Introduction to LRmix Studio

Introduction to LRmix Studio
Oskar Hansson
Statistical methods in forensic genetics
6-7 June 2016, Bologna
EUROFORGEN-NoE is funded by the European Commission
within the 7th Framework Programme
What is LRmix Studio?
• Expert system dedicated to the interpretation
of (complex) forensic DNA mixtures
• Programmed after the likelihood ratio model:
What is LRmix Studio?
• Accommodates for drop-out and drop-in
• Estimates the drop-out rate from the available
data and use this to generate LRs
• Free of charge and open-source (Java)
• Download: http://lrmixstudio.org/download/
– Note: zip-file must first be unpacked into a folder
LRmix Studio – Features
• Computes likelihood ratios for DNA profiles
characterised with autosomal STR kits
• Compares any number of replicates to any number of
reference profiles (tested with up to 3 reference
profiles and 5 replicates)
• Tested for propositions with up to 4 unknowns and a
total of 5 contributors
LRmix Studio – Limitations
• The hypothesized contributors under Hp and Hd
are assumed to be unrelated to each other
• An unknown can be related to the suspect or
victim under Hd
• Uncommon and/or untested scenarios may lead
to unreliable results
• Cannot handle missing data in reference profiles
• Does not deconvolute mixtures
– EuroForMix can be used for this http://euroformix.com/
LRmix Studio graphical user interface
Getting help
Links to a local copy of the manual
Opens up the ‘About’ tab
LRmix Studio user manual
1. What is LRmix Studio?
2. Features
3. Tutorial
3.1 Import sample profiles
3.2 Import or add reference profiles
3.3 Profile summary
3.4 Analysis
3.5 Sensitivity analysis
3.6 Non-contributor tests
3.7 Automatic report generation
4. How to report bugs
5. Join the LRmix Studio user community
References
About LRmix Studio
Software version
Contact: [email protected]
License agreement
Import sample profiles
Tabs are inactivated to
enforce the workflow
Load a previously
analysed case
Manually add
profiles
Clear data and
start fresh
Load the evidence
profile(s)
The sample file
LRmix Studio accept both:
SampleName|Marker|Allele1|Allele2
and:
Sample Name|Marker|Allele 1|Allele 2
Important to use identical marker
names in all files (sample/ref/db)
Not case-sensitive:
Pentad=PentaD=PenTaD=PENTAD
Is sensitive to spaces:
Penta D≠PentaD
Samples and/or
replicates are
stacked
Notepad++
- Advanced non-destructive text editor
LibreOffice Calc (alternatively: OpenOffice Calc)
- Free open-source MS Excel clone
The sample file
LRmix Studio also accept this alternative
GeneMapper sample file format
Sample File
frag_013_G01.fsa
frag_014_G02.fsa
frag_013_G03.fsa
Sample Name
rep.1
rep.2
rep.3
D10S1248
12, 13, 14, , , ,
12, 13, , , , ,
12, 13, , , , ,
Height
220, 203, 66, , , , ,
339, 357, , , , , ,
353, 175, , , , , ,
vWA
16, 19, , , , ,
16, 19, , , , ,
16, 19, , , , ,
Height
223, 387, , , , , ,
139, 352, , , , , ,
247, 327, , , , , ,
D16S539
8, 9, 10, 11, 14, ,
8, 11, 14, , , ,
8, 11, , , , ,
Height
242, 73, 55, 342, 52, , ,
169, 175, 64, , , , ,
218, 193, , , , , ,
Import sample profiles
Next tab activated
A List of all loaded
samples/replicates
The folder from which
the file was loaded
List of markers
Profiles for each
sample/replicate
Import reference profiles
The list of loci is hard-coded,
it’s not derived from the
allele frequency table
Load reference
profile(s)
Homozygotes
Add a profile manually
The reference can be
saved as a csv file
Possible spurious
alleles or additional
contributors
Detects allele
drop-out
Detects e.g. parent
child relations
Profile Summary
Customisable
highlighting
The filters can be
printed
Determine the
number of
contributors
Analysis
Dropout
probability per
contributor
Number of
unknown
contributors
Relatedness
Alleles not in freq
file is assign this
Allele
frequency file
Drop-in
parameter
Threads
(performance)
Theta correction
(Fst or θ)
LR per locus
Final likelihood ratio
Contributors and non-contributors
LRmix Studio manual – “Only profiles relevant to the LR analysis should be
uploaded. If some of the reference profiles are uploaded but later not
used in the analysis tab, they will be considered as non-contributors. Noncontributor profiles can influence the likelihood ratio calculations if the
Fst (θ) correction is not nil.”
Sensitivity Analysis
Vary dropout for
contributors
Drop-out range
Drop-in
parameter
Theta correction
(Fst or θ)
Delete selected
curve
Drop-out Estimation
Vary dropout for
contributors
Number of
iterations
Drop-out range
Drop-in
parameter
Estimated dropout
range is shaded in
the plot
Non-contributor Test
Number of
random persons
to test
Difference
between the
person of interest
and the ‘best’
random person
Check the person that
will be replaced by a
random person
NB! Drop-out and drop-in
values must be > 0
NB! Cannot be used if
relatedness is used
“Performance tests may improve understanding for those unfamiliar with LRs, but they have no
direct bearing on the strength of evidence in a specific case.” - Steele and Balding
Reports
Interpretation and reporting
Other References
Replicate analyses