Statistical modelling of Ion PGM HID STR 10-plex MPS data.

Area of Science:

• Forensic Genetics
• Molecular Biology
• Bioinformatics

Background:

• Massively parallel sequencing (MPS) platforms offer potential advantages for short tandem repeat (STR) analysis.
• Accurate STR profiling relies on understanding and mitigating potential sources of error, including substitution errors, heterozygote imbalance, and stutter.
• Traditional capillary electrophoresis (CE) methods have established quality metrics, necessitating comparable evaluations for MPS-based STR analysis.

Purpose of the Study:

• To evaluate the performance of the Ion PGM platform with the HID STR 10-plex panel for STR marker analysis.
• To investigate quality measures for identifying substitution errors and assess heterozygote balance compared to CE.
• To examine coverage stability, the impact of IonExpress Barcode adapter (IBA) sampling, and the predictive power of allele length and longest uninterrupted stretch (LUS) on stutter and shoulder ratios.

Main Methods:

• Analysis of dilution series and reference profiles using the Ion PGM MPS platform and HID STR 10-plex panel.
• Investigation of quality measures, heterozygote balance, coverage stability, and IonExpress Barcode adapter (IBA) sampling effects.
• Statistical modeling to assess parental longest uninterrupted stretch (LUS) and parental allele length as predictors of stutter and shoulder ratios.
• Application of dynamic thresholds using one-inflated negative binomial (OINB) and geometric models for non-systematic error removal.

Main Results:

• Average marker coverage was not a reliable predictor of template DNA amount due to multiplex identifier (MID) sampling.
• Parental LUS demonstrated superior predictive power for stutter ratio compared to parental allele repeat length, particularly for complex STR markers (e.g., TH01 R² of 0.78 vs. 0.02).
• The OINB and geometric models effectively reduced non-systematic noise, leaving an average of 1.8 and 1.2 systematic errors per STR system, respectively.

Conclusions:

• The Ion PGM platform with the HID STR 10-plex panel can be utilized for STR analysis, but coverage metrics require careful interpretation.
• Parental LUS is a more robust predictor of stutter ratio than parental allele length, especially for challenging STR loci.
• Advanced statistical models effectively mitigate non-systematic errors, enhancing the reliability of MPS-based STR profiling.