Supplementary Materials for - stm.sciencemag.org · years old, if they had an imminently terminal...

www.sciencetranslationalmedicine.org/cgi/content/full/5/203/203ra126/DC1

Supplementary Materials for

A Host-Based RT-PCR Gene Expression Signature to Identify Acute Respiratory Viral Infection

Aimee K. Zaas, Thomas Burke, Minhua Chen, Micah McClain, Bradly Nicholson,

Timothy Veldman, Ephraim L. Tsalik, Vance Fowler, Emanuel P. Rivers, Ronny Otero, Stephen F. Kingsmore, Deepak Voora, Joseph Lucas, Alfred O. Hero, Lawrence Carin,

Christopher W. Woods,* Geoffrey S. Ginsburg*

*Corresponding author. E-mail: [email protected] (G.S.G.); [email protected] (C.W.W.)

Published 18 September 2013, Sci. Transl. Med. 5, 203ra126 (2013)

DOI: 10.1126/scitranslmed.3006280

The PDF file includes:

Methods Table S1. Subject identification. Table S2. Probes and classifier weights, training on H3N2. Table S3. Probes and classifier weights, training on H1N1. Table S4. Classification data for real-world patients. Table S5. Comparison of the host viral infection score to commercially available rapid influenza testing. Fig. S1. Classification of H3N2 infection using microbiological and clinical phenotypes. Fig. S2. Classification of H1N1 infection using microbiological and clinical phenotypes. Fig. S3. Cross-viral validation of RT-PCR classification using clinical and microbiological phenotypes (train on H3N2 cohort and test on H1N1 cohort). Fig. S4. Cross-viral validation of RT-PCR classification using clinical and microbiological phenotypes (train on H1N1 cohort and test on H3N2 cohort). Fig. S5. Classification accuracy remains if H3N2 and H1N1 cohorts are combined, with training on half of the total cohort and testing on half of the total cohort. Fig. S6. Classification of virally infected emergency department subjects.

Supplementary Materials. Methods: Influenza H1N1 and H3N2 Inoculation: After 24 hrs in quarantine, we instilled one of four dilutions (1:10, 1:100, 1:1000, 1:10000) of 106

TCID50 Influenza A bilaterally into the nares of subjects (groups of 4 or 5 for each dilution). The virus was manufactured and processed under current good manufacturing practices (cGMP) by Baxter BioScience, (Vienna, Austria). Every eight hours for the first 5 days following inoculation), we collected blood into RNA PAXGene™ collection tubes (PreAnalytix; Franklin Lakes, NJ) according to manufacturers’ specifications. We obtained nasal lavage samples from each subject daily. This sample was used for qualitative and quantitative influenza PCR to assess the success and timing of infection, as the virus utilized would not grow sufficiently in eggs to be measured by quantitative culture. Blood and nasal lavage collection continued throughout the duration of the quarantine. All subjects received oral oseltamivir (Roche Pharmaceuticals) 75 mg by mouth twice daily as treatment or prophylaxis at 144 hours following inoculation. All patients were negative by rapid antigen detection of a nasal wash sample (BinaxNow Rapid Influenza Antigen; Inverness Medical Innovations, Inc) at time of discharge. We queried patients twice daily using a modified standardized symptom score for upper respiratory tract infection. The symptom score requires subjects to rank symptoms of upper respiratory infection (stuffy nose, scratchy throat, headache, cough, malaise and myalgias) on a scale of 0-3 of “no symptoms”, “just noticeable”, “bothersome but can still do activities” and “bothersome and cannot do daily activities”. For all cohorts, symptom scores were tabulated to determine if subjects became symptomatic from the respiratory viral challenge. Subjects were classified as clinically symptomatic if the sum of symptom scores totaled > 6 over the first 120 hours following inoculation. Subjects were classified as clinically symptomatic AND microbiologically infected if total symptom scores were > 6 over the first 120 hours following inoculation and virus was detected in nasal wash samples taken at 48, 72, 96 or 120 hours following inoculation. Detection of virus in samples taken at 24 hours following inoculation was felt to represent virus from inoculation but not indicative of replicating virus, and was not included in the definition (Supplementary Table 1) Subjects were classified as “asymptomatic, not infected ” if the symptom score was less than 6 over the five days of observation and viral shedding was not documented after the first 24 hours subsequent to inoculation as above. We tabulated the standardized symptom scores at the end of each study to determine both attack rate and time of maximal symptoms (time “T”). Symptom onset was defined as the timepoint at which a symptom score of at least 2 was achieved at 2 consecutive recordings. We performed a second healthy volunteer intranasal challenge using influenza A H3N2 (A/Wisconsin/67/2005) in collaboration with Retroscreen Virology, Ltd (Brentwood, UK) in 17 pre-screened volunteers who provided informed consent. Inoculation procedures, sample collection schemes, duration of quarantine, and antiviral treatment/prophylaxis were identical to the H1N1 study. Pandemic 2009 H1N1 Real-World Cohort: Subjects were recruited from the Duke University Medical Center Emergency Department (DUMC-Level 1 Trauma Center with annual census of 65,000). This study was approved by the Institutional Review Board at each institution and written, informed consent was obtained for all study participants or their legal designates. Subjects were screened between September 1 and December 31, 2009. Subjects were considered for the enrollment if they had a known or suspected influenza infection on the basis of clinical data at the time of screening and if they exhibited two or more signs of systemic inflammation (SIRS) within a 24-hour period. Subjects were excluded if <18

years old, if they had an imminently terminal co-morbid condition, if they had recently been treated with an antibiotic for a viral, bacterial, or fungal infection, or if they were participating in an ongoing clinical trial. Trained study coordinators at each site reviewed and abstracted vital signs, microbiology, laboratory, and imaging results from the initial ED encounter and at 24-hour intervals if patient was admitted. Following hospital discharge, research personnel abstracted the duration of hospitalization, length of ICU stay, in-hospital mortality, timing and appropriateness of antimicrobial administration, and microbiologic-culture results from the medical record. In addition to residual respiratory samples collected as part of routine care, an NP swab was collected from each enrolled subject. Total nucleic acids were extracted from nasal swab or wash isolates with the EZ1 Biorobot and the EZ1 Virus Mini Kit v2.0 (Qiagen). 2009 H1N1 virus was confirmed in 20 ul detection reactions, Qiagen One-Step RT-PCR (Qiagen) reagents on a LightCycler v2.0 (Roche) using the settings and conditions recommended in the CDC Realtime RTPCR (rRTPCR) Protocol for Detection and Characterization of Swine Influenza (version 2009). The primers and probes were as described in the CDC protocol and obtained from Integrated DNA Technologies. We included mRNA expression data obtained concurrently with the H1N1 cohort from 45 gender-matched, healthy controls. RT-PCR Card Development: RT-PCR TLDA Card: Two card designs were utilized in developing the RT-PCR data. Pre-designed probes representing genes comprising the acute respiratory viral signature{Zaas, 2009 #24} as well as a selection of genes noted to provide early classification of symptomatic individuals were selected from the Applied Biosystems probe library for Taqman Low Density Array (TLDA) card design. 48- probe sets were configured on 384 well cards (n=8 samples per card). To test reproducibility, several probes were redundant on the cards and a subset of individuals were run in duplicate. In the initial trial, RNA from each subject from the subjects in the HRV, RSV and H3N2 challenge studies was tested at time 0 (pre-inoculation) and time T (time of maximal symptoms). Following the results of the first application of the TLDA cards, the probeset was revised as further described below. The “second generation” card contained the following primer pairs: 29 pairs representing exons of genes derived from the acute respiratory viral factor; 7 primer pairs representing genes identified through a self-organizing map analysis of the H3N2 data{Huang, #13}; 3 control genes (including GAPDH); and 9 genes derived from the original microarray data. The “second generation” card was used for the analyses reported here. The second generation TLDA card was used to analyze all subjects H3N2 and influenza A H1N1/Brisbane (H1N1) challenges at the pre-inoculation timepoint (Time 0) and the time of maximal symptoms (time T) following inoculation with virus. Raw TLDA data was loaded into RQ Manager 1.2 for determination of Ct values. Next, .sdm-Result Data files (up to 10 plates per batch) were exported from RQ Manager and loaded into RealTime StatMiner 4.1 for data normalization and analysis. Experimental Design .txt or .csv file were generated with Sample Name, Plate, and Experimental Group for each sample data set, and loaded into RealTime StatMiner for DDCT analysis. For all studies, relative expression was quantified using the 2-DDCT methodology with GAPDH as a normalizing gene. 2-DDCT values were log-transformed and classification between symptomatic and asymptomatic individuals determined by ANOVA (time 0 vs time T). Statistical Methods, Elastic Net The working procedure of the ENet classifier is as follows: 1. Pre-processing. We first log transform the 2e-ddct data to obtain the data matrix X of dimension n*p, where n is the number of samples and p is the number of genes. We subtract the sample mean from X and append an all-one column to X to account for the shift (intercept) term in the regression.

2. ENet training. Using the labels of the samples Z and the pre-processed data X, we train a probit regression model Z = sign(Y), Y = X*beta+noise with an Elastic Net prior on the weights (beta). After training we obtain the posterior distribution of the weights (classifier). The ENet classifier favors sparse and grouped variable selection, with a combination of L1 and L2 penalties on the weights. Detailed explanation of the model can be found in. 3. Testing. For an unseen testing sample x, we do the same data pre-processing and the predictive probability can be obtained by integrating out the classifier using the posterior distribution in the probit model. The above predictive probability is between 0 and 1, and classification decision is made by using the threshold 0.5. The final ENet score is obtained by multiplying a constant (e.g., 40) to the predictive probability. Matlab code Name Functionality Main.m Code for generating all the results according to ANALYSIS_PLAN_12_3_10.docx. ENet Cls Fast.m Code for designing the Bayesian Elastic Net classi_er. ENet Cls Call.m Code for calling ENet Cls Fast.m. ENet Cls LOO.m Code for Leave-One-Out cross-validation, by calling ENet Cls Call.m. H3N2PCR Collect.m Code for collecting H3N2 time-course PCR data from raw data _le. H1N1PCR Collect.m Code for collecting H1N1 time-course PCR data from raw data _le. TimeTPCR Collect.m Code for collecting time T (peak symptom) PCR data for H3N2 and H1N1. ED Collect.m Code for collecting the separate ED testing dataset. Gene ROC.m Code for generating the ROC curve. Gene Plot.m Code for plotting the response using different marks for different labels. Train Plot.m Code for plotting the classifer and the response.

Supplementary Table 1. Subject Identification.

Study Subject # Total Score (5 days)

Clinically Symptomatic

Clinically Symptomatic and Microbiologically

Infected H3N2 001 28 Y Y

002 0 N N* 003 0 N N* 004 0 N N 005 45 Y Y 006 30 Y Y 007 31 Y Y 008 34 Y Y 009 5 N N 010 14 Y Y 011 4 N N 012 12 Y Y 013 7 Y Y 014 1 N N 015 7 Y Y 016 1 N N 017 1 N N

H1N1 001 5 N N

002 13 Y N 003 10 Y N 004 0 N N 005 5 N N 006 7 Y Y 007 16 Y N 008 12 Y Y 009 33 Y Y 010 26 Y Y 011 2 N N 012 34 Y Y 013 17 Y Y 014 0 N N 015 4 N N 016 0 N N 017 19 Y Y 018** 6 N N 019 0 N N 020 23 Y Y 021 6 Y Y 022 0 N N 023 0 N N 024 0 N N

* EXCLUDED IN SECONDARY ANALYSIS AS SUBJECTS HAD EVIDENCE OF ANTIBODY SEROCONVERSION AT 28 DAYS FOLLOWING STUDY WITH >4 FOLD INCREASE IN INFLUENZA ANTIBODY TITER AS COMPARED TO BASELINE. ** EXCLUDED IN ALL ANALYSES AS SYMPTOMS BEGAN LATE AND WERE FELT TO BE RELATED TO INFECTION ACQUIRED IN THE FACILITY, NOT PRIMARY INFECTION RELATED TO INOCULATION.

Supplementary Table 2. Probes and classifier weights, training on H3N2.

Gene Name Source Probe ID Weight IFI27 Acute Respiratory

Viral Factor Hs00271467_m1 0.365764

SIGLEC1 Acute Respiratory Viral Factor

Hs00988063_m1 0.256661

IFI44L Acute Respiratory Viral Factor

Hs00199115_m1 0.244360

RSAD2 Acute Respiratory Viral Factor

Hs00369813_m1 0.176117

IFI44 Acute Respiratory Viral Factor

Hs00197427_m1 0.132044

ISG15 Acute Respiratory Viral Factor

Hs01921425_s1 0.013072

LOC26010 Acute Respiratory Viral Factor

Hs00274702_m1 0.011503

LAMP3 Acute Respiratory Viral Factor

Hs00180880_m1 0.011203

SERPING1 Acute Respiratory Viral Factor

Hs00934330_m1 0.010568

IFIT1 Acute Respiratory Viral Factor

Hs01911452_s1 0.007308

OAS3 Acute Respiratory Viral Factor

Hs00196324_m1 0.005626

OASL Acute Respiratory Viral Factor

Hs00984390_m1 0.004477

MX1 Acute Respiratory Viral Factor

Hs00182073_m1 0.004222


Hs00973637_m1 0.003973

LY6E Acute Respiratory Viral Factor

Hs00158942_m1 0.003591

SEPT4 Acute Respiratory Viral Factor

Hs00365352_m1 0.003591

HERC5 Acute Respiratory Viral Factor

Hs00180943_m1 0.003310


Hs00155468_m1 0.003251


Hs00942650_m1 0.002795

XAF1 Downregulated in Asymptomatic

Subjects

Hs00213882_m1 0.002441

ATF3 Acute Respiratory Viral Factor

Hs00910173_m1 0.002359

RTP4 Acute Respiratory Viral Factor

Hs00223342_m1 0.002341


Hs00533665_m1 0.002288

TNFAIP6 Acute Respiratory Viral Factor

Hs01113602_m1 0.001787

GBP1 Acute Respiratory Viral Factor

Hs00977005_m1 0.001726


Hs00242571_m1 0.001685

CTSL1 No differential expression between

symptomatic and asymptomatic

Hs00377632_m1 0.001654


Hs00202721_m1 0.001332

CCL2 Acute Respiratory Viral Factor

Hs00234140_m1 0.001132

DDX58 Acute Respiratory Viral Factor

Hs00204833_m1 0.001127

LILRB2;LILRB1 No differential expression between


Hs00601427_g1 0.000817

ADAR Downregulated in Asymptomatic

Subjects

Hs00241666_m1 0.000485

SOCS1 Downregulated in Asymptomatic

Subjects

Hs00705164_s1 0.000476

C13orf18 No differential expression between


Hs00228336_m1 0.000456

CUZD1 No differential expression between


Hs00222774_m1 0.000403

PRSS21 No differential expression between


Hs00199035_m1 0.000396


Subjects

Hs00751962_s1 0.000382

NOD2 Downregulated in Asymptomatic

Subjects

Hs01550762_g1 0.000326

RPL30 Control Hs00265497_m1 0.000313 GM2A No differential

expression between symptomatic and

asymptomatic

Hs00166197_m1 0.000293

HLA-DOB No differential expression between


Hs00157950_m1 0.000280

NLRP3 Downregulated in Asymptomatic

Subjects

Hs00366465_m1 0.000235

GAPDH Control Hs99999905_m1 0.000231 IL16 Acute Respiratory

Viral Factor Hs00913646_g1 0.000176

ENOSF1 No differential expression between

symptomatic and

Hs00213845_m1 0.000092

asymptomatic NDUFA10 No differential


asymptomatic

Hs00190004_m1 0.000045

PPIA Control Hs99999904_m1 0.000037 SOCS2 Downregulated in

Asymptomatic Subjects

Hs00919620_m1 0.000017

Supplementary Table 3. Probes and classifier weights, training on H1N1.

Gene Name Source Probe ID Weight IFI27 Acute Respiratory


SIGLEC1 Acute Respiratory Viral Factor

Hs00988063_m1 0.022459

IFI44L Acute Respiratory Viral Factor

Hs00199115_m1 0.254542

RSAD2 Acute Respiratory Viral Factor

Hs00369813_m1 0.107491


Hs00197427_m1 -0.000639

ISG15 Acute Respiratory Viral Factor

Hs01921425_s1 0.001004

LOC26010 Acute Respiratory Viral Factor

Hs00274702_m1 0.006461

LAMP3 Acute Respiratory Viral Factor

Hs00180880_m1 0.001565

SERPING1 Acute Respiratory Viral Factor

Hs00934330_m1 0.004257


Hs01911452_s1 0.000955


Hs00196324_m1 0.003852

OASL Acute Respiratory Viral Factor

Hs00984390_m1 0.003782

MX1 Acute Respiratory Viral Factor

Hs00182073_m1 0.002339


Hs00973637_m1 0.002717

LY6E Acute Respiratory Viral Factor

Hs00158942_m1 0.004525

SEPT4 Acute Respiratory Viral Factor

Hs00365352_m1 0.001494

HERC5 Acute Respiratory Viral Factor

Hs00180943_m1 0.001345


Hs00155468_m1 0.002421


Hs00942650_m1 0.002186

XAF1 Downregulated in Asymptomatic

Hs00213882_m1 0.001097

Subjects ATF3 Acute Respiratory


RTP4 Acute Respiratory Viral Factor

Hs00223342_m1 0.002054


Hs00533665_m1 0.001538

TNFAIP6 Acute Respiratory Viral Factor

Hs01113602_m1 0.000948

GBP1 Acute Respiratory Viral Factor

Hs00977005_m1 0.000930


Hs00242571_m1 0.004157

CTSL1 No differential expression between


Hs00377632_m1 0.000953


Hs00202721_m1 0.000953

CCL2 Acute Respiratory Viral Factor

Hs00234140_m1 0.000374

DDX58 Acute Respiratory Viral Factor

Hs00204833_m1 0.000800

LILRB2;LILRB1 No differential expression between


Hs00601427_g1 0.000777

ADAR Downregulated in Asymptomatic

Subjects

Hs00241666_m1 0.000420


Subjects

Hs00705164_s1 -0.000133

C13orf18 No differential expression between


Hs00228336_m1 -0.000216

CUZD1 No differential expression between


Hs00222774_m1 -0.000372

PRSS21 No differential expression between


Hs00199035_m1 -0.000603


Subjects

Hs00751962_s1 -0.001025

NOD2 Downregulated in Asymptomatic

Subjects

Hs01550762_g1 0.000142

RPL30 control Hs00265497_m1 -0.000260 GM2A No differential


asymptomatic

Hs00166197_m1 0.000743

HLA-DOB No differential expression between


Hs00157950_m1 -0.000211

NLRP3 Downregulated in Asymptomatic

Subjects

Hs00366465_m1 0.000340

GAPDH control Hs99999905_m1 0.000409 IL16 Acute Respiratory

Viral Factor Hs00913646_g1 -0.000106

ENOSF1 No differential expression between


Hs00213845_m1 -0.000091

NDUFA10 No differential expression between


Hs00190004_m1 0.000056

PPIA control Hs99999904_m1 -0.000134 SOCS2 Downregulated in

Asymptomatic Subjects

Hs00919620_m1 -0.000180

Supplementary Table 4. Classification data for real-world patients. Patients presenting to the emergency department with febrile illness were tested via blood culture, urine culture, streptococcal urine antigen, wound culture (if appropriate) and respiratory sample viral PCR and/or bacterial culture. Etiologic agent was assigned via a standardized adjudication process. For all patients, blood samples were taken at time of presentation into Paxgene™ tubes (Preanalytix) for RNA extraction. From the larger study, a subset of patients was selected to have RNA tested by the RT-PCR assay described here. Probability refers to the likelihood of a sample being classified as “viral infection”, with a probability of 0.5 or greater assigned as “viral”. Three samples from patients with respiratory viral infection were classified as “non-viral” and four samples from patients with bacterial infection were classified as “viral”. Mean white blood cell count for patients with viral infection was 8.88+/-4.6/mm3 (range 2.8-15.4/ mm3) and mean white blood cell count for patients with bacterial infection was 14.34 +/-7.09/ mm3 (range 2.0-32.7/ mm3). Cx = culture; Ag = antigen.

Sample Pathogen Probability Classification

T101-C H1N1 0.9257 viral T11-C H1N1 0.9502 viral T12-C FluA 0.8225 viral T13-C H1N1 0.8833 viral T14-C H1N1 0.3504 Non-

viral

T15-C H1N1 0.3017 Non-viral

T16-C H1N1 0.9073 viral T17-C H1N1 0.8689 viral T18-C H1N1 0.8348 viral T19-C H1N1 0.8666 viral T2-C H1N1 0.7425 viral T20-C Rhino 0.5601 viral T3-C H1N1 0.5138 viral T4-C H1N1 0.7842 viral T5-C H1N1 0.7510 viral T58-C Rhino 0.4116 Non-

viral

T6-C H1N1 0.6831 viral T60-C H1N1 0.9134 viral T61-C Rhino 0.7154 viral T62-C Rhino 0.7307 viral T63-C H1N1 0.8867 viral

T64-C H1N1 0.9252 viral T65-C H1N1 0.8153 viral T66-C H1N1 0.9129 viral T67-C H1N1 0.9041 viral T7-C H1N1 0.8783 viral T8-C FluA 0.9001 viral T9-C FluA 0.9057 viral

Sample Pathogen Culture Data Probability Classification

T1-C S.aureus Blood cx 0.0878 Non-viral

T10-C S.pneumo Blood cx 0.3458 Non-viral

T100-C S.pneumo Sputum cx 0.0163 Non-viral

T102-C S.pneumo Sputum cx 0.1379 Non-viral

T103-C S.pneumo Urine Ag 0.8855 viral T104-C S.pneumo Blood cx 0.0549 Non-viral T24-C S.aureus Blood cx,

cellulitis 0.4104 Non-viral

T25-C S.aureus Blood cx, cellulitis

0.3426 Non-viral

T26-C S.aureus Blood cx 0.0653 Non-viral T27-C S.aureus Blood cx,

sputum cx 0.0624 Non-viral

T28-C S.aureus Blood cx, Sputum cx

0.7299 viral

T29-C S.pneumo Urine Ag, CXR

0.3938 Non-viral

T30-C S.pneumo Urine Ag, sputum cx

0.7384 viral


0.1127 Non-viral

T32-C S.pneumo Blood cx, Urine Ag

0.3377 Non-viral

T33-C S.pneumo Blood cx, Urine Ag, CXR

0.1288 Non-viral


0.0576 Non-viral


0.0333 Non-viral


0.2982 Non-viral

T37-C S.pneumo Urine Ag 0.1416 Non-viral T38-C Staph Blood cx 0.2603 Non-viral T39-C S.pneumo Urine Ag,

CXR 0.0298 Non-viral


0.1159 Non-viral

T41-C S.pneumo Blood cx, Urine Ag, CXR

0.2425 Non-viral

T42-C S.aureus Blood cx 0.7468 viral

T43-C S.aureus Blood cx, CXR

0.2490 Non-viral


0.1200 Non-viral

T45-C S.pneumo Blood cx, CXR

0.1357 Non-viral

T46-C S.pneumo Sputum Cx, Urine Ag, CXR

0.1167 Non-viral


0.2167 Non-viral

T48-C S.pneumo Pleural fluid cx; urine Ag

0.3705 Non-viral




0.0824 Non-viral


0.0602 Non-viral


0.0417 Non-viral

T55-C S.aureus Blood cx; wound cx

0.1607 Non-viral

T56-C S.aureus Blood cx; wound cx

0.3768 Non-viral

T57-C E.coli Urine cx 0.3171 Non-viral

Supplementary Table 5. Comparison of the host viral infection score to commercially available rapid influenza testing. Probability of detection is the likelihood of a test reporting that a person with influenza A has influenza A (true positive); probability of false alarm indicates that a person who is not sick with influenza A actually has influenza A (false positive). Most rapid tests currently available have a relatively high false negative rate. Data taken from references 3-9 in the manuscript. Assay Tested Sensitivity

(Probability of Detection)

Specificity (1-Probability of False Alarm )

SD Bioline Influenza Ag A/B

0.45 0.99

Influ A/B Respistrip; SD Bioline; Directigen EZ Flu A+B and QuickVue Influenza A+B; QuickVue Influenza A+B

0.32-0.5 0.98

ClearView Exact Influenza A+B

0.19 0.99

QuickVue Influenza A+B; Influenzatop

0.64 0.99

SD Bioline Influenza Ag A/B

0.7 0.97

Supplementary Figures.

Supplementary Figure 1. Classification of H3N2 infection using microbiological and clinical phenotypes. RT-PCR accurately classifies individuals with symptomatic H3N2 infection, when using only individuals who met both clinical and microbiologic definitions of symptomatic infection. Symptomatic individuals (total symptom score > 6 over first 120 hours following inoculation with virus and had microbiologic evidence of infection) are represented with blue dots, asymptomatic individuals with red dots. ENet Score (y axis) represents probability of having viral infection, with a score of 20

indicative of 50% probability. Pd = probability of detection. Pf= probability of false discovery. Individuals with conflicting symptom scores and microbiologic data are excluded in this analysis. Classification error is 0/15 (0%; AUC = 1).

Supplementary Figure 2. Classification of H1N1 infection using microbiological and clinical phenotypes. RT-PCR accurately classifies individuals with symptomatic H1N1 infection when using individuals who met both clinical and microbiologic definition of symptomatic infeciton. Symptomatic

individuals (total symptom score > 6 over first 120hours following inoculation with virus) are represented with blue dots, asymptomatic individuals with red dots. ENet Score (y axis) represents probability of having viral infection, with a score of 20 indicative of 50% probability. Pd = probability of detection. Pf= probability of false discovery. Individuals with conflicting symptom scores and microbiologic data are excluded in this analysis. Classification error is 1/15 (7%).

Supplementary Figure 3. Cross-viral validation of RT-PCR classification using clinical and microbiological phenotypes (train on H3N2 cohort and

test on H1N1 cohort).RT-PCR accurately classifies individuals with symptomatic H3N2 and H1N1 infection when training on one virus and validating in the second cohort. When utilizing subjects who met the clinical AND microbiologic definitions of infection for cross validation, classification error was 1/15 (6.7%) when training on H3N2 and testing on H1N1.

Supplementary Figure 4. Cross-viral validation of RT-PCR classification using clinical and microbiological phenotypes (train on H1N1 cohort and test on H3N2 cohort).RT-PCR accurately classifies individuals with symptomatic H3N2 and H1N1 infection when training on one virus and validating in the second cohort. When utilizing subjects who met the clinical AND microbiologic definitions of infection for cross validation, classification error was 2/15 (13%, 2 false negatives) for training on H3N2 and testing on H1N1.

Supplementary Figure 5. Classification accuracy remains if H3N2 and H1N1 cohorts are combined, with training on half of the total cohort and testing on half of the total cohort. Shown is additional validation of the experimental inoculation data. The data from the H1N1 and H3N2 cohorts were combined into one total set of data, and then randomly partitioned (50% for model training and 50% for testing). We considered 100 random partitions. The mean AUC across these 100 runs was 0.975 +/- 0.034.

Supplementary Figure 6. Classification of virally infected emergency

department subjects. We tested the classifier using only bacterially infected and

virally infected subjects from the emergency department cohort, without the use

of healthy controls. The performance of the classifier in this testing scenario is

similar with misclassification of 4/39 (10.4%) patients with bacterial infection and

only 3/28 patients with viral infection resulting in a sensitivity of 89% (95% CI 72-

98%) and a specificity of 94% (95% CI 86-99%).

Supplementary Materials for - stm.sciencemag.org · years old, if they had an imminently terminal...

Documents

Transcript of Supplementary Materials for - stm.sciencemag.org · years old, if they had an imminently terminal...