Statistical methods for discrimination of STR genotypes using high resolution melt curve data.

Area of Science:

• Forensic Science
• Molecular Biology
• Genetics

Background:

• Current forensic DNA quantification methods detect low template DNA but do not reveal stochastic effects until later stages.
• An earlier genotyping assay could enable critical adjustments before STR amplification, allowing for prompt reporting of exclusionary information.

Purpose of the Study:

• To evaluate the accuracy of Qiagen's principal component analysis (PCA)-based ScreenClust HRM software and a linear discriminant analysis (LDA)-based technique for predicting short tandem repeat (STR) genotypes from high-resolution melt (HRM) data.
• To assess the performance of these methods on STR loci D5S818 and D18S51.

Main Methods:

• Melt curves were generated from STR D5S818 and D18S51 amplicons using a Rotor-Gene Q qPCR instrument and EvaGreen dye.
• PCA-based (ScreenClust HRM) and LDA-based statistical prediction analyses were applied to HRM data to predict genotypes.
• Prediction accuracy was assessed for individual genotypes and genotype groups.

Main Results:

• LDA outperformed PCA for predicting D5S818 genotypes, achieving 58.92% accuracy for individual genotypes and 81.00% for genotype groups.
• For the more heterogeneous locus D18S51, PCA prediction accuracy improved, becoming comparable to LDA (45.10% and 63.46% accuracy, respectively).

Conclusions:

• This study provides foundational data on the performance of prediction modeling for STR genotyping using qPCR-HRM data.
• LDA demonstrates superior accuracy for genotype prediction from HRM data, particularly for less heterogeneous loci.
• Further testing on more common qPCR platforms is recommended to enhance the forensic applicability of HRM assays for genotyping.