
9Department of Public Health Sciences, University of California, Davis, Davis, California
3Department of Chemistry, California State University, Fresno, California 93740
4Department of Biochemistry, Pir Mehr Ali Shah Arid Agriculture University Rawalpindi, Rawalpindi, Pakistan
5National TB Control Program, Islamabad, Pakistan
6Infectious Disease Research Institute, Seattle, Washington
7Departments of Molecular Biology & Biochemistry and Pharmaceutical Sciences
8Department of Medicine, University of California, Irvine, Irvine, California
*Corresponding author. Mailing address: Center for Comparative Medicine, University of California at Davis, Hutchison Rd. and County Rd. 98, Davis, CA 95616. Phone: (530) 752-7200. Fax: (530) 752-7914. E-mail: ihkhan/at/ucdavis.edu.
These authors contributed equally to this work.
Received July 20, 2011; Revisions requested August 12, 2011; Accepted September 24, 2011.
Two billion people are infected with Mycobacterium tuberculosis, the etiological agent of tuberculosis (TB), worldwide. Ten million to 20 million of the infected individuals develop disease per year. TB is a treatable disease, provided that it is diagnosed in a timely manner. The current TB diagnostic methods are subjective, inefficient, or not cost-effective. Antibody-based blood tests can be used efficiently and cost-effectively for TB diagnosis. A major challenge is that different TB patients generate antibodies against different antigens. Therefore, a multiplex immunoassay approach is needed. We have developed a multiplex panel of 28 M. tuberculosis antigen-coated microbeads. Plasma samples were obtained from over 300 pulmonary TB patients and healthy controls in a country where TB is endemic, Pakistan. Multiplex data were analyzed using computational tools by multivariate statistics, classification algorithms, and cluster analysis. The results of antibody profile-based detection, using 16 selected antigens, closely correlated with those of the sputum-based diagnostic methods (smear microscopy and culture) practiced in countries where TB is endemic. Multiplex microbead immunoassay had a sensitivity and specificity of approximately 90% and 80%, respectively. These antibody profiles could potentially be useful for the diagnosis of nonpulmonary TB, which accounts for approximately 20% of cases of disease. Since an automated, high-throughput version of this multiplex microbead immunoassay could analyze thousands of samples per day, it may be useful for the diagnosis of TB in millions of patients worldwide.
More than one-third of the world's population is infected with Mycobacterium tuberculosis (7, 26a). Annually, 10 million to 20 million of these individuals develop clinical symptoms, and about 2 million die of tuberculosis (TB) (4, 17a). The infected host typically mounts a vigorous immune response (25). Nevertheless, 10% of all infections result in active disease within 2 years. Another 10% of cases may experience disease after a latent phase spanning many years (8, 17a). Several Mycobacterium species (e.g., M. tuberculosis, M. bovis, and M. africanum) can infect and cause disease in humans (2, 24). In about 80% of active TB cases, direct involvement of the lung results in pulmonary disease (4a). However, M. tuberculosis can spread to other organs. In approximately 20% of cases, M. tuberculosis may cause nonpulmonary disease in various organ systems (urogenital system, nervous system, digestive system, skeletal system, etc.) with or without the lung involvement (7, 18). TB is a treatable disease, provided that a timely and appropriate diagnosis is made (4a). Commonly used sputum-based methods for pulmonary TB diagnosis are subjective, insensitive, and/or inefficient. Furthermore, for the detection of pediatric pulmonary TB, a major limitation is that children often have difficulty producing usable quantities of sputum.
Sputum smear acid-fast bacillus (AFB) microscopy is recommended by the World Health Organization (WHO) as the first-line diagnostic procedure for pulmonary disease. Although relatively specific, this method is subjective, inconsistent, and not very sensitive (globally, 30 to 70% sensitivity) (26a). Bacterial culture is considered a “gold standard” for TB diagnosis, but because M. tuberculosis is a slow-growing organism, the standard culture methods can take up to 8 to 12 weeks to obtain results (9).
The complete genome sequences of M. tuberculosis (H37Rv, virulent laboratory strain) have been determined (3). More recently, specific and sensitive TB diagnostic tests have been developed by taking advantage of advances in sequencing and annotation of the M. tuberculosis genome, which has revealed approximately 4,000 open reading frames (http://genolist.pasteur.fr/TubercuList/). These diagnostic tests include nucleic acid amplification of M. tuberculosis but are limited to use with processed sputum samples. Disease diagnostics based on blood tests are advantageous because they are minimally invasive, rapid, and cost-effective and are useful for nonpulmonary and pediatric TB. Detection of anti-M. tuberculosis antibodies (plasma or serum) is more suitable for implementation in a variety of clinical laboratory settings. Despite efforts to develop TB diagnostics based on serology, there are challenges facing this approach. Not all patients produce antibodies against the same M. tuberculosis antigens, and exposure to environmental mycobacteria and M. bovis BCG (bacillus Calmette-Guérin) vaccination can potentially lead to confounding results. We reasoned that these challenges can be overcome by a user-friendly and cost-effective multiplex method that employs dozens of M. tuberculosis antigens for detecting profiles of anti-M. tuberculosis antibodies. Detection of antibodies against multiple M. tuberculosis antigens has been fruitful in the detection of M. tuberculosis infection (16).
Ideally, a multiplex platform selected for a clinical diagnostic test should be suitable for the entire process from assay development to clinical validation and implementation. It should additionally be amenable to high throughput, robust, and flexible; readily deployable in low-resource settings; require minimal training; and be cost-effective. A multiplex microbead immunoassay based on the xMAP technology platform (Luminex Corp, Austin, TX) satisfies all of the above-described requirements for a useful infectious disease diagnostic. Discovery platforms such as 2-dimensional protein array (21) are useful in the initial selection of target proteins (antigens) but are inflexible, require sophisticated laboratory infrastructure, and are not cost-effective.
In our study, antibody profiles generated by multiplex microbead immunoassay and multivariate and cluster analyses enabled differentiation of TB patients from healthy indigenous individuals. The xMAP platform used in this study has a high capacity for analysis of hundreds to thousands of samples from patients and control groups per day, making it applicable for use as a first-line diagnostic in countries where TB is endemic.
M. tuberculosis antigens.
Recombinant antigens from 28 M. tuberculosis genes were expressed in Escherichia coli. Plasmid vectors for expression of eight recombinant antigens were obtained from the TB Resource Center at Colorado State University (Fort Collins, CO): Rv3875 (ESAT6), Rv3874 (CFP10), Rv2031c (HspX), Rv3804c (antigen 85a [Ag85a]), Rv1886c (Ag85b), Rv0129c (Ag85c), Rv3841 (Bfrb1), and Rv3418c (GroES). The recombinant antigens were expressed and purified at the University of California, Davis (UCD; Davis, CA), and the University of Arid Agriculture (UAAR), Rawalpindi, Pakistan, as described elsewhere (http://www.cvmbs.colostate.edu/mip/tb/sop.htm). Several other recombinant antigens (n = 14) were expressed and purified at the Infectious Disease Research Institute (IDRI; Seattle, WA) as previously described (13): Rv2875 (MPT70), Rv1984c (CFP21), Rv1980c (MPT64), Rv0934 (P38 or PstS1), Rv1860 (MPT32), Rv0054, Rv3874-Rv3875 (CFP10-ESAT) fusion, Rv3873, Rv3619, Rv2220, Rv0831c, Rv1009, Rv1099, and Rv2032. The antigens Rv1926c, Rv2878c, Rv1677, Rv3881c, Rv1566c, and Rv3507 were expressed and purified at the University of California, Irvine (Irvine, CA), as previously described (10, 11).
Sample groups.
Blood samples from TB patients and healthy individuals were obtained under the protocols approved through the institutional review boards (IRBs) at UCD and UAAR. The HIV status of healthy controls and TB patients was unknown. Samples were obtained from the four main groups described below.
(i) Healthy control samples.
The first control group (n = 87) comprised individuals of both sexes (ages, 21 to 25 years) from the same geographical area as the TB patients (Rawalpindi/Islamabad, Pakistan, where TB is endemic). This group was termed the “indigenous healthy (IH) group.” BCG coverage in Pakistan since 1978 has been over 80% (19a). The second group of healthy donors of both sexes (ages, 21 to 65 years) comprised individuals from the California Central Valley (Delta Blood Bank, Stockton, CA) and was termed the “nonindigenous healthy (NH) group” (n = 146); the BCG status of this group is unknown.
(ii) TB patient samples.
Plasma samples were collected from AFB microscopy-positive (AFB+) TB patients (n = 243) at the Federal Government TB Hospital, Rawalpindi, Pakistan. At this hospital, TB diagnosis is primarily based on AFB microscopy (on the basis of WHO-recommended criteria). Three consecutive, daily sputum specimens were used to determine positivity by AFB microscopy. A subgroup of AFB+ patients consisted of those undergoing directly observed treatment, short course (DOTS), termed the “DOTS group.” Blood samples were drawn at four time points: (i) before DOTS (including all the AFB+ patients, n = 243) and (ii) 2 months (n = 52), (iii) 4 months (n = 31), and (iv) 6 months (n = 21) post-DOTS initiation. Patients from whom post-DOTS samples were taken were subsets of the original AFB+ patients (n = 243). Because visits to the Federal Government TB Hospital, Rawalpindi, by the original AFB+ patients decreased over time due to poor compliance, there was a progressive decrease in the number of samples available at the later time points. Also included in the study are a group of TB patients that were negative by AFB microscopy but positive by culture, designated the “AFB/Cul+ group” (n = 17).
Blood sample collection, processing, and storage.
A 5-ml blood sample was collected from each individual through venipuncture and placed into a Vaccutainer tube (EDTA; catalog no. 367899; BD, NJ). Plasma was separated by centrifugation at 1,000 × g, 10 min, room temperature) within 2 h of collection and immediately frozen in aliquots at −80°C until use.
Microbead coating with M. tuberculosis antigens.
Microbead sets (Luminex Corp, Austin, TX) were coupled through carbodiimide linkages with each of the M. tuberculosis recombinant antigens as previously described (1416).
Multiplex assay conditions and data collection.
Multiplex assays were performed and data (median fluorescence intensity [MFI]) were collected as previously described (16). Cutoff values were calculated for each antigen-coated microbead set; MFI values that represented the reactivity of each microbead set to healthy control samples from the IH group were used for calculations, as follows: mean MFI + (2 × standard deviation).
Statistical analysis.
Multivariate analysis was performed to select useful antigens. Fold increases of antibodies in patient samples were calculated across conditions by fitting a linear model to the data (antibodies to 28 M. tuberculosis antigens with technical duplicates) for each patient relative to the healthy groups. A significantly positive signal was detected in at least one comparison between control and patient groups (e.g., IH versus AFB+ groups). Differential expression across the multiple comparisons was detected by the F test; separate F tests were performed for each antibody; and P values were adjusted using the Benjamini-Hochberg procedure to correct for multiple comparisons (1, 5, 8).
Cluster analysis was performed to define natural groupings of antigens and TB patients (AFB+ group) analyzed with respect to healthy controls (IH group) as previously described (20). Briefly, data sets with which to perform the cluster analysis were generated by scaling the MFI levels of each sample in the AFB+ group with reference to the baseline value for each microbead set. Hierarchical cluster analysis was performed using a Euclidean distance metric, without standardization. For visual depiction, results of the cluster analysis were presented as dendrograms, and heat maps were generated by a combination of codes using the Matlab (19) and SigmaPlot (23) programs.
To calculate assay predictive values, sensitivity, and specificity, multiplex data were first analyzed by two regression model-based classification algorithms (random forest [RF] and sequential minimal optimization [SMO]). These analyses were performed on data from IH controls and AFB+ patients with 20-fold cross validation employing Waikato Environment for Knowledge Analysis (WEKA; version 3.6.2) software (6, 12). Analysis by classification algorithms yielded the following measures: true positive (TP), true negative (TN), false positive (FP), and false negative (FN). Calculations were performed using the following formulas: sensitivity = TP/(TP + FN), specificity = TN/(FP + TN), positive predictive value (PPV) = TP/(TP + FP), and negative predictive value (NPV) = TN/(FN + TN). Because the total number of samples in the AFB/Cul+ group was relatively low (n = 17) and, therefore, analysis by classification algorithms was unsuitable, sensitivity calculations for this group were performed on TP and FN values determined from the cutoff points for tier 1 antigens (Fig. 1).
Fig. 1.
Fig. 1.
Capacity of M. tuberculosis antigen-coated microbead sets to discriminate between TB patients (AFB+ group) and healthy controls (IH group). Black filled circles and red filled squares, average reactivity (MFI) of each antigen-coated microbead set to IH (more ...)
Detection of antibodies in TB patients.
We examined the serological response to 28 recombinant M. tuberculosis antigens in patients with TB (n = 243) diagnosed by AFB microscopy (AFB+ group) and IH controls (n = 87). Using a multiplex microbead immunoassay and multivariate analysis, antibodies against 16 M. tuberculosis antigens were detected in the AFB+ group (Table 1). Results for the NH controls (n = 146) were similar (Table 1). The other 12 antigens either reacted nonspecifically (Rv1009, Rv1099, Rv3507, Rv1677) or displayed negligible reactivity (Rv3874-Rv3875 fusion, Rv3619, Rv2220, Rv2032, Rv1984, Rv3873, Rv1566c, Rv3418c) (data not shown).
Table 1.
Table 1.
Fold increases in patient plasma antibodiesa
Important antigens.
The 16 useful antigens were further divided, on the basis of the antibody detection signal (MFI fold change [FC]), into two categories, tier 1 and tier 2 (Table 1). The extent of discrimination between antibody-positive and -negative samples was much larger for tier 1 antigens than for those in tier 2 (Fig. 1). Differences in antibody responses in the two groups of TB patients (AFB+ and AFB/Cul+) were also noted (Fig. 2; Table 1). For example, antigen Rv0054 (tier 1) was of high value for the AFB+ group of patients (5.66 FC and 42% positive) but was not significant in the AFB/Cul+ group (1.32 FC and 12% positive).
Fig. 2.
Fig. 2.
Percentage of patients positive for antibodies to tier 1 antigens relative to the baseline values in Fig. 1. Black bars and red bars, AFB+ and AFB/Cul+ patient groups, respectively.
Antibodies in TB patients post-DOTS group.
Over a 6-month period post-DOTS, antibody profiles did not change for the post-DOTS group as a whole or in individual patients (data not shown). The impact of DOTS on the outcome of disease in these patients is not known; it is assumed to be an effective treatment.
Antibody profiles.
The serology information based on cutoff values for each antigen-coated microbead set (Fig. 1) was incorporated in the cluster analysis first to identify antibody-positive patient samples in the AFB+ group and then to identify reactivity with specific antigens and classify antigens and patients into natural groupings by computer-assisted clustering.
(i) Antigen clusters.
The relative strength of reactivity of each antigen in individual patient samples is shown in a heat map by color intensity (red) above the cutoff values (black) (Fig. 3). Profiles of antibodies against the same 16 antigens identified by the multivariate analysis emerged as two major clusters (clusters 1 and 2; Fig. 3); clusters 1 and 2 were the same as tiers 1 and 2 (Table 1), respectively.
Fig. 3.
Fig. 3.
Natural groupings (clusters) of antigens and patients (AFB+ group). M. tuberculosis antigens are listed at the top of the heat map, and two major clusters of antigens are shown. Patient clusters are indicated by color-coded dendrograms on the left side (more ...)
(ii) Patient clusters.
TB patients with similar antibody profiles classified into seven major clusters (Fig. 3). Antibodies against Rv3881 can be visualized in the largest number of patients, followed by antibodies against the rest of the antigens, as in Fig. 2. In most of the patient clusters (e.g., clusters 1 to 6), each cluster was dominated by antibodies against at least one tier 1 antigen (Fig. 3). Because tier 1 antigens demonstrated strong discrimination between TB patients and healthy controls (Fig. 1), these patient clusters were well-defined. Patient cluster 7 was an exception; antibodies against no single antigen dominated, and most of the patients contained antibodies against cluster 2 antigens.
Predictive modeling.
Two classification algorithms (i.e., RF and SMO) were used to classify antibody-positive and -negative samples among the patient and healthy control samples. Data from these classification analyses were used for determining predictive values, as well as assay sensitivity and specificity.
(i) Positive and negative predictive values.
The ability of the multiplex microbead immunoassay to predict whether individuals were infected with M. tuberculosis (PPV) or healthy (NPV) was assessed. PPVs for tier 1 antigens alone and tier 1 plus tier 2 antigens together were 91 to 95% and 93%, respectively. NPVs were 77 to 78% and 78 to 80%, respectively (Table 2).
Table 2.
Table 2.
Positive and negative predictive values, sensitivity, and specificitya
(ii) Assay sensitivity and specificity.
The sensitivity of the multiplex microbead immunoassay (tier 1 antigens) for the detection of M. tuberculosis infection in the AFB+ group was 91 to 92%. Similar levels of sensitivity were observed with tier 1 and tier 2 antigens together (92 to 93%) (Table 2). Sensitivity for the detection of M. tuberculosis infection in the AFB/Cul+ group of patients was 88% (tier 1 antigens). The specificity of the multiplex immunoassay assay ranged from 75 to 85% for tier 1 antigens and was 79% for tier 1 and tier 2 antigens together (Table 2).
Sample sets and data analysis.
Samples from healthy individuals from the same geographical region (IH group) as the TB patients were used as controls. A second control group (NH group) showed that the differences in background levels in two geographically distinct healthy populations were minimal. Data were analyzed with three statistical approaches, each highlighting a specific aspect of the diagnostic power of the multiplex microbead immunoassay coupled with appropriate computational tools. First, multivariate statistical methods were used to compare the antibody levels between the study groups. Linear modeling with the empirical Bayesian method (8) enabled identification of antibodies with high fold changes and significant P values (Table 1). Second, a clustering scheme was used to categorize the profiles of antibodies (and their levels) with natural groupings of antigens and patients (Fig. 3). Third, two distinctly different but high-performing algorithms (RF, SMO) were used for classification of antibody-positive and -negative samples (Table 2). An approach based on the receiver optimization characteristic (ROC) curve is useful for binary distinction of a single antibody. In contrast, the regression-based classification algorithms used in this study provided a powerful summary of prediction and performance in the presence of multiple positive signals. In the development of a computational data-mining system, the above-described algorithms provided the most accurate results with data from several hundred samples (unpublished data) analyzed by multiplex serodetection (10 infectious pathogens) in a validated mouse model of infectious diseases (20).
Anti-M. tuberculosis antibody profiles.
Two independent statistical approaches used for the selection of antigens useful for serodetection of M. tuberculosis infection in AFB+ patients (multivariate and cluster analysis) identified the same sets of antigens. In fact, antigens in tiers 1 and 2 were identical to those in clusters 1 and 2, respectively. In addition, the ability of antigen-coated microbead sets to discriminate between antibody-positive and -negative samples (Fig. 1) is consistent with the value of each antigen prioritized on the basis of multivariate and cluster analyses (Table 1 and Fig. 3). We did not observe significant changes in the antibody profiles post-DOTS. This observation suggests that once generated, antibodies are sustained over many months.
Antigens used in this study were chosen on the basis of previous studies that have shown the utility of M. tuberculosis antigens in serodiagnosis of TB in various immunoassay formats either individually (e.g., enzyme-linked immunosorbent assay) or in combination (e.g., two-dimensional chip array) (13, 26, 27). More recently, a comprehensive multiplex study based on two-dimensional chip array analysis of gene expression products of over 4,000 open reading frames in the M. tuberculosis genome was reported (17). Results showed that only a subset of M. tuberculosis antigens (about a dozen) may be useful in serodetection of TB. The selected antigens were as follows: Rv3881, Rv3804, Rv1860, Rv2031c, Rv0934, Rv1980c, Rv1411, Rv3616c, Rv3864, Rv0632, Rv2873, Rv3874, and Rv1984c. The first five of these selected antigens (Rv3881, Rv3804, Rv1860, Rv2031c, Rv0934) are common with those in tier 1 reported in this study, and the sixth antigen (Rv1980c) is in tier 2. There were a few differences in the results of the two multiplex studies. For example, we identified three additional antigens (Rv0054, Rv1886c, and Rv0129c) to be highly discriminatory between TB patients and healthy individuals (Fig. 1). Additionally, antibodies against Rv3874 in our study were not highly discriminatory between TB patients and healthy individuals; therefore, it was included in tier 2 (Fig. 1). Antibodies against Rv1984c were not at all detectable in our study. These differences in results may be due to the different technology platforms used in the two studies. Five other selected antigens (Rv1411, Rv3616c, Rv3864, Rv0632, and Rv2873) listed in the previous study (17) were not included in our study.
Sensitivity and specificity.
Data analyses based on two classification algorithms show that the assay sensitivity for TB patients in the AFB+ group was approximately 90%, while the assay specificity was approximately 80%. Due to a smaller number of patients (n = 17), assay sensitivity for the AFB/Cul+ group was calculated on the basis of cutoff values for each microbead set, and it was similar to that for the AFB+ group (88%). Although inclusion of tier 2 antigens (e.g., Rv3874 and Rv3875) somewhat reduced assay specificity and a relatively small number of patients contained antibodies against them (Fig. 3), Rv3874 and Rv3875 are useful antigens since they are specific to M. tuberculosis/M. bovis and are absent in environmental mycobacteria and BCG. Certain other antigens (e.g., Rv3804c, Rv1886c, Rv0129c, and Rv1860) are also found in environmental mycobacteria. Therefore, detection of antibodies against these antigens may contribute to a decrease in assay specificity.
Overall, the following conclusions can be drawn: (i) the sensitivity and specificity of multiplex microbead immunoassay strongly suggest that this approach will be useful for TB serodiagnostics in regions of the world where TB is endemic, (ii) antigens selected in this study are sensitive for the detection of M. tuberculosis antibodies in TB patients, and (iii) the inclusion of multiple antigens is necessary to achieve specificity and sensitivity comparable to those of sputum methods of smear microscopy and culture.
Results presented in this report clearly demonstrate that the selected antigens and the xMAP technology platform are highly suitable for TB serodiagnostics in pulmonary TB patients. Most importantly, although nonpulmonary TB patients were not included in this study, it is reasonable to propose that this multiplex microbead immunoassay will also be useful for the serodiagnosis of this manifestation of M. tuberculosis infection. Approximately 20% of TB cases do not have lung involvement (or have minimal lung involvement) (22). These cases may go undetected by conventional diagnostic methods for long periods of time. A research grant has been awarded by the Regional Office for the Eastern Mediterranean (EMRO) of WHO to perform a field validation study of the multiplex serodiagnostic system in Pakistan. We will assess application of this novel assay for the detection of several forms of TB, including pediatric and nonpulmonary cases, by analysis of a total of 1,200 TB patients and matched controls.
ACKNOWLEDGMENTS
We thank Marila Gennaro for providing M. tuberculosis antigen Rv3875 for initial experiments, a critical reading of the manuscript, and making helpful suggestions.
This work was supported by funding (to P.A.L., I.H.K., and A.K.), in whole or part, from the National Academy of Sciences, the U.S. Agency for International Development, the U.S. Department of State, and the Higher Education Commission of Pakistan. V. V. K. was supported in part by a grant for Research Infrastructure for Minority Institutions (P20MD002732).
Any opinions, findings, conclusions, or recommendations expressed are those of the authors and do not necessarily reflect the views of the project sponsors.
Footnotes
[down-pointing small open triangle]Published ahead of print on 5 October 2011.
1. Bolstad B. M., Irizarry R. A., Astrand M., Speed T. P. 2003. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19:185–193. [PubMed]
2. Brosch R., et al. 2002. A new evolutionary scenario for the Mycobacterium tuberculosis complex. Proc. Natl. Acad. Sci. U. S. A. 99:3684–3689. [PubMed]
3. Cole S. T., et al. 1998. Deciphering the biology of Mycobacterium tuberculosis from the complete genome sequence. Nature 393:537–544. [PubMed]
4. Corbett E. L., et al. 2003. The growing burden of tuberculosis: global trends and interactions with the HIV epidemic. Arch. Intern. Med. 163:1009–1021. [PubMed]
4a. Davies D. H., et al. 2005. Profiling the humoral immune response to infection by using proteome microarrays: high-throughput vaccine and diagnostic antigen discovery. Proc. Natl. Acad. Sci. U. S. A. 102:547–552. [PubMed]
5. Dudoit S., Popper-Shaffer J., Boldrick J. C. 2002. Multiple hypothesis testing in microarray experiments, U.C. Berkeley Division of Biostatistics Working Paper Series. University of California, Berkeley, Berkeley, CA.
6. Frank E., Hall M., Trigg L., Holmes G., Witten I. H. 2004. Data mining in bioinformatics using Weka. Bioinformatics 20:2479–2481. [PubMed]
7. Frieden T. R., Sterling T. R., Munsiff S. S., Watt C. J., Dye C. 2003. Tuberculosis. Lancet 362:887–899. [PubMed]
8. Gentleman R. C., et al. 2004. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol. 5:R80. [PMC free article] [PubMed]
9. Getahun H., Harrington M., O'Brien R., Nunn P. 2007. Diagnosis of smear-negative pulmonary tuberculosis in people with HIV infection or AIDS in resource-constrained settings: informing urgent policy changes. Lancet 369:2042–2049. [PubMed]
10. Goulding C. W., et al. 2004. Gram-positive DsbE proteins function differently from Gram-negative DsbE homologs. A structure to function analysis of DsbE from Mycobacterium tuberculosis. J. Biol. Chem. 279:3516–3524. [PubMed]
11. Goulding C. W., Perry L. J. 2003. Protein production in Escherichia coli for structural studies by X-ray crystallography. J. Struct. Biol. 142:133–143. [PubMed]
12. Hall M., et al. 2009. The WEKA data mining software: an update, p. 10–18 In Special Interest Group on Knowledge Discovery and Data Mining Explorer Newsletter, vol. 11 Association for Computing Machinery, New York, NY.
13. Ireton G. C., et al. 2010. Identification of Mycobacterium tuberculosis antigens of high serodiagnostic value. Clin. Vaccine Immunol. 17:1539–1547. [PMC free article] [PubMed]
14. Khan I. H., et al. 2005. Simultaneous serodetection of 10 highly prevalent mouse infectious pathogens in a single reaction by multiplex analysis. Clin. Diagn. Lab. Immunol. 12:513–519. [PMC free article] [PubMed]
15. Khan I. H., et al. 2006. Simultaneous detection of antibodies to six nonhuman-primate viruses by multiplex microbead immunoassay. Clin. Vaccine Immunol. 13:45–52. [PMC free article] [PubMed]
16. Khan I. H., et al. 2008. Profiling antibodies to Mycobacterium tuberculosis by multiplex microbead suspension arrays for serodiagnosis of tuberculosis. Clin. Vaccine Immunol. 15:433–438. [PMC free article] [PubMed]
17. Kunnath-Velayudhan S., et al. 2010. Dynamic antibody responses to the Mycobacterium tuberculosis proteome. Proc. Natl. Acad. Sci. U. S. A. 107:14703–14708. [PubMed]
17a. Lawn S. D., Zumila A. I. 2011. Tuberculosis. Lancet 378:57–72. [PubMed]
18. LoBue P. A., Enarson D. A., Thoen T. C. 2010. Tuberculosis in humans and its epidemiology, diagnosis and treatment in the United States. Int. J. Tuberc. Lung Dis. 14:1226–1232. [PubMed]
19. Mathworks Inc 2005. Matlab, 12th ed. Mathworks Inc., Natick, MA.
19a. Pakistan Institute of Legislative Development Transparency May 2010. Immunization in Pakistan—briefing paper. Pakistan Institute of Legislative Development and Transparency, Islamabad, Pakistan.
20. Ravindran R., et al. 2010. Validation of multiplex microbead immunoassay for simultaneous serodetection of multiple infectious agents in laboratory mouse. J. Immunol. Methods 363:51–59. [PubMed]
21. Reimer U., Reineke U., Schneider-Mergener J. 2002. Peptide arrays: from macro to micro. Curr. Opin. Biotechnol. 13:315–320. [PubMed]
22. Steingart K. R., et al. 2007. A systematic review of commercial serological antibody detection tests for the diagnosis of extrapulmonary tuberculosis. Postgrad. Med. J. 83:705–712. [PMC free article] [PubMed]
23. Systat Software, Inc. 2005. SigmaPlot, 9th ed. Systat Software, Inc., San Jose, CA.
24. Walls T., Shingadia D. 2004. Global epidemiology of paediatric tuberculosis. J. Infect. 48:13–22. [PubMed]
25. Walzl G., Ronacher K., Hanekom W., Scriba T. J., Zumla A. 2011. Immunological biomarkers of tuberculosis. Nat. Rev. Immunol. 11:343–354. [PubMed]
26. Weldingh K., Rosenkrands I., Okkels L. M., Doherty T. M., Andersen P. 2005. Assessing the serodiagnostic potential of 35 Mycobacterium tuberculosis proteins and identification of four novel serological antigens. J. Clin. Microbiol. 43:57–65. [PMC free article] [PubMed]
26a. WHO 2010. Global tuberculosis control report. WHO, Geneva, Switzerland.
27. Wu X., et al. 2010. Comparison of antibody responses to seventeen antigens from Mycobacterium tuberculosis. Clin. Chim. Acta 411:1520–1528. [PubMed]
Articles from Clinical and Vaccine Immunology : CVI are provided here courtesy of
American Society for Microbiology (ASM)