Maximizing Allele Detection by Selecting
Optimal Analytical Thresholds
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
Christine A. Rakay
Joli Bregu
Cheng-Tsung Hu
Catherine M. Grgicak
American Academy of
Forensic Sciences February
2013
Steps During DNA Interpretation
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
Validation
studies &
Literature
Application of
Thresholds
Comparison to
Known(s)
Effect of AT’s on Data Analysis
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
2 males, 1:19 at 2ng, 1ul 3130 prep volume and 5s injection
Summary of Methods
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
ISHI Mixture Interpretation Workshop, 2012. http://www.cstl.nist.gov/strbase/mixture.htm
89% of respondents
use an AT between
50 and 150 RFU
Summary of Methods
Use data from negatives (i.e.
samples with no DNA)
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118




Use data from
DNA dilution
series


Method 1.
◦ Kaiser (IUPAC 1976)
 Winefordner 1983 and Krane 2007
Method 2.
◦ Currie (IUPAC 1995)
 Winefordner 1983
Method 3.
◦ Example in SWGDAM Guidelines
Method 4.
◦ Largest observed noise peak
Method 5.
◦ Miller & Miller. Statistics for Analytical Chemistry (Ellis
Horwood & Prentice Hall)
 IUPAC 1997 ElectroAnalytical Committee
Method 6.
◦ 1997 IUPAC ElectroAnalytical Committee Recommendations
J. Bregu et al. Analytical Thresholds and Sensitivity: Establishing RFU Thresholds for Forensic DNA Analysis. JFS (2013) 1
pg 120-129.
Method 1 to 4 - Negatives
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
-Negative sample run with an internal size standard (not shown) using manufacturer’s
recommended protocol
Negative = extraction or amplification negative
15
Baseline is never below 0 RFU
Processed data!
0
15
0
15
0
15
0
Green and Blue channels seem
‘quieter’ than yellow and red
Method 5 to 6 – Positives (Standard Curves)
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
Regression of positive samples (i.e. single source samples)
 Amplified 0.0625-4ng dilution series, injected 5s using manufacturer’s recommended protocol
 Plot of Input DNA (ng) versus average peak height (per color) – with error bars
◦ If a peak was homozygous, the RFU was divided by 2

•
•
•
The points at 2 and 4 ng fall off
the line (PCR efficiency
approaching a plateau)
The error bars become larger
with increased DNA input
A weighted linear regression is
within the linear range (i.e.
0.0625 – 1 ng) was used.
Summary of Results
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
Method
Origin
Analytical
Threshold for green
5s injection
example
1
Negatives
7
2
Negatives
4
3
Negatives
18
4
Negatives
6
5
DNA Series
31
6
DNA Series
39
False non-labeling of alleles (Drop-out)
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
Single source 0.125ng, 1ul 3130 prep volume
200
80
40
20
0
Drop-out with Respect to ATs - <0.5 ng DNA
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
Frequency of Drop-out
1
■ locus DO
■ allelic DO
■ sum (# loci exhibiting
DO)
0.8
0.6
freqDO(locus ) 
0.4
# hetloci (2alleleDO )
total # hetloci
freqDO(allele ) 
0.2
# hetloci (1alleleDO )
total # hetloci
0
0
50
100
150
200
Analytical Threhold (RFU)
-As AT increases, locus DO increases, while allele DO stabilizes after 50 RFU then starts to
decrease after AT of ~150 RFU.
-Although a higher AT (i.e. >150 RFU) begins to decrease the number of loci where allele DO
occurs (less stochastic variation),
-Locus DO increases, resulting in an overall increase in DO with AT for Low-template samples
Balancing Type I and Type II Errors – < 0.5ng
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
-AT’s have a large effect on the ability to detect/label alleles.
-Red = high level of allele drop-out, blue=low levels of allele drop-out.
- To take a ‘conservative’ approach and utilize high AT values leads to a substantial
level of Type II errors for low-level samples (i.e. <1000RFU).
Impact of ATs on STs
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
500 – 63 pg
Description of
Calculation
Frequency of
Allele Dropout
Peak Height of Largest
Surviving Allele at a locus
exhibiting allele drop-out
(RFU)
Frequency of
Locus Dropout
Negatives
Kaiser’s
0.025
126
0.000
Negatives
Kaiser’s
0.041
126
0.002
Negatives
Max observed
noise peak
0.168
170
0.028
Positives
IUPAC
0.230
240
0.074
50
150
200
N/A
N/A
N/A
0.246
0.246
0.184
229
349
548
0.074
0.593
0.816
Analytical
Threshold
(RFU)
N/A = not applicable
Baselines Positives ≠ Baselines Negatives
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
High
input of
DNA
30
0
30
Neg amp
control
0
30
Low
input of
DNA
0
More on Baselines and Noise
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
•
•
This is not instrument baseline/noise
Single source DNA data amplified from 0.0625 – 2 ng
• Differentiated ‘noise’ from artifact
• -A, pull-up, stutter (+ or -), spikes, dye artifacts
Plotted RFU of the known/expected peak versus the highest ‘noise’ peak
High noise with >0.5 ng of DNA, higher AT needed for higher-template samples
Mixture 1:9
Proportion of minor allele labled
•
•
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
AT= 50
ATM2
> 0.5 ng
<0.5 ng
0.0
0.1
0.2
0.3
Proportion of Loci with Noise >AT
0.4
Injection Times
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
Amplification positive sample (2 sec)
120
High-template
injection times
injection times
# of noise peaks
# of noise peaks
height noise peaks
height noise peaks
blue
100
Number of peaks
Low-template
green
80
yellow
60
red
40
20
0
<500
500-1000 1000-1500 1500-2000 2000-2500
AT
AT
Avg allele height (RFU)
Amplification positive samples (10 sec)
blue
green
yellow
red
200
180
160
140
120
100
80
60
40
20
0
blue
green
yellow
red
<500
500-1000
1000-1500
1500-2000
2000-2500
2500-3000
3000-3500
3500-4000
4000-4500
4500-5000
5000-5500
5500-6000
6000-6500
6500-7000
7000-7500
200
180
160
140
120
100
80
60
40
20
0
Number of peaks
Number of peaks
Amplification positive samples (5 sec)
Avg allele height (RFU)
Avg allele height (RFU)
Conclusions
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118
• Baseline does not remain constant between negatives and samples
with a significant amount of DNA
• There may be amplification ‘noise’ that cannot be characterized as
known artifact (i.e. bleed-through, spike, stutter, etc)
• Optimal ATs will be dependent on the DNA amplification mass
• Optimal AT for DNA samples amplified with < 0.5 ng was 10-20
RFU.
• An AT of 50 resulted in ~ 20% Type II error rate. An AT of
150 resulted in ~80% error rate.
• To minimize error for DNA samples amplified with > 0.5 ng the AT
needs to be increased by a factor of 2.5 - 5 (i.e. 50 RFU)
• Thresholds designed for/by samples containing optimal masses are not
optimal for low-template DNA interpretation
• Samples amplified with sub-optimal masses require special
interpretation schemes/methods
Acknowledgements
Boston University School of Medicine
Program in Biomedical Forensic Sciences
72 E. Concord Street, Boston, MA 02118

Thanks to the following Boston University BMFS students,
◦ Christine A. Rakay
◦ Joli Bregu
◦ Kevin Hu
◦ Thank-you
◦ Robin Cotton, Charlotte Word, Michael Coble, John Butler, Desmond Lun
◦ Supported by
◦ NIJ2008-DN-BX-K158 training grant awarded by the National Institute of Justice, Office
of Justice Programs, U.S. Department of Justice. The opinions, findings and
conclusions or recommendations expressed in this presentation are those of the
authors and do not necessarily reflect those of the Department of Justice
◦ Boston University, Biomedical Forensic Sciences Program
◦ [email protected]