Đánh giá các kỹ thuật xét nghiệm dùng trong dịch tễ học phân tử

Steps used to validate a new molecular strain typing test for epidemiologic investigation of infectious diseases Assess whether or not the Demonstrate that the typing information generat

National Institute of Infectious Disease

Validity of a new test:

• Sensitivity

• Specificity

Validity: ability of a test to correctly

predict or identify those who truly

have the characteristic the test is

trying to detect, and exclude those

who do not have the characteristic.

• In molecular epidemiology, a new test is validated by its ability to discriminate strains

that are epidemiologically related from

those that are not

Validity is determined by

comparison of the observed

results to a reference standard,

“truth”, “gold standard”.

In molecular epidemiology, validity is determined empirically

Steps used to validate a new molecular strain typing test for

epidemiologic investigation of infectious diseases

Assess whether or not the

Demonstrate that the typing information generated by the test is indistinguishable for all isolates from persons with disease in a recognized outbreak.

Select appropriate comparison isolates (geographic and temporal controls), and show that the typing information from these isolates is distinct from that of the outbreak isolates.

Steps used to validate a new test in molecular epidemiology—

cont.

• Ascertain fidelity of the typing

information used

• temporal stability of the taxonomic unit;

• clonality of the isolates obtained from a single host)

• Perform new analysis in

Steps used to validate a new test in molecular epidemiology—cont

If outbreak occurrence is uncertain:

Show that isolates that do not belong

to the clonal group

do not have the same epidemiologic association.

Ascertain fidelity of the typing

information.

If possible, eliminate the identified or putative risk factor and evaluate if this will lead to control

or amelioration of the problem.

For molecular epidemiology, a strain typing test that cannot yield any epidemiologically useful or meaningful information, no matter how simple, discriminating, or taxonomically relevant, is not valid!

Final test of validity of a molecular typing technique:

Lecture 4, Part 2:

Analysis of similarity and relatedness

Principles of Molecular Epidemiology

National Institute of Infectious Disease

January 16, 2017

 Describe cladistic vs phenetic methods of classifying microbes.

 Understand the appropriate applications of similarity coefficient calculations

to analyze patterns generated from strain-typing methods

 Describe different ways to measure reliability of the relationships portrayed

by a dendrogram

 Name different analytical tools needed to conduct molecular epidemiologic investigations

What can you do with this pattern?

All molecular techniques used to type organisms can be divided

into 3 general methods:

1) Direct comparison of nucleotide sequences.

2) Gel electrophoretic fingerprinting methods (e.g., REA, RFLP/Southern

blot hybridization, PFGE, and some PCR-generated patterns)

3) Hybridization matrix patterns (e.g., spoligotyping)

Questions about strain relatedness that arise in epidemiologic investigations

 Determining relatedness between two or more strains isolated from seemingly unrelated infected persons or contaminated sources

 Distinguishing strain typing data on the basis of variations within a range of such data (e.g., the number of bands in electrophoretic

patterns, or nucleotide substitutions in DNA sequences)

Questions about strain relatedness that arise in epidemiologic

 Identifying hidden groupings in a large collection of

strain typing data

 Selecting criteria for assigning a new pattern or

sequence into existing sets of strain typing data

Types of errors:

• Molecular epidemiology: chance of erroneously rejecting a subtyping assignment that concludes there is no epidemiologic relationship.

Type 1 probability error:

The chance of erroneously

rejecting a null hypothesis

that is in fact true

• Molecular epidemiology: chance of failing

to reject a subtyping assignment that concludes there is no epidemiologic relationship

Type 2 probability error:

The chance of erroneously

failing to reject a null

hypothesis that is indeed false

Similarity (or difference) analysis of patterns (electrophoretic, hybridization matrix) and

sequences all take into consideration methods that minimize these probability errors.

• Similarity or difference (distance) is compared between OTUs

Phenetic

methods

(numerical

taxonomy)

Phenetic methods of classification used in epidemiology

 Methods that measure similarity or difference (dissimilarity, distance) between individual OTUs.

 Clustering methods based on similarity (distance) index that identify patterns among a collection of OTUs

Epidemiologic applicability of the phenetic classification

methods

 For epidemiologic applications, the “correctness” of the OTU

assignment is ultimately based on how well this assignment explains and solves the epidemiologic problem posed

 The advantage of the phenetic methods applied to epidemiology is that for a given pathogen, the predictability of a classification scheme can be maximized and validated empirically through multiple

epidemiologic studies.

Similarity coefficient calculation:

Simple matching index:

Lane position

1 2 3 4 5 6 7 8 9 10

OTU-A OTU-B cell

Similarity calculations based on different indices for the above example

a = 5 (band present in both OTUs)

b = 1 (band present only in OTU-B)

c = 2 (band present only in OTU-A)

d = 2 (band absent from both OTUs)

(1.0 = identical)

 Simple matching index: S = 0.70

 Sokal and Sneath’s index: Sss = 0.82

 Jaccard index: SJ = 0.62

 Dice index or coefficient: SD = 0.77

Multiple OTUs

Which similarity coefficient to use?

Cluster analysis

Comparing relatedness among multiple strains (Cluster analysis)

 Nearest neighbor (also called single linkage clustering)

 Farthest neighbor (also called complete linkage

clustering)

averages (UPGMA) cluster analysis

 Ordination analysis : non-hierarchical cluster analysis

method (e.g., principal component analysis or PCA)

Algorithms for electrophoretic banding pattern analysis

•Jaccard or Dice coefficients

Comparison of band

positions (binary

character state)

•Pearson product-moment correlation coefficient

Comparison of width

or intensity of bands

(continuous character

state; curve-based)

Dice (Opt:2.00%) (Tol 2.0%-2.0%) (H>0.0% S>0.0%) [0.0%-100.0%]

G4 G7 G8 G11 G11 G4 G4 G6 G7 G9

8 8 8 9 10 10 10 10 10 10

D6 D11 D12 D15 D14 D6 D6 D8 D9 D13

A A A A A A A2 A A A

Dendrogram (phenogram) constructed from IS6110 RFLP analysis of

M tuberculosis isolates from Sao Paulo, Brazil (Ferrazoli et al): Dice

coefficient

Pearson correlation (Opt:2.00%) [0.0%-100.0%]

A A A A A A2 A A A A

Dendrogram (phenogram) constructed from IS6110 RFLP analysis of M

tuberculosis isolates from Sao Paulo, Brazil (Ferrazoli et al): Pearson

coefficient

Assessing reliability of relatedness measures

 Phenetic methods depict mathematical relationships that attempt to predict biological relationships (evolutionary or epidemiologic).

Evolutionary vs epidemiologic relatedness

 Based on a model or consensus definition

 Based on evolutionary relatedness data

 Based on empirically-validated data

Phylogenetic tree based on 1278 core genes of 186 E coli strains (Kaas et al, 2012)

Phylogenetic tree of E coli O157:H7 by their core

genes (Kaas RS et al, 2012)

“What we observe is not nature itself, but nature

exposed to our method of questioning.”

Assessing reliability of relatedness measures—cont.

 Their “true” relationship needs to be empirically

determined In some situations, this can be done (e.g., outbreaks) If this cannot be done (e.g., relationship of multiple nucleic acid sequences—alignments), the data points need to be examined for their reliability by a

stochastic method.

Measures of reliability of data points

 Resampling methods:

Data points from the original data set containing n data points

are randomly and repeatedly sampled until new sample sets,

each containing n points are created

Example of resampling of nucleotide sequences—bootstrapping:

OTU

A: 5 ’-atgggcgacttcatcacgatgaggtcaggaggccactatt ref

B: 5 ’-atgggctacttcttcacgatcaggtcaggaggccactatt

C: 5 ’-atcggcgacttcatcacgatgaggtgtggaggccactatt

D: 5 ’-aagggcgacttcatcaccatgaggtcaggaggccactata

E: 5 ’-atgggcgattttaccactttgaggtcaggtggccggtatt

F: 5 ’-atggcttgctttataacgattaggtgagaaggccactatt

G: 5 ’-cagggcgacttcatcttagcctggtcagcaggccacgatt

Dendrogram generated from the original data set

resampled

Resampling methods—cont.

 Degree of deviation from the original tree among the pseudosamples measures the reliability of the original tree If there is no deviation, then the original tree can be said to be unaffected by any stochastic effects.

Typical software requirements in a laboratory for database analysis involving a molecular epidemiology project

Study design methods Software

power, sample size calculation EpiInfo

questionnaire design EpiInfo

data entry, storage, line listing EpiInfo, Access., Excel, etc.

data analysis EpiInfo

Advanced statistical methods EpiInfo, STATA, SAS, SPSS, R

Capturing and storing pattern images tiff, jpeg, gif, etc

Image normalization, similarity/distance

and cluster analysis, storage, tree generation GelCompar, Molecular Analyst

Sequence alignment ClustalX

Command-line search programs used to analyze

high-throughput sequences

 Nucleic acid sequence

assembly to create contigs

Comparison Tool (ACT)

BRIG (reference comparison)

(phage genes)