Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Phenotype prediction accuracy – A Swedish perspective

Junker, Klara ; Staadig, Adam ; Sidstedt, Maja LU ; Tillmar, Andreas and Hedman, Johannes LU (2019) In Forensic Science International: Genetics Supplement Series 7(1). p.384-386
Abstract

Methods for SNP-based phenotype prediction have recently been developed, but prediction accuracy data for several populations and regions are missing. We analysed the accuracy of hair and eye colour predictions for 111 individuals residing in Sweden, using the ForenSeq system and the MiSeq FGx instrument (Verogen). Observed colours were compared to predicted colours, using the colour with the highest probability value for each prediction. Overall, 80% of eye colour predictions were correct, but the system failed to predict intermediate/green eye colour in our cohort. For hair colour, 58% of predictions were correct, and the majority of incorrect predictions were related to brown hair. To assess if prediction accuracy could be improved... (More)

Methods for SNP-based phenotype prediction have recently been developed, but prediction accuracy data for several populations and regions are missing. We analysed the accuracy of hair and eye colour predictions for 111 individuals residing in Sweden, using the ForenSeq system and the MiSeq FGx instrument (Verogen). Observed colours were compared to predicted colours, using the colour with the highest probability value for each prediction. Overall, 80% of eye colour predictions were correct, but the system failed to predict intermediate/green eye colour in our cohort. For hair colour, 58% of predictions were correct, and the majority of incorrect predictions were related to brown hair. To assess if prediction accuracy could be improved by the exclusion of predictions with low probabilities, we applied a threshold of ≥0.7. The threshold improved eye colour prediction, from 80% to 85% correct predictions, whereas hair colour prediction accuracy was virtually unaffected (58% versus 57% correct predictions). In summary, the phenotype prediction accuracy was acceptable in our cohort and the use of a threshold was only useful for eye colour predictions.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; and
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
EVC, Massively parallel sequencing, Phenotype, SNP
in
Forensic Science International: Genetics Supplement Series
volume
7
issue
1
pages
384 - 386
publisher
Elsevier
external identifiers
  • scopus:85073055664
ISSN
1875-1768
DOI
10.1016/j.fsigss.2019.10.022
language
English
LU publication?
yes
id
c343d0c9-e05f-4c32-b869-91ef0c0d18ad
date added to LUP
2019-10-25 12:19:30
date last changed
2022-04-18 18:30:05
@article{c343d0c9-e05f-4c32-b869-91ef0c0d18ad,
  abstract     = {{<p>Methods for SNP-based phenotype prediction have recently been developed, but prediction accuracy data for several populations and regions are missing. We analysed the accuracy of hair and eye colour predictions for 111 individuals residing in Sweden, using the ForenSeq system and the MiSeq FGx instrument (Verogen). Observed colours were compared to predicted colours, using the colour with the highest probability value for each prediction. Overall, 80% of eye colour predictions were correct, but the system failed to predict intermediate/green eye colour in our cohort. For hair colour, 58% of predictions were correct, and the majority of incorrect predictions were related to brown hair. To assess if prediction accuracy could be improved by the exclusion of predictions with low probabilities, we applied a threshold of ≥0.7. The threshold improved eye colour prediction, from 80% to 85% correct predictions, whereas hair colour prediction accuracy was virtually unaffected (58% versus 57% correct predictions). In summary, the phenotype prediction accuracy was acceptable in our cohort and the use of a threshold was only useful for eye colour predictions.</p>}},
  author       = {{Junker, Klara and Staadig, Adam and Sidstedt, Maja and Tillmar, Andreas and Hedman, Johannes}},
  issn         = {{1875-1768}},
  keywords     = {{EVC; Massively parallel sequencing; Phenotype; SNP}},
  language     = {{eng}},
  number       = {{1}},
  pages        = {{384--386}},
  publisher    = {{Elsevier}},
  series       = {{Forensic Science International: Genetics Supplement Series}},
  title        = {{Phenotype prediction accuracy – A Swedish perspective}},
  url          = {{http://dx.doi.org/10.1016/j.fsigss.2019.10.022}},
  doi          = {{10.1016/j.fsigss.2019.10.022}},
  volume       = {{7}},
  year         = {{2019}},
}