sep 24 08 fwf seminar eda final
Post on 01-Dec-2021
10 Views
Preview:
TRANSCRIPT
9/28/2008
1
New diagnostic methods for infectious diseases
Shigetoshi Eda, PhD
Research Assistant ProfessorCenter for Wildlife Health
Department of Forestry, Wildlife, and Fisheries
A B
Fig. 1. (A) Lab-on-a-Chip developed at the Oak Ridge National Laboratory (ORNL) and (B) a hand-held device developed based on the Chip at the Sandia National Laboratories, Courtesy of Dr. Robert Foote at ORNL..
A BB
Fig. 1. (A) Lab-on-a-Chip developed at the Oak Ridge National Laboratory (ORNL) and (B) a hand-held device developed based on the Chip at the Sandia National Laboratories, Courtesy of Dr. Robert Foote at ORNL..
Wednesday, September 24, 2008 - 160 Plant Biotech
New diagnostic test for Johne’s disease in cattle
I. Past - present
Topics
Point-of-care diagnostic device for infectious diseases
II. Present -
Part 1
New diagnostic test for Johne’s disease in cattle
9/28/2008
2
Pathogen or Etiologic agent ---
Bacteria, Mycobacterium avium subsp. paratuberculosis (MAP)
Host --- Primarily ruminants in livestock and wildlife. Many other animals can be carriers or sporadically infected.
Signs --- Diarrhea and weigh loss. Often cows will show no signs of the disease for first 3-5 years of infection.
Pathology --- The primary site of infection is small intestine.
Transmission --- Oral-fecal transmission
Why studyJohne’s disease?
Highly prevalent worldwide; in the US, Africa, Australia, New Zealand and Europe the prevalence of infected cattle herds varies from 10-60%.
Annual loss of more than $200 million to the US dairy industry. These loss are mainly due to reduced milk production and early culling.
The causative bacteria of Johne’s disease is suspected to cause or worsen a human disease, called Crohn’s disease
Obstacles in Johne’s disease control
Antibiotics are expensive and require a long course of treatment: not a practical option for the livestock industry.
Vaccines have had limited success.
So, current management is based on: Diagnosis and culling of infected animals.
However …
9/28/2008
3
Problems in Johne’s disease diagnosis
Detection of bacteria in fecesa. Culture
Time consuming -- up to 4 months (!) Expensive -- $20/sample
b. DNA detectionLabor intensive – requires a skilled examinerExpensive -- $25/sample
Detection of antibodies in serumRapid -- Half a dayCheap -- $5/sampleEasy -- but very low sensitivity!
Antibody detection VERY LOW sensitivity!
Antibody detection (ELISA)
Author Year Sensitivity (%) Specificity
Reichel 1999 31.1 97.9
Stable 2002 25 94
McKenna 2005 16.6 97.113.9 95.9
Sweeney 2006 13.5 98.1
Wells 2006 27.5 95.327.8 99.7
Sensitivity at30% means
Only 1 in 3 diseased cattle can be detected and removed.
9/28/2008
4
The remaining cattle will spread the disease to the healthy cattle in the herd.
Help me…
So, we urgently dneed a more
sensitive test!
9/28/2008
5
What’s wrong with current antibody-detection tests?
This is Volkswagen Golf.
Not hard to recognize.
Also, an emblem and badge are available on its surface that allow for provide easy identification of the car.
What’s wrong with current antibody-detection tests?
But this is what current tests use as antigens.
These internal parts are shared with other types of Volkswagen cars, and interfere with our car identification.
Our idea ….
Extract just the surface antigens
from the bacteria without destroying the organisms and use them for antibody detection.
9/28/2008
6
Approach
Since the surface of the MAP bacterium is oily, we tested
Lipids
Surface
is oily, we tested several organic solutions for gentle extraction of MAP surface antigens.
Internal space
MethodBacilli of MAP
Mixed with buffers or organic solutions
Gentle agitation
MAP
solvent
MAPCentrifuged
Immobilized the antigens in the supernatant
Reacted with serum from JD-negative or JD-positive cow
MAP
JD-positive serumJD-negative serum
Detection of antibody binding by ELISA
Extraction of surface antigens
Before
After
9/28/2008
7
IgG binding to MAP surface antigens extracted with various organic solutions
0.8
1
1.2
1.4
ding
at 4
15 n
m)
0.8
1
1.2
1.4
ding
at 4
15 n
m)
0.8
1
1.2
1.4
ding
at 4
15 n
m)
0.8
1
1.2
1.4
ding
at 4
15 n
m)
0
0.2
0.4
0.6
0.8
MeOH EtOH ProOH ACN Acetone DCM CHCl3 Ether Hexane
IgG
bind
(Abs
orba
nce
a
0
0.2
0.4
0.6
0.8
MeOH EtOH ProOH ACN Acetone DCM CHCl3 Ether Hexane
IgG
bind
(Abs
orba
nce
a
MeOH: methanol, EtOH: ethanol, ProOH: iso-propanol, ACN, acetonitrile, DCM, dichloromethane, CHCl3: chloroform
: JD-positive serum: JD-negative serum: No serum
0
0.2
0.4
0.6
0.8
MeOH EtOH ProOH ACN Acetone DCM CHCl3 Ether Hexane
IgG
bind
(Abs
orba
nce
a
0
0.2
0.4
0.6
0.8
MeOH EtOH ProOH ACN Acetone DCM CHCl3 Ether Hexane
IgG
bind
(Abs
orba
nce
a
MeOH: methanol, EtOH: ethanol, ProOH: iso-propanol, ACN, acetonitrile, DCM, dichloromethane, CHCl3: chloroform
: JD-positive serum: JD-negative serum: No serum : JD-positive serum: JD-negative serum: No serum
Effects of various ethanol concentrations on IgG binding to MAP or MAA antigens
1.2
1.4
1.6
1.8
ng 415
nm)
1.2
1.4
1.6
1.8
ng 415
nm)
1.2
1.4
1.6
1.8
ng 415
nm)
1.2
1.4
1.6
1.8
ng 415
nm)
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60 70 80 90 100
EtOH (%)
IgG
bind
in(A
bsor
banc
e at
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60 70 80 90 100
EtOH (%)
IgG
bind
in(A
bsor
banc
e at
: MAP antigens : MAA antigens : No antigen
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60 70 80 90 100
EtOH (%)
IgG
bind
in(A
bsor
banc
e at
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60 70 80 90 100
EtOH (%)
IgG
bind
in(A
bsor
banc
e at
: MAP antigens : MAA antigens : No antigen
Diagnostic specificity and sensitivity of EVELISA EVELISA= Ethanol-Vortex ELISA
3.0
4.0
5.0
(S/P
val
ue)
3.0
4.0
5.0
(S/P
val
ue)
3.0
4.0
5.0
(S/P
val
ue)
3.0
4.0
5.0
(S/P
val
ue)
0
1.0
2.0
IgG
bind
ing
0
1.0
2.0
IgG
bind
ing
JD-negative serum (n=38)
JD-positive serum (n=51)
0
1.0
2.0
IgG
bind
ing
0
1.0
2.0
IgG
bind
ing
JD-negative serum (n=38)
JD-positive serum (n=51)
9/28/2008
8
Diagnostic specificity and sensitivity of EVELISA
3.0
4.0
5.0
(S/P
val
ue)
3.0
4.0
5.0
(S/P
val
ue)
3.0
4.0
5.0
(S/P
val
ue)
3.0
4.0
5.0
(S/P
val
ue)
0
1.0
2.0
IgG
bind
ing
0
1.0
2.0
IgG
bind
ing
JD-negative serum (n=38)
JD-positive serum (n=51)
0
1.0
2.0
IgG
bind
ing
0
1.0
2.0
IgG
bind
ing
JD-negative serum (n=38)
JD-positive serum (n=51)
100% sensitivity !!
The novel test improved the sensitivity of Johne’s disease diagnosis from 30% to 100%.
Part 1 Summary
Currently being evaluated by a veterinary diagnostic company for future (~two years) marketing.
Once in market, the test would have a positive impact on Johne’s disease control efforts.
Part 1 Summary
Angus Journal
Funding
9/28/2008
9
Part 2
Point-of-care diagnostic device for infectious diseases
A BA BBA
Fig. 1. (A) Lab-on-a-Chip developed at the Oak Ridge National Laboratory (ORNL) and (B) a hand-held device developed based on the Chip at the Sandia National Laboratories, Courtesy of Dr. Robert Foote at ORNL..
A
Fig. 1. (A) Lab-on-a-Chip developed at the Oak Ridge National Laboratory (ORNL) and (B) a hand-held device developed based on the Chip at the Sandia National Laboratories, Courtesy of Dr. Robert Foote at ORNL..
Main idea
Acquire and analyze Multiplex data for diagnosis of infectious diseases.
Achieve this using:
1. a ‘Microfluidic’ Flow Cytometry;
2. Machine Learning analysis of the data.
Why Multiplex analysis?y p y
9/28/2008
10
Why multiplex analysis?
antigen A
antigen B
antigen CAntibody production against
30% of cases
50% of cases
40% of cases
antigen D
antigen E
against
antigen F
early stage
late stage
antigen Gvirulent strain infection
avirulent strain infection
Antigen A Antigen B Antigen C Total
Why multiplex analysis?
30% 50% 40% 100%
Improved accuracy of diagnostic tests
Why multiplex analysis?
antigen A
antigen B
antigen CAntibody production against
30% of cases
50% of cases
40% of cases
Additional information for effective medical treatments
antigen D
antigen E
against
antigen F
early stage
late stage
antigen Gvirulent strain infection
avirulent strain infection
9/28/2008
11
How we identify candidate antigens?Phage Display Library (PDL)
• Library of genetically modified phage• Each phage clone express a specific sequence of peptide• The diversity of peptides in the PDL is higher than one billion
Peptide
• Quick, easy, and high-throughput screening method• Unbiased --- No preceding knowledge required• Can be used for any infectious disease applications
Library of billions of different peptides
Phage
Advantage
PDL ~ Screening method ~
Antibodies in diseased individuals
PDL ~ Screening method ~
Antibodies in diseased individuals
9/28/2008
12
PDL ~ Screening method ~
Antibodies in diseased individuals
PDL ~ Screening method ~
• Antibody binding assay• Read peptide sequence
Preliminary data
Eda et al. 2004 “Selection of peptides recognized by human antibodies against the surface of Plasmodium falciparum (malaria)-infected erythrocytes“
9/28/2008
13
OK, so Multiplex analysis maybe important.
How can we do such assay?
Microfluidic Flow CytometryMicrofluidic Flow Cytometry is a portable, rapid method that is capable of conducting Multiplex assays in a single run.
A BA BBApplications of flow cytometry:
• Counting and identification of microorganisms
Fig. 1. (A) Lab-on-a-Chip developed at the Oak Ridge National Laboratory (ORNL) and (B) a hand-held device developed based on the Chip at the Sandia National Laboratories, Courtesy of Dr. Robert Foote at ORNL..
Fig. 1. (A) Lab-on-a-Chip developed at the Oak Ridge National Laboratory (ORNL) and (B) a hand-held device developed based on the Chip at the Sandia National Laboratories, Courtesy of Dr. Robert Foote at ORNL..
• Host leukocyte counting
• Biomarker detection
• Antibody detection High-throughput analysis is also possible.
Microfluidic Flow Cytometry (contd.)Typical method:
Different size of beads
Dif ferent ant igens on dif ferent size of beads
Mix the beads (* one test tube for mult i ant igen analysis)
Dif ferent size of beads
Dif ferent ant igens on dif ferent size of beads
Mix the beads (* one test tube for mult i ant igen analysis)Mix the beads (* one test tube for mult i- ant igen analysis)
React with blood samples
React then with f luorescing- secondary ant ibody
Flow cytometric analysis of ant ibody binding
Mix the beads (* one test tube for mult i- ant igen analysis)
React with blood samples
React then with f luorescing- secondary ant ibody
Flow cytometric analysis of ant ibody binding
9/28/2008
14
Microfluidic Flow Cytometry (contd.)
Expected result:
ding
leve
l
Antigen BAntigen E
Antigen FAntigen C
Size of beads
Antib
ody
bind
Antigen AAntigen G
Antigen D
0
10000
20000
30000
IgG
bin
ding
(flu
ores
cent
inte
nsity
)
7 8 9 10 11 12
0.0
0.4
0.8
log(Test)
0 0 2 0 4 0 6 0 8 1
00.
40.
8
RO
C(t)
A B C
D
10000
20000
30000
40000
50000
60000
70000
80000
90000
10000
20000
30000
40000
50000
60000
70000
80000
90000
Preliminary data
Fig. 1. Binding of IgG in bovine sera to MAP-antigen-coated magnetic beads.
A. Magnetic beads coated with MAP antigen were treated with no serum (1), JD-negative serum (2), orJD-positive serum (3). IgG molecules bound to the beads were labeled with FITC-labeled secondaryantibody and detected by using a flow cytometer.
B. Magnetic beads coated with MAP antigen were treated with 30 JD-negative sera (1), or 30 JD-positive serum (2). IgG molecules bound to the beads were labeled with FITC-labeled secondaryantibody and detected by using a flow cytometer.
C. Log transformation data of Fig. 3B. Dotted line: JD-negative sera, Solid line: JD-positive sera
D. ROC analysis of Fig. 3B, AUC=0.991 with standard error = 0.012
01 2 3
I 0 0.2 0.4 0.6 0.8 1
t0
1 20
1 2
Complicated data × number of samples
9/28/2008
15
OK, so Flow-Cytometric Multiplexanalysis seems doable.
But how can we analyze such complicated data?
Machine learningSome of these methods have been used for classification of complicated data…
Artificial Neural Network(ANN)
Support Vector Machine(SVM)
Ensemble LearningRandom Forest, Logistic boost, Model Tree etc.
Machine learningcan convert complicated data into simple classifications
Complicated Simple
Hidden layerOutput layer
Input layer
Man
Woman
Output layer
9/28/2008
16
Preliminary dataClinical data from Crohn’s disease (CD) patients and healthy individuals (n=73)
Each sample was tested for antibody binding to Antigens A-F using a flow cytometer.
Randomly selected 20 data were used for training of machine learning algorithms and the remaining data (n=53) were used for accuracy evaluation.
Antigen A Antigens A-F
Machine learning - SVM ANN
Ensemble learningRandomForest LogiBoost ModelTree
t test 0.057 0.009 0.029 - - -
Accuracy (%) 138.6 141.1 147 151.4 152.3 156.3
Part 2 Summary
Encouraging data in each step: high-throughput screening, microfluidics, and machine learning methods.
This approach can be used for development of a diagnostic test for any infectious disease.
A ‘point-of-care’ (hand-held) diagnostic device is desirable for diagnosis, particularly of wild animals.
Oak Ridge National LaboratoryRobert Foote
AcknowledgementsUniversity of Tennessee Knoxville
Cathy ScottGraham HicklingStacey PattersonSteve OliverDilip Patel
University of PennsylvaniaRobert H. Whitlock
The Scripps Research InstituteIrwin Sherman
Supported by:
Robert FooteRobert ShawEdward Uberbacher #2007-3504-18462
United States Department of AgricultureJohn P. BannantineW. Ray Waters
Centers for Disease Control and Prevention
Patricia Wilkins
top related