Reverse transcription-quantitative PCR (RT-qPCR): Reporting and minimizing the uncertainty in data accuracy

Reverse transcription-quantitative PCR (RT-qPCR)

Reporting and minimizing the uncertainty in data accuracy

Ann CuypersEnvironmental BiologyCentre for Environmental SciencesHasselt UniversityBelgium

Hallo

Reverse transcription and quantitative (real-time) PCR

Gene expression analysis

Steady state mRNA levels

Highly sensitive to technical variation

Accuracy and precision depends on

Minimizing technical errors

Normalization to stably expressed reference genes

Remans et al. (2014) Plant Cell Commentary

2

Outline

MIQE guidelines

Selecting reference genes for RT-qPCR


3

MIQE guidelines

Uniform standard for reporting qPCR data

Bustin et al. (2010): Practical implementation of MIQE

International consortium of academic scientists

4

MIQE guidelines checklist

Sample/Template details Checklist

Source If cancer, was biopsy screened for adjacent normal tissue?

Method of preservation Liquid N2/RNAlater/formalin

Storage time (if appropriate) If using samples >6 months old

Handling fresh/frozen/formalin

Extraction method TriZol/columns

RNA: DNA-free Intron-spanning primers/no RT control

Concentration Nanodrop/ribogreen/microfluidics

RNA: integrity Microfluidics/3':5' assay

Inhibition-free Method of testing

Assay optimisation/validation

Accession number RefSeq XX_1234567

Amplicon details exon location, amplicon size

Primer sequence even if previously published

Probe sequence* identify LNA or other substitutions

In silico BLAST/Primer-BLAST/m-fold

empirical primer concentration/annealing temperature

Priming conditions oligo-dT/random/combination/target-specific

PCR efficiency dilution curve

Linear dynamic range spanning unknown targets

Limits of detection LOD detection/accurte quantification

Intra-assay variation copy numbers not Cq

RT/PCR

Protocols detailed description, concentrations, volumes

Reagents supplier, Lot number

Duplicate RT DCq

NTC Cq & melt curves

NAC DCq beginning:end of qPCR

Positive control inter-run calibrators

Data analysis

Specialist software e.g., QBAsePlus

Statistical justification e.g., biological replicates

Transparent, validated

normalisation e.g., GeNorm summary

5

Lack of adherence...

6

0.5

1

2

4

8

0 100 250 500

Rel

ati

ve

RB

OH

Fex

pre

ssio

n

Correct interpretation?

Normalized data

No further data available

7

-2

µM Zn

Outline

MIQE guidelines



8


Golden standard

Multiple reference genes

Validated minimal expression variation

Selection flowchart

Select genes to validate Different sources

Validate candidate reference genes Minimum 10 genes

Using the same cDNA as for GOI measurements

Apply evaluation algorithm (geNorm, Normfinder, GRAYNORM)

Revalidation of chosen reference genes Related or repeated experiments

9

Selection flowchart

10Remans et al (2014) Plant Cell Commentary

1. SELECT 2. VALIDATE

3. REVALIDATE

Outline

MIQE guidelines



11

Uncertainty in Data Accuracy

Origin?

Quantification

MinimizingA new algorithm for selecting reference genes: GrayNorm

Reporting

Histogram

Table

12

Uncertainty in data accuracy: origin

SAMPLE 1Control

SAMPLE 2Treated

Technical variation t1 t2SAMPLE-SPECIFIC

Gene of interest (GOI)

Reference gene (REF)

t1RQGOI

t1RQREF

Measurements:

t2RQGOI

t2RQREF

Normalization: t1/t1NRQGOI t2/t2NRQGOI

13


SAMPLE 1Control

SAMPLE 2Treated

Technical variation t1 = 1 t2 = 2SAMPLE-SPECIFIC



1RQGOI

1RQREF

Measurements:

2RQGOI

2RQREF

Normalization: 1NRQGOI 1NRQGOI

Reference genes correct for sample-specific technical variation

Example: RNA input for SAMPLE 1 = 1/2 RNA input for SAMPLE 2

14


SAMPLE 1Control

SAMPLE 2Treated

Technical variation t1 = 1 t2 = 2SAMPLE-SPECIFIC


Reference genes correct for sample-specific technical variation

ASSUMPTION: perfect reference genes

15

Example: RNA input for SAMPLE 1 = 1/2 RNA input for SAMPLE 2


SAMPLE 1Control

SAMPLE 2Treated

Biological variationin reference gene

b1 b2SAMPLE-SPECIFIC



RQGOI

b1RQREF

Measurements:

RQGOI

b2RQREF

Normalization: 1/b1NRQGOI 1/b2NRQGOI

Imperfect reference genes

16


Example: REF expression in SAMPLE 1 = 1/2 REF expression in SAMPLE 2

SAMPLE 1Control

SAMPLE 2Treated


b1 = 1 b2 = 2SAMPLE-SPECIFIC



RQGOI

1RQREF

Measurements:

RQGOI

2RQREF

Normalization: 1/1NRQGOI 1/2NRQGOI

Reference gene-specific biological variationis inversely imposed on GOI

17


SAMPLE 1Control

SAMPLE 2Treated


t1 t2

SAMPLE-SPECIFIC



t1RQGOI

t1b1RQREF

Measurements:

t2RQGOI

t2b2RQREF


Technical variation

b1 b2

18

Reference genes correct for technical variation,but impose biological variation on GOI


SAMPLE 1Control

SAMPLE 2Treated

Reference gene (REF) t1b1RQREF t2b2RQREF

Technical and biological variation

RQREF of near perfect reference genes

RQREF of experiment with high technical quality

Two possible assumptions

19


Assumption 1: perfect reference genes – no BIOLOGICAL variation

SAMPLE 1Control

SAMPLE 2Treated

Biological variationin reference geneBiological variationin reference gene

t1 t2

SAMPLE-SPECIFIC



t1RQGOI

t1b1RQREF

Measurements:

t2RQGOI

t2b2RQREF


Technical variation

b1 = b2 b2 = b1

Normalized data are accurate20


Assumption 2: perfect technical experiment – no TECHNICAL variation

SAMPLE 1Control

SAMPLE 2Treated

Biological variationin reference geneBiological variationin reference gene

t1 = t2 t2 = t1

SAMPLE-SPECIFIC



t1RQGOI

t1b1RQREF

Measurements:

t2RQGOI

t2b2RQREF


Technical variation

b1 b2

Non-normalized data are more accurate21

Uncertainty in data accuracy: quantification

The “truth” lies between normalized and non-normalized data

Normalized data: correction for technical variation

Non-normalized data: no biological variation is imposed

“Gene expression sensitivity” (GES) test

Statistics on normalized data

Statistics on non-normalized data

BOTH SHOULD BE SIGNIFICANT

22

Uncertainty in data accuracy: minimize!


Normalized data: correction for technical variation

Non-normalized data: no biological variation is imposed

Distance between normalized and non-normalized data

Uncertainty

Created by using the normalisation factor = 1/NF

GrayNorm algorithm: combination of reference genes with lowest deviation from 1 of 1/NF

23

0.5

1

2

4

8

0 100 250 500

Rel

ati

ve

RB

OH

Fex

pre

ssio

n

Uncertainty in data accuracy


24

0.5

1

2

4

8

0 100 250 500

Rel

ati

ve

RB

OH

Fex

pre

ssio

n

µM Zn

-2

µM Zn

-2

Normalized relative quantities Normalized relative quantities

Non-normalized relative quantities

Uncertainty in Data Accuracy

Origin?

Quantification

MinimizingA new algorithm for selecting reference genes: GrayNorm

Reporting

Histogram

Table

25

Data representation

Histogram of normalized and non-normalized data

Statistics on normalized data

Statistics on non-normalized data (sensitivity analysis)

Both should be significant!

26

0.5

1

2

4

8

0 100 250 500

Rel

ati

ve

RB

OH

Fex

pre

ssio

n

µM Zn

-2

*

*

*

**

Data representation

Table of normalized and non-normalized data

Supplement

Provide “resolution” values: 1/NF per condition

27

Time (h) Genotype 1 Genotype 2

RESOLUTION

0 2 2472

1.00 ± 0.140.46 ± 0.090.95 ± 0.111.26 ± 0.43

1.00 ± 0.080.79 ± 0.170.90 ± 0.270.65 ± 0.21

GOI

0 2 2472

1.00 ± 0.090.34 ± 0.050.29 ± 0.040.45 ± 0.16

1.00 ± 0.090.44 ± 0.030.33 ± 0.090.68 ± 0.18

28